Skip to Content

Rule

Medicare Program; Comprehensive Care for Joint Replacement Payment Model for Acute Care Hospitals Furnishing Lower Extremity Joint Replacement Services

Document Details

Information about this document as published in the Federal Register.

Published Document

This document has been published in the Federal Register. Use the PDF linked in the document sidebar for the official electronic format.

Start Preamble Start Printed Page 73274

AGENCY:

Centers for Medicare & Medicaid Services (CMS), HHS.

ACTION:

Final rule.

SUMMARY:

This final rule implements a new Medicare Part A and B payment model under section 1115A of the Social Security Act, called the Comprehensive Care for Joint Replacement (CJR) model, in which acute care hospitals in certain selected geographic areas will receive retrospective bundled payments for episodes of care for lower extremity joint replacement (LEJR) or reattachment of a lower extremity. All related care within 90 days of hospital discharge from the joint replacement procedure will be included in the episode of care. We believe this model will further our goals in improving the efficiency and quality of care for Medicare beneficiaries with these common medical procedures.

DATES:

These regulations are effective on January 15, 2016, and applicable on April 1, 2016 when the first model performance period begins.

Start Further Info

FOR FURTHER INFORMATION CONTACT:

Claire Schreiber, Claire.Schreiber@cms.hhs.gov, 410 786 8939.

Gabriel Scott, Gabriel.Scott@cms.hhs.gov, 410 786 3928.

End Further Info End Preamble Start Supplemental Information

SUPPLEMENTARY INFORMATION:

Electronic Access

This Federal Register document is also available from the Federal Register online database through Federal Digital System (FDsys), a service of the U.S. Government Printing Office. This database can be accessed via the internet at http://www.gpo.gov/​fdsys/​.

Alphabetical List of Acronyms

Because of the many terms to which we refer by acronym, abbreviation, or short form in this final rule, we are listing the acronyms, abbreviations and short forms used and their corresponding terms in alphabetical order.

µSA Micropolitan Statistical Area

ACE Acute Care Episode

ACO Accountable Care Organization

APM Alternative Payment Model

ASC Ambulatory Surgical Center

ASPE Assistant Secretary for Planning and Evaluation

BPCI Bundled Payments for Care Improvement

CAH Critical Access Hospital

CBSA Core-Based Statistical Area

CCN CMS Certification Number

CFR Code of Federal Regulations

CJR Comprehensive Care for Joint Replacement

CMHC Community Mental Health Center

CMI Case Mix Index

CMMI Center for Medicare and Medicaid Innovation

CMP Civil Monetary Penalty

CMS Centers for Medicare & Medicaid Services

CoPs Conditions of Participation

CPCi Comprehensive Primary Care Initiative

CPT Current Procedural Terminology

CSA Combined Statistical Area

DME Durable Medical Equipment

DMEPOS Durable Medical Equipment, Prosthetics, Orthotics, and Supplies

eCQM Electronic Clinical Quality Measures

EFT Electronic funds transfer

ESRD End-Stage Renal Disease

FFS Fee-for-service

GAAP Generally Accepted Accounting Principles

GEM General Equivalence Mapping

GPCI Geographic Practice Cost Index

HAC Hospital-Acquired Condition

HACRP Hospital-Acquired Condition Reduction Program

HCAHPS Hospital Consumer Assessment of Healthcare Providers and Systems

HCC Hierarchical Condition Category

HCPCS Healthcare Common Procedure Coding System

HHA Home health agency

HHPPS Home Health Prospective Payment System

HHRG Home Health Resource Group

HHVBP Home Health Value-Based Purchasing

HIT Health Information Technology

HIQR Hospital Inpatient Quality Reporting

HLMR HCAHPS Linear Mean Roll Up

HOOS Hip Dysfunction and Osteoarthritis Outcome Score

HOPD Hospital outpatient department

HRR Hospital Referral Region

HRRP Hospital Readmissions Reductions Program

HVBP Hospital Value Based Purchasing Program

ICD-9-CM International Classification of Diseases, 9th Revision, Clinical Modification

ICD-10-CM International Classification of Diseases, 10th Revision, Clinical Modification

IPPS Inpatient Prospective Payment System

IPF Inpatient psychiatric facility

IRF Inpatient rehabilitation facility

KOOS Knee Injury and Osteoarthritis Outcome Score

LEJR Lower extremity joint replacement

LOS Length of stay

LTCH Long term care hospital

LUPA Low Utilization Payment Adjustment

MAC Medicare Administrative Contractor

MACRA Medicare Access and Chip Reauthorization Act of 2015

MAPCP Multi-Payer Advanced Primary Care Practice model

MCC Major Complications or Comorbidities

MCCM Medicare Care Choices Model

MDH Medicare-Dependent Hospital

MedPAC Medicare Payment Advisory Commission

MIPS Merit-based Incentive Payment System

MP Malpractice

MPFS Medicare Physician Fee Schedule

MSA Metropolitan Statistical Area

MS-DRG Medical Severity Diagnosis-Related Group

NPI National Provider Identifier

NPP Nonphysician Practitioner

NPRA Net Payment Reconciliation Amount

NQF National Quality Forum

OCM Oncology Care Model

OPPS Outpatient Prospective Payment System

PAC Post-Acute Care

PBPM Per Beneficiary Per Month

PE Practice Expense

PGP Physician Group Practice

PHA Partial hip arthroplasty

PPS Prospective Payment System

PRO Patient-Reported Outcome

PROMIS Patient-Reported Outcomes Measurement Information Systems

PRO-PM Patient-Reported Outcome Performance Measure

QIO Quality Improvement Organization

RAC Recovery Audit Contractor

RRC Rural Referral Center

RSCR Risk-Standardized Complication Rate

RSRR Risk-Standardized Readmission Rate

RVU Relative Value Unit

SCH Sole Community Hospital

SNF Skilled nursing facility

THA Total hip arthroplasty

TIN Taxpayer identification number

TKA Total knee arthroplasty

TP Target price

VR-12 Veterans Rand 12 Item Health Survey

Table of Contents

I. Executive Summary

A. Purpose

B. Summary of the Major Provisions

1. Model Overview: LEJR Episodes of Care

2. Model Scope

3. Payment

4. Similar, Previous, and Concurrent Models

5. Overlap With Ongoing CMS Efforts

6. Quality Measures and Reporting Requirements

7. Data Sharing Process

8. Beneficiary Protections

9. Financial Arrangements and Program Policy Waivers

C. Summary of Economic Effects

II. Background

A. General Background

B. Acronym of This Model

C. Public Comments Received in Response to the CJR Proposed Rule

III. Provisions of the Proposed Regulations and Analysis of and Responses to Public CommentsStart Printed Page 73275

A. Definition of the Episode Initiator and Selected Geographic Areas

1. Background

2. Definition of Episode Initiator

3. Financial Responsibility for the Episode of Care

4. Geographic Unit of Selection and Exclusion of Selected Hospitals

a. Overview and Options for Geographic Area Selection

b. MSA Selection Methodology

(1) Exclusion of Certain MSAs

(2) Selection Strata

(a) MSA Average Wage-Adjusted Historic LEJR Episode Payments

(b) MSA Population Size

(c) Analysis of Strata

(3) Factors Considered But Not Used in Creating Proposed Strata

(4) Sample Size Calculations and the Number of Selected MSAs

(5) Method of Selecting MSAs

B. Episode Definition for the CJR Model

1. Background

2. Clinical Dimension of Episodes of Care

a. Definition of the Clinical Conditions Included in the Episode

b. Definition of Related Services Included in the Episode

3. Duration of Episodes of Care

a. Beginning the Episode and Beneficiary Care Inclusion Criteria

b. Middle of the Episode

c. End of the Episode

C. Methodology for Setting Episode Prices and Paying Model Participants Under the CJR Model

1. Background

2. Performance Years, Retrospective Episode Payment, and Two-Sided Risk Model

a. Performance Period

b. Retrospective Payment Methodology

c. Two-Sided Risk Model

3. Adjustments to Payments Included in Episode

a. Treatment of Special Payment Provisions Under Existing Medicare Payment Systems

b. Treatment of Payment for Services That Extend Beyond the Episode

c. Pricing Adjustment for High Payment Episodes

4. Episode Price Setting Methodology

a. Overview

b. Pricing Features

(1) Different Target Prices for Episodes Anchored by MS-DRG 469 vs. MS-DRG 470

(2) Three Years of Historical Data

(3) Trending of Historical Data to the Most Recent Year of the Three

(4) Update Historical Episode Payments for Ongoing Payment System Updates

(a) Inpatient Acute Services Update Factor

(b) Physician Services Update Factor

(c) IRF Services Update Factor

(d) SNF Services Update Factor

(e) HHA Services Update Factor

(f) Other Services Update Factor

(5) Blend Hospital-Specific and Regional Historical Data

(6) Define Regions as U.S. Census Divisions

(7) Normalize for Provider-Specific Wage Adjustment Variations

(8) Combination of CJR Episodes Anchored by MS-DRGs 469 and 470

(9) Discount Factor

c. Approach To Combine Pricing Features

5. Use of Quality Performance in the Payment Methodology

a. Background

b. Implementation of Quality Measures for in the Payment Methodology

(1) General Selection of Quality Measures

(2) Adjustment to the Payment Methodology for Voluntary Submission of Data for Patient-Reported Outcome Measure

(3) Measure Risk-Adjustment and Calculations

(4) Applicable Time Period

(5) Criteria for Applicable Hospitals and Performance Scoring

(a) Identification of Applicable Hospitals for the CJR Model

(b) Methodology to Determine Performance on the Quality Measures

(c) Methodology to Link Quality and Payment

(i) Background

(ii) Alternatives Considered To Link Quality and Payment

(iii) Threshold Methodology and Final Policy to Link Quality and Payment

6. Process for Reconciliation

a. Net Payment Reconciliation Amount

b. Payment Reconciliation

7. Adjustments for Overlaps With Other CMMI Models and CMS Programs

a. Overview

b. CJR Beneficiary Overlap With BPCI Episodes

c. Accounting for CJR Reconciliation Payments and Repayments in Other Models and Programs

d. Accounting for Per Beneficiary Per Month (PBPM) Payments in the Episode Definition

e. Accounting for Overlap With Other Medicare Initiatives Involving Shared Savings and Total Cost of Care Models

8. Limits or Adjustments to Hospital Financial Responsibility

a. Overview

b. Limit on Raw NPRA Contribution to Repayment Amounts and Reconciliation Payments

(1) Limit on Raw NPRA Contribution to Repayment Amounts

(2) Limit on Raw NPRA Contribution To Reconciliation Payments

c. Policies for Certain Hospitals To Further Limit Repayment Responsibility

d. Hospital Responsibility for Increased Post-Episode Payments

9. Appeal Procedures

a. Payment Processes

b. Calculation Error

c. Dispute Resolution

(1) Limitations on Review

(2) Matters Subject To Dispute Resolution

(3) Dispute Resolution Process

10. Financial Arrangements and Beneficiary Incentives

a. Financial Arrangements

(1) CJR Sharing Arrangement Requirements

(2) Participation Agreement Requirements

(3) Gainsharing Payment and Alignment Payment Conditions and Restrictions

(4) Documentation and Maintenance of Records

b. Beneficiary Incentives Under the CJR Model

11. Waivers of Medicare Program Rules

a. Overview

b. Post-Discharge Home Visits

c. Billing and Payment for Telehealth Services

d. SNF 3-Day Rule

e. Waivers of Medicare Program Rules to Allow Reconciliation Payment or Repayment Actions Resulting From the Net Payment Reconciliation Amount

12. Enforcement Mechanisms

D. Quality Measures and Display of Quality Metrics Used in the CJR Model

1. Background

a. Purpose of Quality Measures in the CJR Model

b. Public Display of Quality Measures in the CJR Model

2. Quality Measures for Performance Year 1 (CY 2016) and Subsequent Years

a. Hospital-Level Risk-Standardized Complication Rate (RSCR) Following Elective Primary Total Hip Arthroplasty (THA) and/or Total Knee Arthroplasty (TKA) (NQF #1550)

(1) Background

(2) Data Sources

(3) Cohort

(4) Inclusion and Exclusion Criteria

(5) Risk-Adjustment

(6) Calculating the Risk-Standardized Complication Rate and Performance Period

b. Hospital-Level 30-Day, All-Cause Risk-Standardized Readmission Rate (RSRR) Following Elective Primary Total Hip Arthroplasty (THA) and/or Total Knee Arthroplasty (TKA) (NQF #1551)

(1) Background

(2) Data Sources

(3) Cohort

(4) Inclusion and Exclusion Criteria

(5) Risk-Adjustment

(6) Calculating the Risk-Standardized Readmission Rate and Performance Period

c. Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) Survey

(1) Background

(2) Data Sources

(3) Cohort

(4) Inclusion and Exclusion Criteria

(5) Case-Mix-Adjustment

(6) HCAHPS Scoring

(7) Performance Period

d. Applicable Time Period

3. Possible New Outcomes for Future Measures

a. Hospital-Level Performance Measure(s) of Patient-Reported Outcomes Following Elective Primary Total Hip and/or Total Knee Arthroplasty

(1) Background

(2) Data Sources

(3) Cohort

(4) Inclusion and Exclusion Criteria

(5) Outcome

(6) Risk-Adjustment (If Applicable)

(7) Calculating the Risk-Standardized Rate

(8) Performance Period

(9) Requirements for Successful Submission of THA/TKA Voluntary Data

b. Measure That Captures Shared Decision-Making Related to Elective Primary Total Hip and/or Total Knee ArthroplastyStart Printed Page 73276

c. Future Measures Around Care Planning

d. Future Measures for Use of Health IT and Health Enforcement Exchange

4. Form, Manner and Timing of Quality Measure Data Submission

5. Display of Quality Measures and Availability of Information for the Public From the CJR Model

E. Data Sharing

1. Overview

2. Beneficiary Claims Data

3. Aggregate Regional Data

4. Timing and Period of Baseline Data

5. Frequency and Period of Claims Data Updates for Sharing Beneficiary-Identifiable Claims Data During the Performance Period

6. Legal Permission To Share Beneficiary-Identifiable Data

F. Monitoring and Beneficiary Protection

1. Introduction and Summary

2. Beneficiary Choice and Beneficiary Notification

3. Monitoring for Access to Care

4. Monitoring for Quality of Care

5. Monitoring for Delayed Care

G. Coordination With Other Agencies

IV. Evaluation Approach

A. Background

B. Design and Evaluation Methods

C. Data Collection Methods

D. Key Evaluation Research Questions

E. Evaluation Period and Anticipated Reports

V. Collection of Information Requirements

VI. Regulatory Impact Analysis

A. Statement of Need

B. Overall Impact

C. Anticipated Effects

1. Overall Magnitude of the Model and Its Effects on the Market

2. Effects on the Medicare Program

a. Assumptions and Uncertainties

b. Analyses

c. Further Consideration

3. Effects on Beneficiaries

4. Effects on Small Entities

5. Effects on Small Rural Hospitals

6. Unfunded Mandates

D. Alternatives

E. Accounting Statement

F. Conclusion

Regulations Text

I. Executive Summary

A. Purpose

The purpose of this final rule is to implement a new payment model called the Comprehensive Care for Joint Replacement (CJR) model under the authority of the Center for Medicare and Medicaid Innovation (CMMI). Section 1115A of the Social Security Act (the Act) authorizes CMMI to test innovative payment and service delivery models to reduce program expenditures while preserving or enhancing the quality of care furnished to Medicare, Medicaid, and Children's Health Insurance Program beneficiaries. The intent of the CJR model is to promote quality and financial accountability for episodes of care surrounding a lower-extremity joint replacement (LEJR) or reattachment of a lower extremity procedure.[1] CJR will test whether bundled payments to acute care hospitals for LEJR episodes of care will reduce Medicare expenditures while preserving or enhancing the quality of care for Medicare beneficiaries. We anticipate the CJR model will benefit Medicare beneficiaries by improving the coordination and transition of care, improving the coordination of items and services paid for through Medicare Fee-For-Service (FFS), encouraging more provider investment in infrastructure and redesigned care processes for higher quality and more efficient service delivery, and incentivizing higher value care across the inpatient and post-acute care (PAC) spectrum spanning the episode of care. We will test the CJR model for 5 performance periods, beginning April 1, 2016, and ending December 31, 2020. Under FFS, Medicare makes separate payments to providers and suppliers for the items and services furnished to a beneficiary over the course of treatment (an episode of care). With the amount of payments dependent on the volume of services delivered, providers may not have incentives to invest in quality improvement and care coordination activities. As a result, care may be fragmented, unnecessary, or duplicative.

We have previously used our statutory authority under section 1115A of the Act to test bundled payment models such as the Bundled Payments for Care Improvement (BPCI) initiative. Bundled payments, for multiple services in an episode of care, hold participating organizations financially accountable for an episode of care. They also allow participants to receive payment, in part, based on the reduction in expenditures for Medicare arising from their care redesign efforts.

We believe the CJR model will further the mission of CMMI and the Secretary's goal of increasingly paying for value rather than for volume,[2] because it will promote the alignment of financial and other incentives for all health care providers and suppliers caring for a beneficiary during an LEJR episode. In the CJR model, the acute care hospital that is the site of surgery will be held accountable for spending during the episode of care. Participant hospitals will be afforded the opportunity to earn performance-based payments by appropriately reducing expenditures and meeting certain quality metrics. They will also gain access to data and educational resources to better understand LEJR patients' PAC needs and associated spending. Payment approaches that reward providers that assume financial and performance accountability for a particular episode of care create incentives for the implementation and coordination of care redesign between hospitals and other providers and suppliers.

The CJR model requires the participation of hospitals in multiple geographic areas that might not otherwise participate in the testing of bundled payments for episodes of care for LEJR procedures. Other episode-based, bundled payment models being tested by the Centers for Medicare & Medicaid Services (CMS), such as the BPCI initiative, are voluntary in nature. Interested participants must apply to such models to participate. To date, we have not tested an episode payment model with bundled payments in which providers are required to participate. We recognize that realizing the full potential of new payment models will require the engagement of an even broader set of providers than have participated to date, providers who may only be reached when new payment models are applied to an entire class of providers of a service. As such, we are interested in testing and evaluating the impact of a bundled payment approach for LEJR procedures in a variety of circumstances, especially among those hospitals that may not otherwise participate in such a test.

This model will allow CMS to gain experience with making bundled payments to hospitals who have a variety of historic utilization patterns; different roles within their local markets; various volumes of services; different levels of access to financial, community, or other resources; and various levels of population and health provider density including local variations in the availability and use of different categories of PAC providers. We believe that by requiring the participation of a large number of hospitals with diverse characteristics, the CJR model will result in a robust data set for evaluation of this bundled payment approach, and will stimulate the rapid development of new evidence-based knowledge. Testing the model in this manner will also allow us to learn more about patterns of inefficient utilization of health care services and how to incentivize the improvement of Start Printed Page 73277quality for common LEJR procedure episodes. This learning potentially could inform future Medicare payment policy.

This final rule implements a model focused on episodes of care for LEJR procedures. We chose LEJR episodes for the CJR model because as discussed in depth in section III.C. of this final rule, these are high-expenditure, high utilization procedures commonly furnished to Medicare beneficiaries,[3] where significant variation in spending for procedures is currently observed. The high volume of episodes and variation in spending for LEJR procedures create a significant opportunity to test and evaluate the CJR model that specifically focuses on a defined set of procedures. Moreover, there is substantial regional variation in PAC referral patterns and the intensity of PAC provided for LEJR patients, thus resulting in significant variation in PAC expenditures across LEJR episodes initiated at different hospitals. The CJR model will enable hospitals to consider the most appropriate PAC for their LEJR patients. The CJR model additionally will offer hospitals the opportunity to better understand their own processes with regard to LEJR, as well as the processes of post-acute providers. Finally, while many LEJR procedures are planned, the CJR model will provide a useful opportunity to identify efficiencies both for when providers can plan for LEJR procedures and for when the procedure must be performed urgently.

The following is a summary of the comments received on the proposed model as a whole, including the authority for the model and general comments on CMS' implementation of the CJR model at this time and our responses.

Comment: A commenter stated that while the proposed rule emphasized the learning CMS hoped to gain from implementing and testing the CJR model, it made inadequate mention of the potential benefits to beneficiaries, providers, hospitals, and other stakeholders. Other commenters contended that bundled payment models encourage hospitals to engage in care stinting and potentially stifle innovation.

Response: We appreciate the commenters' concerns. We refer readers to section III.F. of this final rule for discussion of monitoring and beneficiary protections under this model which we believe will address the commenters' concerns about care stinting. We expect that the CJR model will benefit not just CMS, but also beneficiaries, hospitals, and other providers in the health care system. The goals of this model are to improve the quality of care furnished to beneficiaries and reduce spending during LEJR episodes. Beneficiaries would directly benefit from improved care coordination and care redesign activities that reduce readmissions and complications rates, for example, as well as provide an improved care experience during the inpatient hospitalization and post-discharge period. Hospitals also stand to benefit from the CJR model, in the form of the opportunity to earn reconciliation payments if successful under the model, and a structured incentive to redesign care processes for beneficiaries receiving LEJR procedures. For example, section III.C.11. of this final rule details waivers of Medicare program rules that would allow hospitals to test additional ways to introduce flexibility into care processes and improve the quality of care for beneficiaries. In addition, providers and suppliers across the spectrum of care provided during an LEJR episode could also benefit from the care redesign strategies as well as the financial arrangements as detailed in section III.C.10. of this final rule. Finally, we disagree with commenters that the CJR model will stifle innovation for care furnished during an LEJR episode. We proposed, and are finalizing in this final rule, a payment methodology that will account for changes in care patterns and utilization trends for LEJR episodes by updating the historical performance periods used throughout the model, as described in section III.C.4. of this final rule. In addition, the CJR financial incentives would be consistent with clinical practices that result in reductions of spending during LEJR episodes, allowing hospitals that engage in such practices to earn reconciliation payments and engage with other providers furnishing services during the episode, as discussed in section III.C.10. of this final rule.

Comment: Several commenters questioned CMS' legal authority to require participation in a model. Commenters stated that CMS lacks the legal authority to compel participation in a model, and that CMS misreads section 1115A(a)(5) of the Act as the legal basis for compelling providers in selected Metropolitan Statistical Area (MSAs) to participate in the CJR model. A commenter stated that language in the Act has never been interpreted to afford the Secretary the authority to compel provider participation in a Medicare demonstration project or model, and that the Congress intended for model tests to be voluntary, not mandatory, when authorizing CMS to test new models. The commenter noted that requiring providers to participate in a model that would encompass a substantial proportion of a particular service would render the statutory distinction between testing and expanding models meaningless. The commenter also expressed concern about the model's potential effect on beneficiaries' appeal rights. Several commenters stated that CMS is sidestepping the legal safeguards designed to prevent the Agency from imposing novel or haphazard models on providers prior to adequate testing and evaluation. Commenters also claimed that CMS had exceeded its statutory authority because under section 1115A of the Act, providers are precluded from appealing their selection in a model, raising further concern that CMS is overreaching by requiring participation in the CJR model. Commenters also noted that there is no precedent for a CMS demonstration or model that requires providers to participate. Finally, several commenters stated that CMS has reversed the intended sequence of testing and then expanding models.

Response: We disagree with commenters that we lack the legal authority to test the CJR model as proposed and specifically, to require the participation of selected hospitals. We note that although CJR will be the first Innovation Center model in which acute care hospitals are required to participate, we refer readers to the 2016 Home Health Prospective Payment System (HHPS) Final Rule, which finalizes the Home Health Value-Based Purchasing (HHVBP) model. Home health agencies in selected states will be required to participate in the HHVBP model beginning in January 2016.

We believe that both section 1115A and the Secretary's existing authority to operate the Medicare program authorize the CJR model as we have proposed and are finalizing it. Section 1115A of the Act authorizes the Secretary to test payment and service delivery models intended to reduce Medicare costs while preserving quality. The statute does not Start Printed Page 73278require that models be voluntary, but rather gives the Secretary broad discretion to design and test models that meet certain requirements as to spending and quality. Although section 1115A(b) of the Act describes a number of payment and service delivery models that the Secretary may choose to test, the Secretary is not limited to those models. Rather, models to be tested under section 1115A of the Act must address a defined population for which there are either deficits in care leading to poor clinical outcomes or potentially avoidable expenditures. Here, the CJR model addresses a defined population (FFS Medicare beneficiaries undergoing LEJR procedures) for which there are potentially avoidable expenditures (arising from less than optimal care coordination). For the reasons described elsewhere in this rule, we have determined that it is necessary to test this model among varying types of hospitals that have not chosen to voluntarily participate in another episode payment model such as BPCI. As noted elsewhere in this final rule, we are testing an episode approach for LEJR episodes through the voluntary BPCI models. We have designed the CJR model to require participation by hospitals in order to avoid the selection bias inherent to any model in which providers may choose whether to participate. Such a design will allow for testing of how a variety of hospitals will fare under an episode payment approach, leading to a more robust evaluation of the model's effect on all types of hospitals. We believe this is the most prudent approach for the following reasons. The information gained from testing of the CJR model will allow CMS to more comprehensively assess whether LEJR episode payment models are appropriate for any potential national expansion. We will have evaluation information on results for providers who are participating in such models voluntarily (under BPCI) as well as for hospitals that are required to participate in CJR. Under CJR, we will have tested and evaluated such a model across a wide range of hospitals representing varying degrees of experience with episode payment. We believe it is important to gain knowledge from a variety of perspectives in considering whether and which models merit national expansion. Thus, the CJR model meets the criteria required for initial model tests.

Moreover, the Secretary has the authority to establish regulations to carry out the administration of Medicare. Specifically, the Secretary has authority under both sections 1102 and 1871 of the Act to implement regulations as necessary to administer Medicare, including testing this Medicare payment and service delivery model. We note that while CJR will be a model, and not a permanent feature of the Medicare program, the model will test different methods for delivering and paying for services covered under the Medicare program, which the Secretary has clear legal authority to regulate. The proposed rule went into great detail about the provisions of the proposed CJR model, enabling the public to fully understand how the proposed model was designed and could apply to affected providers. We acknowledge section 1115A(d)(2) of the Act, which states that there shall be no administrative or judicial review of, among other things, “the selection of organizations, sites, or participants to test . . . models selected,” as well as the commenter's concern that this provision would preclude a participant hospital from appealing its selection as a participant in the CJR model. However, it is precisely because the model will impose new requirements upon participant hospitals that we undertook notice and comment rulemaking to implement it.

In response to the comment indicating that we misread section 1115A(a)(5) of the Act, we believe that the commenter misunderstood the reference to that provision in the proposed rule. The reference to section 1115A(a)(5) of the Act was made in the context of the discussion of selecting certain MSAs within which we will test the model. We do not rely on section 1115A(a)(5) of the Act specifically as the authority for a model in which participation is not voluntary; rather, as noted previously, we rely on section 1115A of the Act as a whole, as well as the Secretary's existing authority to carry out her duties and administer the Medicare program.

We disagree with commenters that implementing the CJR model will negatively affect beneficiaries' appeal rights. We note that normal claims processes will continue under this model, including beneficiary and provider appeal rights. We also refer readers to section III.C.9. of this final rule for discussion of hospital appeals procedures under the CJR model.

With regard to the comment about CMS sidestepping safeguards designed to prevent imposing haphazard models prior to appropriate vetting and testing, we reiterate that we have undertaken rulemaking to solicit comprehensive public input on all aspects of the CJR model. In addition, as previously noted, the CJR model has been designed to limit selection bias, which will allow for more robust evaluation results across a variety of providers.

We note that this is a new model, not an expansion of an existing model. We disagree with the commenters who believe that we have reversed the order of testing and expansion of Innovation Center models. As permitted by section 1115A of the Act, we are testing the CJR model within specified limited geographic areas. The fact that the model will require the participation of certain hospitals does not mean it is not an initial model test. If the model is successful such that it meets the statutory requirements for expansion, and the Secretary determines that expansion is warranted, we would undertake rulemaking to implement the expansion, as required by section 1115A(c) of the Act.

Comment: Several commenters questioned how the proposed CJR model relates to the potential for expansion of BPCI. Commenters also noted that CMS included language in the FY 2016 IPPS/LTCH PPS proposed rule requesting public input on an eventual expansion of BPCI.

Response: CMMI's three major priorities include testing new payment and service delivery models, evaluating results and advancing best practices, and engaging stakeholders. Since 2011, we have been working to develop and test models of bundling Medicare payments under the authority of section 1115A of the Act. Consistent with its ongoing commitment to develop new models and refine existing models based on additional information and experience, we may modify existing models or test additional models under our authority under section 1115A of the Act. The CJR model is a new, additional episode payment model being tested under the authority of section 1115A of the Act. As such, it is not an expansion of the BPCI initiative, which needs further evaluation to determine its impact on both Medicare cost and quality before the Secretary can determine whether the findings from the evaluation of the initiative demonstrate that it meets all criteria for expansion, consistent with the requirements of section 1115A(c) of the Act, and that, based on these findings and other pertinent factors, expansion is warranted.

In the FY 2016 IPPS/LTCH PPS proposed rule (80 FR 24414 through 24418), we solicited public comments regarding policy and operational issues related to a potential expansion of the BPCI initiative in the future. We explained that as we initiated discussions about potential expansion, we continued to value stakeholder Start Printed Page 73279engagement within the framework of our three priorities. With respect to expansion, section1115A(c) of the Act, as added by section 3021 of the Affordable Care Act, provides the Secretary with the authority to expand through rulemaking the duration and scope of a model that is being tested under section 1115A(b) of the Act, such as the BPCI initiative (including implementation on a nationwide basis), if the following findings are made, taking into account the evaluation of the model under section 1115A(b)(4) of the Act: (1) The Secretary determines that the expansion is expected to either reduce Medicare spending without reducing the quality of care or improve the quality of patient care without increasing spending; (2) the CMS Chief Actuary certifies that the expansion would reduce (or would not result in any increase in) net Medicare program spending; and (3) the Secretary determines that the expansion would not deny or limit the coverage or provision of Medicare benefits. The decision of whether or not to expand BPCI will be made by the Secretary in coordination with CMS and the Office of the Chief Actuary based on whether findings about the initiative meet the statutory criteria for expansion under section 1115A(c) of the Act. We did not propose an expansion of any of the BPCI models or any policy changes associated with those models in the FY 2016 IPPS/LTCH PPS proposed rule.

Although BPCI and the CJR model both include testing episode payment for LEJR episodes of care, CJR differs from BPCI in significant ways, as detailed throughout this final rule. Providers elected to participate in BPCI, and were given a choice of various design features, such as the clinical episodes included and the episode length. The CJR model was designed in part based on feedback and experience from BPCI, and will provide additional information on the impact of episode payment for LEJR episodes across a variety of hospitals, including those who may not have elected to participate in the model. As previously discussed in this section, it is necessary to require participation in the CJR model in order to avoid the selection bias inherent to any voluntary model. When the CJR model begins on April 1, 2016, we will be testing both episode payment models concurrently for a period of time, as well as many other payment and service delivery models, in order to gain information about the most successful strategies to improve the quality of care and reduce spending. The different design features of BPCI and the CJR model will aid us in evaluating the success of episode-based payment across a range of provider types and in a range of geographic areas. As evaluation results addressing the impact of each model on Medicare quality and cost become available, the Secretary will review this information to determine whether the findings from the evaluation of the model demonstrate that it meets all criteria for expansion, consistent with the requirements of section 1115A(c) of the Act, and that, based on these findings and other pertinent factors, expansion is warranted.

Comment: Many commenters requested changes to the BPCI model in response to the proposed rule. Commenters also requested clarification on how BPCI awardees would be transitioned into the CJR model; for example, which performance year policies would apply to the new model participants.

Response: We will not address comments about BPCI policies in this final rule. We will address commenters' suggestions on BPCI through our usual processes for informing BPCI participants and the public of any changes to BPCI. As discussed in section III.A of this final rule, all Inpatient Prospective Payment System (IPPS) hospitals in the selected MSAs that are not participating in BPCI Model 1 or Phase II of Models 2 or 4 for LEJR episodes would be included in the CJR model. We intend for the current performance year's policies to be in effect for any new entrants in the CJR model. We also note that an acute care hospital formerly participating in BPCI for the LEJR episode will have likely established care coordination and redesign strategies for success. As such, it would not be necessary to grant such hospitals additional time to transition from BPCI into the CJR model.

Comment: Numerous commenters requested that physicians who enter into sharing arrangements with CJR hospitals qualify as eligible professionals under the Medicare Access and Chip Reauthorization Act of 2015 (MACRA) beginning in 2019. A commenter requested that all CJR collaborators qualify as eligible professionals under MACRA. Several commenters outlined wholly different structures for the proposed CJR model, including provisions that would allow for the CJR model to qualify as an alternative payment model (APM) under MACRA.

Response: We interpret commenters' requests as follows: That collaborators under the CJR model would be able to meet the requirements that would otherwise apply under the Merit-based Incentive Payment System (MIPS) or, alternatively, qualify as APM participants under section 1833(z)(2) of the Act (and therefore be excluded from MIPS) through their participation in CJR. We further interpret commenters' requests as follows: That CJR would include eligible alternative payment entities, and therefore that eligible professionals in CJR would potentially be qualifying APM participants. We note that the statute specifies which types of individuals qualify as eligible professionals (EPs) under section 1848(k)(3)(B) of the Act or as MIPS EPs under section 1848(q)(1)(C) of the Act. We plan to develop regulations under MACRA through notice and comment rulemaking. We will be releasing further guidance on the implementation of MACRA, and through such guidance, will be clarifying the parameters for eligibility under MACRA.

Comment: Several commenters presented different episode payment models for CMS' consideration to be tested in addition to or instead of the CJR model, or suggested such major changes to the proposed CJR model design elements that the result of their adoption would be a wholly different test of episode payment than CMS proposed. A few commenters recommended that CMS consider testing a model that emphasizes the role of PAC providers in managing episode care for beneficiaries, instead of just the hospital. Such a model would assign financial responsibility during an episode to a PAC entity with capabilities to coordinate care across a wide range of post-acute settings. Other commenters suggested that CMS test a model that would create physician-led organizations to manage financial risk for LEJR episodes of care, instead of assigning risk to hospitals. These organizations would receive prospective episodic payments and allocate such payments among the providers and suppliers furnishing care to beneficiaries during an LEJR episode. Several commenters recommended CMS implement a population-based model similar to an Accountable Care Organization (ACO) model, in lieu of an episode-based payment model. Finally, a commenter requested that instead of including rural and low-volume hospitals in the CJR model, CMS develop a model tailored to this subset of providers.

Response: We appreciate the suggestions for alternatives to the CJR model design that were recommended by the commenters, including the details and rationale provided about many features of those models. We are Start Printed Page 73280not adopting these approaches to an episode payment model under this final rule as we did not propose the design elements of such models for public notice and comment nor did we propose the additional policies that would be required to implement such features that do not rely on existing Medicare definitions (for example, the definition of a physician-led organization to manage risk). However, we note that we are constantly considering modifications to existing models and designing new models under our testing authority under section 1115A of the Act, taking into consideration stakeholder input received through many channels, including public comments on this proposed rule and the FY 2016 IPPS/LTCH PPS proposed rule discussion item on potential BPCI expansion considerations, as well as feedback from providers participating in existing models. We note that potential modifications to the CJR model would go through notice and comment rulemaking as necessary. As we consider developing additional payment service and delivery models, we will continue to engage with stakeholders and review all of the information available to us about alternative approaches to episode payment that could be tested.

B. Summary of the Major Provisions

1. Model Overview: LEJR Episodes of Care

LEJR procedures are currently paid under the IPPS (IPPS) through one of two Medicare Severity-Diagnosis Related Groups (MS-DRGs): MS-DRG 469 (Major joint replacement or reattachment of lower extremity with Major Complications or Comorbidities (MCC)) or MS-DRG 470 (Major joint replacement or reattachment of lower extremity without MCC). Under the CJR model, as described further in section III.B of this final rule, episodes will begin with admission to an acute care hospital for an LEJR procedure that is assigned to MS-DRG 469 or 470 upon beneficiary discharge and paid under the IPPS and will end 90 days after the date of discharge from the acute care hospital. This episode of care definition offers operational simplicity for providers and CMS. The episode will include the LEJR procedure, inpatient stay, and all related care covered under Medicare Parts A and B within the 90 days after discharge, including hospital care, PAC, and physician services.

2. Model Scope

We have finalized that participant hospitals will be the episode initiators and bear financial risk under the CJR model. In comparison to other health care facilities, hospitals are more likely to have resources that will allow them to appropriately coordinate and manage care throughout the episode, and hospital staff members are already involved in hospital discharge planning and PAC recommendations for recovery, key dimensions of high quality and efficient care for the episode. We require all hospitals paid under the IPPS in selected geographic areas to participate in the CJR model, with limited exceptions. Eligible beneficiaries who elect to receive care at these hospitals will automatically be included in the model. We have selected geographic areas based on a stratified random sampling methodology within strata using the following criteria: historical wage adjusted episode payments and population size. Our geographic area selection process is detailed further in section III.A of this final rule.

3. Payment

We will test the CJR model for 5 performance years. We have finalized an alternative start date for the model from the timeline set forth in the proposed rule. As discussed in further detail in section III.C.2.a. of this final rule, the first performance year for the CJR model will begin on April 1, 2016 and end on December 31, 2016. During these performance years we will continue paying hospitals and other providers and suppliers according to the usual Medicare FFS payment systems. However, after the completion of a performance year, the Medicare claims payments for services furnished to the beneficiary during the episode, based on claims data, will be combined to calculate an actual episode payment. The actual episode payment is defined as the sum of related Medicare claims payments for items and services furnished to a beneficiary during a CJR episode. The actual episode payment will then be reconciled against an established CJR target price that is stratified based on the beneficiary's fracture status, with consideration of additional payment adjustments based on quality performance, post-episode spending, and policies to limit hospital financial responsibility. The amount of this calculation, if positive, will be paid to the participant hospital. This payment will be called a reconciliation payment. If negative, we will require repayment from the participant hospital. Medicare will require repayment of the difference between the actual episode payments and the CJR target price from a participant hospital if the CJR target price is exceeded.

We will make reconciliation payments to participant hospitals that achieve quality outcomes and cost efficiencies relative to the established CJR target prices in all performance years of the model. We will also phase in the requirement that participant hospitals whose actual episode payments exceed the applicable CJR target price pay the difference back to Medicare beginning in performance year 2. Under this final rule, Medicare will not require repayment from hospitals for performance year 1 for actual episode payments that exceed their target price in performance year 1.

We will also limit how much a hospital can gain or lose based on its actual episode payments relative to target prices. We have also put in place additional policies to further limit the risk of high payment cases for all participant hospitals and for special categories of participant hospitals as described in section III.C. of this final rule.

4. Similar, Previous, and Concurrent Models

The CJR model is informed by other models and demonstrations currently and previously conducted by CMS and will explore additional ways to enhance coordination of care and improve the quality of services through bundled payments. We recently announced the Oncology Care Model (OCM), a new voluntary payment model for physician practices administering chemotherapy. Under OCM, practices will enter into payment arrangements that include financial and performance accountability for episodes of care surrounding chemotherapy administration to cancer patients. We plan to coordinate with other payers to align with OCM in order to facilitate enhanced services and care at participating practices. More information on the OCM can be found on CMMI's Web site at: http://innovation.cms.gov/​initiatives/​Oncology-Care/​. Medicare tested innovative approaches to paying for orthopedic services in the Medicare Acute Care Episode (ACE) demonstration, a prior demonstration, and is currently testing additional approaches under BPCI. Both of these models have also informed the design of the CJR model.

Under the authority of section 1866C of the Act, we conducted a 3-year demonstration, the ACE Demonstration. The demonstration used a prospective global payment for a single episode of care as an alternative approach to payment for service delivery under traditional Medicare FFS. The episode Start Printed Page 73281of care was defined as a combination of Part A and Part B services furnished to Medicare FFS beneficiaries during an inpatient hospital stay for any one of a specified set of cardiac and orthopedic MS-DRGs. The MS-DRGs tested included 469 and 470, which are included in the CJR model. The discounted bundled payments generated an average gross savings to Medicare of $585 per episode for a total of $7.3 million across all episodes (12,501 episodes) or 3.1 percent of the total expected costs for these episodes. After accounting for increased PAC costs that were observed at two sites, Medicare saved approximately $4 million, or 1.72 percent of the total expected Medicare spending. More information on the ACE Demonstration can be found on CMMI's Web site at: http://innovation.cms.gov/​initiatives/​ACE/​.

We are currently testing the BPCI initiative. The BPCI initiative is comprised of four related payment models, which link payments for multiple services that Medicare beneficiaries receive during an episode of care into a bundled payment. Under the initiative, entities enter into payment arrangements with CMS that include financial and performance accountability for episodes of care. Episodes of care under the BPCI initiative begin with either—(1) An inpatient hospital stay; or (2) PAC services following a qualifying inpatient hospital stay. The BPCI initiative is evaluating the effects of episode-based payment approaches on patient experience of care, outcomes, and cost of care for Medicare FFS beneficiaries. Each of the four models tests LEJR episodes of care. While final evaluation results for the models within the BPCI initiative are not yet available, we believe that CMS' experiences with BPCI support the design of the CJR model. Under section 1115A(c) of the Act, the Secretary may, taking into consideration an evaluation conducted under section 1115A (b)(4) of the Act, “through rulemaking, expand (including implementation on a nationwide basis) the duration and the scope of a model that is being tested under” CMMI's authority. CJR is not an expansion of BPCI, and BPCI may be expanded in the future. We published a discussion item soliciting public comment on a potential future expansion of one or more of the models within BPCI in the FY2016 IPPS rule, 80 FR 24414 through 24418. CJR will not be an expansion or modification of BPCI; nor does it reflect comments received in response to the proposed rule for the 2016 IPPS Rule. CJR is a unique model that tests a broader, different group of hospitals than BPCI. It is necessary to provide CMS with information about testing bundled payments to hospitals that are required to participate in an APM. For a discussion of why we are requiring hospitals to participate in the CJR model, see section III.A. of this final rule.

The CJR model's design was informed to a large degree by our experience with BPCI Model 2. BPCI's Model 2 is a voluntary episode payment model in which a qualifying acute care hospitalization initiates a 30, 60 or 90 day episode of care. The episode of care includes the inpatient stay in an acute care hospital and all related services covered under Medicare Parts A and B during the episode, including PAC services. More information on BPCI Model 2 can be found on CMMI's Web site at: http://innovation.cms.gov/​initiatives/​BPCI-Model-2/​.

Further information of why elements of the OCM, the ACE Demonstration, and BPCI Model 2 were incorporated into the design of the CJR model appears later in this final rule.

5. Overlap With Ongoing CMS Efforts

We are excluding from participation in CJR certain hospitals participating in the risk-bearing phase of BPCI Models 2 and 4 for LEJR episodes, as well as acute care hospitals participating in BPCI Model 1. We are not excluding beneficiaries in CJR model episodes from being included in other Innovation Center models or CMS programs, such as the Medicare Shared Savings Program (Shared Savings Program), as detailed later in this final rule. We will account for overlap, that is, where CJR beneficiaries are also included in other models and programs, to ensure the financial policies of CJR are maintained and results and spending reductions are attributed to the correct model or program.

6. Quality Measures and Reporting Requirements

We are adopting two hospital-level quality of care measures for the CJR model. Those measures include a complications measure and a patient experience survey measure. We will use these measures in the model pay-for-performance payment methodology, as well as to test the success of the model in achieving its goals under section 1115A of the Act and to monitor for beneficiary safety. We intend to publicly report this information on the Hospital Compare Web site. Additionally, we will encourage the voluntary submission of data to support the development of a hospital-level measure of patient-reported outcomes following an elective primary total hip (THA) or total knee arthroplasty (TKA) through incorporation of the measure in the composite quality scoring methodology described in III.C.5. of this final rule.

7. Data Sharing Process

We will share data with participant hospitals upon request throughout the performance period of the CJR model to the extent permitted by the HIPAA Privacy Rule and other applicable law. We will share upon request both raw claims-level data and claims summary data with participants. This approach will allow participant hospitals without prior experience analyzing claims to use summary data to receive useful information, while allowing those participant hospitals who prefer raw claims-level data the opportunity to analyze claims. We will provide hospitals with up to 3 years of retrospective claims data upon request that will be used to develop their target price, as described in section III.C. of this final rule. In accordance with the HIPAA Privacy Rule, we will limit the content of this data set to the minimum data necessary for the participant hospital to conduct quality assessment and improvement activities and effectively coordinate care of its patient population.

8. Beneficiary Protections

Under the CJR model, beneficiaries retain the right to obtain health services from any individual or organization qualified to participate in the Medicare program. Under the CJR model, eligible beneficiaries who receive services from a participant hospital will not have the option to opt out of inclusion in the model. We require participant hospitals to supply beneficiaries with written information regarding the design and implications of this model as well as their rights under Medicare, including their right to use their provider of choice. We will also make a robust effort to reach out to beneficiaries and their advocates to help them understand the CJR model.

We also will use our existing authority, if necessary, to audit participant hospitals if claims analysis indicates an inappropriate change in delivered services. Beneficiary protections are discussed in greater depth in section III.E. of this final rule.

9. Financial Arrangements and Program Policy Waivers

We will hold participant hospitals financially responsible for CJR LEJR episodes as participants in the model as discussed in section III.C.6. of this final Start Printed Page 73282rule. Specifically, only these hospital participants will be directly subject to the requirements of this final rule for the CJR model. Participant hospitals will be responsible for ensuring that other providers and suppliers collaborating with the hospital on LEJR episode care redesign are in compliance with the terms and conditions of the model, including any applicable program policy waivers.

Several of the Medicare program policy waivers outline the conditions under which SNFs and physicians could furnish and bill for certain services furnished to CJR beneficiaries where current Medicare programs rules will not permit such billing. We draw the attention of SNFs and physicians to these waivers, which are included in section III.C.11.b.(5). of this final rule.

C. Summary of Economic Effects

As shown in our impact analysis, we expect the CJR model to result in savings to Medicare of $343 million over the 5 performance years of the model. We note that a composite quality score will be calculated for each hospital in order to determine eligibility for a reconciliation payment and whether the hospital qualifies for quality incentive payments that will reduce the effective discount percentage experience by the hospital at reconciliation for a given performance year.

More specifically, in performance year 1 of the model, we estimate a Medicare cost of approximately $11 million, as hospitals will not be subject to downside risk in the first year of the model. As we introduce downside risk beginning in performance year 2 of the model, we estimate Medicare savings of approximately $36 million. In performance year 3 of the model, we estimate Medicare savings of $71 million. In performance years 4 and 5 of the model, we will move from target episode pricing that is based on a hospital's experience to target pricing based on regional experience, we estimate Medicare savings of $120 million and $127 million, respectively.

As a result, we estimate the net savings to Medicare to be $343 million over the 5 performance years of the model. We anticipate there will be a broader focus on care coordination and quality improvement for LEJR episodes among hospitals and other providers and suppliers within the Medicare program that will lead to both increased efficiency in the provision of care and improved quality of the care provided to beneficiaries.

We note that under section 1115A(b)(3)(B) of the Act, the Secretary is required to terminate or modify a model unless certain findings can be made with respect to savings and quality after the model has begun. If during the course of testing the model it is determined that termination or modification is necessary, such actions will be undertaken through rulemaking as necessary.

II. Background

A General Background

This final rule finalizes the implementation of a new innovative health care payment model under the authority of section 1115A of the Act. Under the model, called the CJR model, acute care hospitals in certain selected geographic areas will receive bundled payments for episodes of care where the diagnosis at discharge includes a lower extremity joint replacement (LEJR) or reattachment of a lower extremity that was furnished by the hospital. The bundled payment will be paid retrospectively through a reconciliation process; hospitals and other providers and suppliers will continue to submit claims and receive payment via the usual Medicare FFS payment systems. All related care covered under Medicare Part A and Part B within 90 days after the date of hospital discharge from the joint replacement procedure will be included in the episode of care. We believe this model will further our goals of improving the efficiency and quality of care for Medicare beneficiaries for these common medical procedures.

B. Acronym of This Model

We have changed the acronym of this model to “CJR” and have updated all references in this rule and the regulations to reflect this change.

C. Public Comments Received in Response to the CJR Proposed Rule

We received approximately 400 timely pieces of correspondence containing multiple comments on the CJR proposed rule. We note that some of these public comments were outside of the scope of the proposed rule. These out-of-scope public comments are mentioned but not addressed with the policy responses in this final rule. Summaries of the public comments that are within the scope of the proposed rule and our responses to those public comments are set forth in the various sections of this final rule under the appropriate heading.

III. Provisions of the Proposed Regulations and Analysis of and Responses to Public Comments

A. Definition of the Episode Initiator and Selected Geographic Areas

1. Background

The CJR model is different from BPCI because it would require participation of all hospitals (with limited exceptions) throughout selected geographic areas, which would result in a model that includes varying hospital types. However, a discussion of BPCI is relevant because its design informs and supports the proposed CJR model. The BPCI model is voluntary, and under that model we pay a bundled payment for an episode of care only to entities that have elected to participate in the model. We are interested in testing and evaluating the impact of an episode payment approach for LEJRs in a variety of other circumstances, including among those hospitals that have not chosen to voluntarily participate because we have not tested bundled payments for these hospitals previously. This would allow CMS and participants to gain experience testing and evaluating episode-based payment for LEJR procedures furnished by hospitals with a variety of historic utilization patterns; roles within their local markets; volume of services provided; access to financial, community, or other resources; and population and health care provider density. Most importantly, participation of hospitals in selected geographic areas will allow CMS to test bundled payments without introducing selection bias such as the selection bias inherent in the BPCI model due to self-selected participation.

2. Definition of Episode Initiator

Under the CJR model, as described further in section III.B. of this final rule, episodes will begin with admission to an acute care hospital for an LEJR procedure that is paid under the IPPS through Medical Severity Diagnosis-Related Group (MS-DRG) 469 (Major joint replacement or reattachment of lower extremity with MCC) or 470 (Major joint replacement or reattachment of lower extremity without MCC) and end 90 days after the date of discharge from the hospital. For the CJR model, we proposed that hospitals would be the only episode initiators. For purposes of CJR, the term “hospital” means a hospital as defined in section 1886(d)(1)(B) of the Act. This statutory definition of hospital includes only acute care hospitals paid under the IPPS. We proposed that all acute care hospitals in Maryland would be Start Printed Page 73283excluded from CJR. The state of Maryland entered into an agreement with CMS, effective January 1, 2014, to participate in CMS' new Maryland All-Payer Model. In order to implement the Maryland All-Payer Model, CMS waived certain requirements of the Act, and the corresponding implementing regulations, as set forth in the agreement between CMS and Maryland. Specifically, under the Maryland All-Payer Model, Maryland acute care hospitals are not paid under the IPPS or Outpatient Prospective Payment System (OPPS) but rather are paid under rates set by the state. Following the model's performance period, Maryland will transition to a new model that incorporates the full spectrum of care, not just hospital services. As such, with respect to Maryland hospitals, CMS intends to test and develop new payment and delivery approaches that can incorporate non-hospital services in a manner that accounts for Maryland's unique hospital rate setting system and permit Maryland to develop its own strategy to incentivize higher quality and more efficient care across clinical situations within and beyond hospitals, including but not limited to LEJR episodes of care. We proposed that because Maryland hospitals are not paid under the IPPS or OPPS, payments to Maryland hospitals will be excluded in the regional pricing calculations as described in section III.C.4. of this final rule. We sought comment on whether there were potential approaches for including Maryland acute care hospitals in CJR. In addition, we sought comment on whether Maryland hospitals should be included in CJR in the future upon any termination of the Maryland All-Payer Model.

The following is a summary of the comments received and our responses.

Comment: Several commenters commented on the proposed exclusion of Maryland hospitals in the All-Payer model from the model. A commenter requested that if we are considering approaches for including Maryland acute care hospitals in the CJR model that we ensure that the inclusion of such hospitals would not jeopardize the current all-payer system in Maryland. If such an approach were to be developed, the commenter noted that it would welcome the opportunity to participate in the CJR model and further stated that it is confident that it would be successful under the CJR model in helping to further to goals of providing high quality care at lower costs to better patient outcomes and population health. Another commenter noted that Maryland's All-Payer Model Agreement is focused on holding hospitals accountable for improving care, improving health, and reducing the total cost of hospital care for all payers. Under the All-Payer model, Maryland has shifted its long-standing hospital rate-setting system from a volume-based system, focused on cost per case, to a global population-based system that incorporates performance requirements for quality and outcomes. The Maryland system will be held accountable for the total cost of care for Medicare patients under its contract with CMS and thus already has two-sided risk for hospital costs. The commenter stated that Maryland wants to work with CMS to develop a unique approach to achieving the goals of the model, but under the All-Payer model. Lastly, another commenter expressed confusion if we were announcing a plan to have Maryland transition to a new model that incorporates the full spectrum of care, not just hospital services.

Response: Under the All-Payer model, Maryland has facilitated the movement of regulated hospital revenue into population-based payment reimbursement under a hospital global budget model. We appreciate the state's efforts to move away from volume-based payments and to focus on reducing total cost of care and improving quality of care, and we have seen improvement on these areas in the first year of the All-Payer model. However, we remain concerned that certain aspects of the All-Payer Model make it challenging for Maryland to be included in other payment and delivery innovations being launched by the CMS Innovation Center. As we anticipate testing more models across the country, we do not want Maryland to fall behind in payment and delivery innovation. We are very interested in Maryland's strategy to be accountable for total cost of care beyond hospital services, which we intend to implement under the All-Payer model in 2019. We note that we are not announcing a new model for Maryland in this rule, but rather the CMS Innovation Center looks forward to working with Maryland on its total cost of care model.

Comment: Several commenters agreed with CMS that Maryland hospitals should not be included in our definition of “hospital” and, instead, the state of Maryland should be allowed to develop its own strategy to encourage higher quality care and efficiencies across clinical settings.

Response: We agree that for the purposes of the CJR model, the term “hospital” should only encompass hospitals currently paid under the IPPS and we are finalizing as proposed to exclude Maryland hospitals from the CJR model.

Final Decision: After consideration of the public comments we received, we are finalizing, for purposes of the CJR model, the term “hospital” to mean a hospital subject to the IPPS as defined in section 1886(d)(1)(B) of the Act. This statutory definition of hospital includes only acute care hospitals paid under the IPPS, thus excluding Maryland hospitals from participating in CJR and excluding payments to Maryland hospitals in regional pricing calculations described in section III.C.4 of this final rule. This definition will be codified in § 510.2

We proposed to designate IPPS hospitals as the episode initiators to ensure that all Medicare FFS LEJR services furnished by participant hospitals in selected geographic areas to beneficiaries who do not meet the exclusion criteria (specified in section III.B.3. and section III.C.7. of this final rule) are included in the CJR model. Given that our proposal that the LEJR episode begins with an admission to a hospital paid under the IPPS that results in a discharge assigned to MS-DRG 469 or 470, we further believed that utilizing the hospital as the episode initiator is a straightforward approach for this model because the hospital furnishes the LEJR procedure. In addition, we noted our interest in testing a broad model in a number of hospitals under the CJR model in order to examine results from a more generalized payment model. Thus, we believed it is important that, in a model where hospital participation is not voluntary, all Medicare FFS LEJR episodes that begin at the participant hospital in a selected geographic area should be included in the model for beneficiaries that do not meet the exclusion criteria specified in section III.B.3. of this final rule and are not LEJR BPCI episodes that we are excluding as outlined in this section and also in section III.C.7 of this final rule. This is best achieved if the hospital is the episode initiator. Finally, as described in the following sections that present our proposed approach to geographic area selection, this geographic area selection approach relies upon our definition of hospitals as the entities that initiate episodes. We sought comment on our proposal to define the episode initiator as the hospital under CJR. However, commenters generally commented on our proposal to define the episode initiator as the hospital in tandem with comments regarding the proposal that the hospital also be the entity financially responsible for the episode of care under CJR. As such, comments regarding the proposed Start Printed Page 73284episode initiator and the entity financially responsible for the episode of care are summarized in section III.A.2. of this final rule.

3. Financial Responsibility for the Episode of Care

BPCI Model 2 participants that have entered into agreements with CMS to bear financial responsibility for an episode of care include acute care hospitals paid under the IPPS, health systems, physician-hospital organizations, physician group practices (PGPs), and non-provider business entities that act as conveners by coordinating multiple health care providers' participation in the model. Thus, our evaluation of BPCI Model 2 will yield information about how results for LEJR episodes may differ based on differences in which party bears financial responsibility for the episode of care. For the CJR model, we proposed to make hospitals financially responsible for the episode of care.

Although we proposed that hospitals would bear the financial responsibility for LEJR episodes of care under CJR, because there are LEJR episodes currently being tested in BPCI Model 1, 2, 3 or 4, we believed that participation in CJR should not be required if it would disrupt testing of LEJR episodes already underway in BPCI models. Therefore, we proposed certain exceptions for instances where IPPS hospitals located in an area selected for the model are active participant hospitals or episode initiators for LEJR episodes as of July 1, 2015, and exceptions for LEJR episodes initiated by other providers or suppliers under certain BPCI models.

The following is a summary of the comments received and our responses.

Comment: Most commenters expressed overall support for the CJR model, with some commenters noting that the CJR model could help to transform care delivery through improved care coordination and financial accountability. Several commenters further expressed support for our proposal to designate hospitals as the episode initiators and the entity financially responsible for the episode of care under the CJR model. These commenters agreed that hospitals should bear the responsibility of implementing the CJR model and further agreed with being able to share this responsibility with “collaborators” through gainsharing agreements. The commenters noted that the themes surrounding responsibility and cost in conjunction with quality as presented in the proposed rule were encouraging and show a continued focus on bettering outcomes and patient engagement while lowering costs.

Response: We thank the commenters for their support. As noted in the proposed rule, the intent of the CJR model is to promote quality and financial accountability for episodes of care surrounding a lower-extremity joint replacement (LEJR) or reattachment of a lower extremity procedure. We anticipate the CJR model would benefit Medicare beneficiaries by improving the coordination and transition of care, improving the coordination of items and services paid for through Medicare FFS, encouraging more provider investment in infrastructure and redesigned care processes for higher quality and more efficient service delivery, and incentivizing higher value care across the inpatient and PAC spectrum spanning the episode of care (80 FR 41198).

Comment: Several commenters disagreed with the proposal for the CJR model to limit financial responsibility for the episode of care to only hospitals. Commenters advocated for PGPs or orthopedic surgeons to be financially responsible, while other commenters advocated for PAC entities to be financially responsibility for the episode of care. Commenters listed a variety of reasons why orthopedic physician groups and/or PAC providers should be financially responsible for the episode of care. Some commenters stated that the episode initiator for the CJR model should be a physician, as key clinical decisions about care within the episode are made by physicians, including determining what kind of follow-up care is needed. A few commenters stated that the episode initiator should be the PAC provider, similar to BPCI Model 3, since much of the reduction in CJR episode costs could occur through changes in PAC utilization. A few commenters stated that CMS should distribute program risk across all providers within the episode of care and not delegate that function to the hospital because during a CJR episode, ideal care and successful care coordination involve multiple providers across the care continuum and is especially dependent on PAC providers. Finally, several commenters stated that with gainsharing there is greater opportunity for the physician to participate in patient care redesign, but that unless the physician is also financially responsible, physician involvement in the full care redesign would be less than ideal.

Response: As noted in the proposed rule (80 FR 41204 through 41205), because the CJR model is testing a more generalizable model by including providers that might not participate in a voluntary model, we believe it is most appropriate to identify a single type of provider to bear financial responsibility for making repayment to CMS under the CJR model as one entity needs to be ultimately responsible for ensuring that care for CJR model beneficiaries is appropriately furnished and coordinated in order to avoid fragmented approaches that are often less effective and more costly. Hospitals play a central role in coordinating episode-related care and ensuring smooth transitions for beneficiaries undergoing LEJR procedures. Most hospitals already have some infrastructure in place related to patient and family education and health information technology as hospitals receive incentive payments for the adoption and meaningful use of interoperable health information technology (HIT) and qualified electronic health records (EHRs). In addition, hospitals are required by the hospital Conditions of Participation (CoPs) to have in effect a discharge planning process that applies to all patients (§ 482.43). As part of the discharge planning process, hospitals are required to arrange for the initial implementation of the discharge plan (§ 482.43(c)(3)), which includes coordinating with PAC providers, a function usually performed by hospital discharge planners or case managers. Thus hospitals can build upon already established infrastructure, practices, and procedures to achieve efficiencies under this episode payment model. Many hospitals also have recently heightened their focus on aligning their efforts with those of community providers to provide an improved continuum of care due to the incentives under other CMS models and programs, including ACO initiatives such as the Medicare Shared Savings Program, and the Hospital Readmissions Reduction Program (HRRP), establishing a base for augmenting these efforts under the CJR model. Hospitals are also more likely than other providers and suppliers to have an adequate number of episode cases to justify an investment in episode management for this model, have access to resources that would allow them to appropriately manage and coordinate care throughout the LEJR episode, and hospital staff is already involved in discharge planning and placement recommendations for Medicare beneficiaries, and more efficient PAC service delivery provides substantial opportunities for improving quality and reducing costs under CJR.

We considered requiring treating physicians (orthopedic surgeons or Start Printed Page 73285others) or their associated PGPs, if applicable, to be financially responsible for the episode of care under the CJR Model. However, the services of providers and suppliers other than the hospital where the acute care hospitalization for the LEJR procedure occurs would not necessarily be furnished in every LEJR episode. For example, that physicians of different specialties play varying roles in managing patients during an acute care hospitalization for a surgical procedure and during the recovery period, depending on the hospital and community practice patterns and the clinical condition of the beneficiary and could not be assumed to be included in every LEJR episode. This variability would make requiring a particular physician or PGP to be financially responsible for a given episode very challenging. If we were to assign financial responsibility to the operating physician, it is likely that there would be significant variation in the number of relevant episodes that could be assigned to an individual person. Where the physician was included in a PGP, episodes could be aggregated to this group level but this would not be possible for all cases and would likely still have multiple instances with physicians with a very low volume of cases. We acknowledge that providers and suppliers with low volumes of cases may not find it in their financial interests to make systematic care redesigns or engage in an active way with the CJR model. We expect that low volume hospitals may achieve less savings compared to their target episode payments for the simple reason that they would not find the financial incentive present in the CJR sufficiently strong to cause them to shift their practice patterns. While this concern is present in low volume hospitals, it is much more likely to occur if physicians are financially responsible for episode costs because physicians typically do not have the case volume to justify an investment in the infrastructure needed to adequately provide the care coordination services required under the CJR model (such as dedicated support staff for case management), which leads us to believe that as a result, the model would be less likely to succeed.

Although the BPCI initiative allows a PGP and PAC providers to have financial responsibility for episodes of care, the physician groups and PAC providers electing to participate in BPCI have done so because their business structure supports care redesign and other infrastructure necessary to bear financial responsibility for episodes and is not necessarily representative of the typical group practice or PAC provider. Most of the PGPs in BPCI are not bearing financial responsibility, but are participating in BPCI as partners with convener organizations, which enter into agreements with CMS on behalf of health care providers, through which they accept financial responsibility for the episode of care. The PAC providers in BPCI are not at risk for episodes that include more than just the post-anchor hospital discharge period. The incentive to invest in the infrastructure necessary to accept financial responsibility for the entire CJR episode of care, starting at admission to an acute care hospital for an LEJR procedure that is paid under the IPPS MS-DRG 469 or 470 and ending 90 days after the date of discharge from the hospital, would not be present across all PGPs and PAC providers. Thus we do not believe it would be appropriate to designate PGPs or PAC providers to bear the financial responsibility for making repayments to CMS under the CJR model where participation is mandatory, rather than voluntary in nature, potentially causing this model to be less likely to succeed. We may consider, through future rulemaking, other episode of care models in which PGPs or PAC providers are financially responsible for the costs of care.

Comment: Several commenters suggested that conveners—non-provider business entities that coordinate multiple health care providers' participation in the model—should also be allowed to bear financial responsibility for episodes of care under the proposed CJR model. A commenter suggested that instead of making hospitals responsible for managing payments and costs, a management organization should be designated or created to manage the costs and payments.

Response: In the BPCI initiative, participants have entered into a variety of relationships with entities above the hospital level. Some of these relationships are ones where the financial responsibility is borne by an entity other than the hospital, such as a parent organization (known as awardee conveners). Other relationships between hospitals and other organizations (known as facilitator conveners) are more managerial or consultative where financial responsibility remains with the episode initiator (for example, the hospital). We acknowledge the important role that conveners play in the BPCI initiative with regard to providing infrastructure support to hospitals and other entities initiating episodes in BPCI. The convener relationship (where another entity assumes financial responsibility) may take numerous forms, including contractual (such as a separate for-profit company that agrees to take on a hospital's financial risk in the hopes of achieving financial gain through better management of the episodes) and through ownership (such as when financial responsibility is borne at a corporate level within a hospital chain). However, we proposed that for the CJR model we would hold only the participant hospitals financially responsible for the episode of care. This is consistent with the goal of evaluating the impact of bundled payment and care redesign across a broad spectrum of hospitals with varying levels of infrastructure and experience in entering into risk-based reimbursement arrangements. If conveners were included as participants in CJR, we may not gain the knowledge of how a variety of hospitals can succeed in relationship with CMS in which they bear financial risk for the episode of care.

While we proposed that the participant hospital be financially responsible for the episode of care under CJR, we agreed that effective care redesign for LEJR episodes requires meaningful collaboration among acute care hospitals, PAC providers, physicians, and other providers and suppliers within communities to achieve the highest value care for Medicare beneficiaries. We believe it may be essential for key providers and suppliers to be aligned and engaged, financially and otherwise, with the hospitals, with the potential to share financial responsibility with those hospitals. As such, CJR participant hospitals may enter into relationships with other entities in order to manage the episode of care or distribute risk. We refer readers to section III.C.10 of this final rule for further discussion of financial arrangements between participant hospitals and other providers and suppliers. Depending on a hospital's current degree of clinical integration, new and different contractual relationships among hospitals and other health care providers and suppliers may be important, although not necessarily required, for CJR model success in a community. We acknowledge that financial incentives for other providers and suppliers may be important aspects of the model in order for hospitals to partner with these providers and suppliers and incentivize certain strategies to improve episode efficiency.

As noted in the proposed rule (80 FR 41261), in addition to providers and Start Printed Page 73286suppliers with which the participant hospital may want to enter into financial arrangements to share risks and rewards, we expect that participant hospitals may choose to engage with organizations that are neither providers nor suppliers to assist with matters such as: Episode data analysis; local provider and supplier engagement; care redesign planning and implementation; beneficiary outreach; CJR beneficiary care coordination and management; monitoring participant hospital compliance with the terms and conditions of the CJR model; or other model-related activities. These organizations may play important roles in a hospital's plans to implement the CJR model based on the experience these organizations may bring to the hospital's successful participation in the model, such as prior experience with bundled payment initiatives, care coordination expertise, familiarity with the local community, and knowledge of Medicare claims data. All relationships established between participant hospitals and these organizations for purposes of the CJR model would only be those permitted under existing law and regulation, meaning that gainsharing agreements between hospitals and organizations that are neither providers nor suppliers are not permitted. Hospital relationships with organizations other than providers and suppliers would be based solely on the ability of such organizations to directly support the participant hospitals' CJR model implementation.

Comment: Numerous commenters urged CMS to implement the CJR model on a voluntary basis, rather than requiring hospitals to participate. Commenters observed that the CJR model was unprecedented, unjustified, and risky for beneficiaries, because it was the first time CMS would require participation of providers who may not have the interest, experience, capability, or infrastructure to carry out what is necessary for an experiment whose outcomes are unknown. Other commenters claimed that some of the hospitals in the selected MSAs would not be prepared for model participation due to a lack of resources to better coordinate care, insufficient infrastructure, low patient volumes, and lack of negotiating power in their communities, among other reasons. A few commenters disagreed with designating hospitals as financially responsible for the episode of care under CJR if the hospital cannot withdraw its participation if it cannot thrive under the model. The commenters stated that absent readmissions, hospitals have limited influence over other, non-surgical costs associated with joint replacements, such as PAC, rehabilitation, home care, doctors' visits, and more. Conversely, a commenter wrote that there may be some hospitals not in the selected MSAs that would like to participate in CJR and would be precluded from doing so unless CMS opens the model to other hospitals who volunteered to participate. Several commenters requested that CMS continue to test voluntary payment models so that providers can continue to tailor bundled payment reforms to their particular patient populations, practice settings, markets, infrastructure, and administrative resources. A commenter stated that requiring participation in the CJR model may preclude testing of alternative, potentially more effective, approaches. Another commenter contended that requiring participation in this model for providers who may also be participating in a voluntary payment model could create confusion and competing incentives. Commenters further questioned the appropriateness of requiring participation in CJR, given that hospitals may not have contractual agreements with other providers and suppliers furnishing services during an episode. Finally, several commenters contended that the CJR model could result in beneficiary harm; a commenter stated that because participation in the CJR model is required, CMS should be held responsible for any harm to beneficiaries as a result of the model.

Response: We appreciate the views of the commenters on our proposal for required participation in the CJR model test of LEJR episode payment. We recognize that the CJR model represents the first time the Innovation Center will require hospital participation in a payment model being tested under section 1115A of the Act, and we have engaged in rulemaking to ensure robust opportunity for public notice and comment on the model and its design. This model will allow CMS to gain experience with making bundled payments to hospitals who have a variety of historic utilization patterns; different roles within their local markets; various volumes of services; different levels of access to financial, community, or other resources; and various levels of population and health provider density including local variations in the availability and use of different categories of PAC providers. We believe that by requiring the participation of a large number of hospitals with diverse characteristics, the CJR model will result in a robust data set for evaluation of this bundled payment approach, and will stimulate the rapid development of new evidence-based knowledge. Testing the model in this manner will also allow us to learn more about patterns of inefficient utilization of health care services and how to incentivize the improvement of quality for common LEJR procedure episodes. Finally, requiring participation removes selection bias and gives CMS a better, more accurate picture of the effects of the model for consideration of any potential expansion on a national scale.

We have multiple years of experience with several types of large voluntary episode payment models where we have successfully collaborated with participants on implementation of episode payment in a variety of settings for multiple clinical conditions. We believe the relatively narrow scope of the model (LEJR episodes only), the phasing in of full financial responsibility over multiple years of the model, and our plan to engage with hospitals to help them succeed under this model through the provision of claims data, will aid hospitals in succeeding under the CJR model. As discussed in section III.C.2. of this final rule, we are also finalizing that the model's first performance period will begin April 1, 2016, instead of on January 1, 2016 as originally proposed. The longer notice of the final model policies before implementation will provide hospitals with more time to prepare for participation by identifying care redesign opportunities, beginning to form financial and clinical partnerships with other providers and suppliers, and using data to assess financial opportunities under the model.

We acknowledge commenters' concern that some hospitals not in a selected MSA may desire to participate in the CJR model. We also note that CMS will continue to test voluntary bundled payment models, including those already undergoing testing through the BPCI initiative, which offered several open periods over the past few years where interested hospitals and other organizations could join. We expect that many providers will continue to engage in initiatives such as BPCI, and may also participate in other emerging models in the coming years. The coexistence of voluntary initiatives such as BPCI alongside new models in which providers are required to participate will provide CMS, providers, and beneficiaries with multiple opportunities to benefit from various care redesign and payment reform initiatives. We will also continue Start Printed Page 73287to explore alternative approaches that may also prove effective in improving care for beneficiaries while reducing spending.

We disagree that requiring participation in the CJR model could create confusion and competing incentives for hospitals already participating in voluntary initiatives. We note that simultaneous testing of multiple bundled payment models is appropriate in many situations, depending on the care targeted under each model. Section III.C.7. of this final rule lays out our policies for accounting for overlap between models and contains discussion of the potential synergies and improved care coordination we expect will ensue through allowing for hospitals and beneficiaries to be engaged in more than one initiative simultaneously.

We appreciate that not all hospitals will have contractual arrangements with providers and suppliers furnishing services to beneficiaries during LEJR episodes. However, this final rule lays out the various financial arrangements that will be permitted under the CJR model, to allow hospitals the opportunity to engage with other providers and suppliers and to form clinical and financial partnerships. Section III.C.10. of this final rule details the requirements for these financial arrangements. Although hospitals will not be required to form financial relationships with other providers and suppliers, we expect many will do so in order to help align the clinical and financial incentives of key providers and suppliers caring for CJR model beneficiaries.

Finally, we do not see how participation in the CJR model, in and of itself, would lead to beneficiary harm and that if beneficiary harm were to occur, that CMS would be responsible. First, and most importantly, we note that under the model, providers and suppliers are still required to provide all medically necessary services, and beneficiaries are entitled to all benefits that they would receive in the absence of the model. Second, we note that we have employed many payment systems, such as IPPS, and payment models, such as BPCI and ACOs, that include similar economic incentives to promote efficiency, and we have not determined that beneficiaries have been harmed by those systems and models. Third, we note that CMS has numerous tools and monitoring plans which are both specific to this model and common to all FFS Medicare. These include audits, monitoring of utilization and outcomes within the model, and the availability of Quality Improvement Organization (QIOs) and 1-800-MEDICARE for reporting beneficiary concerns, among other protections. The CJR model includes monitoring to ensure beneficiary access, choice, and quality of care is maintained under the model. We refer readers to section III.F. of this final rule for discussion of beneficiary protections and monitoring under the CJR model. The model pricing structure, discussed in III.C. of this final rule, also includes features to protect against such potential harm, such as responsibility for post-episode spending increases, stop-gain policies that set a maximum threshold a hospital can earn for savings achieved during episodes, and other policies as detailed in that section. In summary, we note that this payment model does not constrain the practice of medicine and we do not expect clinical decisions to be made on the basis of the payment amount.

Comment: Several commenters stated that all states selected to participate in the proposed HHVBP should be exempted from having to participate in the CJR model. Commenters stated that forcing HHAs to participate in two mandatory models simultaneously is harsh and punitive and would likely skew the results of both models in areas of overlap.

Response: Only participant hospitals under the CJR model are financially responsible to CMS for the episode of care. HHAs will continue to be paid the FFS amount that they would otherwise receive for beneficiaries included in the CJR model. Therefore, there is no reason to exempt hospitals located in MSAs selected for participation in CJR that are also located in states selected for participation in the HHVBP model.

Comment: Many commenters expressed concern with the interaction between BPCI and the proposed joint replacement model due to instances where LEJR episodes excluded from CJR due to BPCI would cause a low volume issue for certain hospitals. Other commenters stated that the proposed CJR model penalizes providers that are voluntarily participating in the BPCI initiative and suggested that CMS allow hospitals in selected MSAs to be allowed to choose between participation in BPCI and the joint replacement model.

Response: Because there are LEJR episodes currently being tested in BPCI Models 1, 2, 3 and 4, we noted in the proposed rule that we believed that participation in CJR should not be required if it would disrupt testing of LEJR episodes already underway in BPCI models. Therefore, we proposed that IPPS hospitals located in an area selected for the model that are active Model 1 BPCI participant hospitals as of July 1, 2015, or episode initiators for LEJR episodes in the risk-bearing phase of Model 2 or 4 of BPCI as of July 1, 2015, would be excluded from participating in CJR during the time that their qualifying episodes are included in one of the BPCI models. We clarify that we will utilize current information on BPCI participation to determine whether a given hospital is included in CJR. For example, if a hospital elected to participate in the LEJR episode under BPCI Model 2 in September 2015, that hospital would not be included in CJR during the time that their qualifying episodes are included in BPCI. Likewise, we proposed that if the participant hospital is not an episode initiator for LEJR episodes under BPCI Model 2, then LEJR episodes initiated by other providers or suppliers under BPCI Model 2 or 3 (where the surgery takes place at the participant hospital) would be excluded from CJR. Otherwise qualifying LEJR episodes (that is, those that are not part of a Model 3 BPCI LEJR episode or a Model 2 PGP-initiated LEJR episode) at the participant hospital would be included in CJR. We are testing a model where participation is not voluntary; therefore, it would not be appropriate for hospitals in selected MSAs to be allowed to choose between participation in BPCI and the joint replacement model. If hospitals were allowed to voluntarily participate in the CJR model, this would introduce selection bias and hamper CMS' ability to analyze how such a payment model potentially would work on a national scale. In addition, a hospital interested in participating in a voluntary model had the opportunity under BPCI. In response to concerns regarding the interaction between BPCI and CJR and potential for too few LEJR episodes at a given hospital to remain under the CJR model, low volume concerns are discussed and addressed in section III.A.4.b of this final rule.

Comment: Some commenters requested CMS to allow hospitals participating in ACOs that achieved shared savings in recent performance periods, Shared Savings Program ACOs (Track 2 and Track 3), and full-risk ACOs (such as Next Generation ACO), to opt-out of participation in the CJR model.

Response: As we previously noted and in the proposed rule, many hospitals have recently heightened their focus on aligning their efforts with those of community providers to provide an improved continuum of care due to the incentives under other CMS models and programs. Therefore, hospitals that are already involved in ACO initiatives and Start Printed Page 73288the HRRP have already established a base for augmenting these efforts under the CJR model (80 FR 41205). Therefore, we see no compelling reason why hospitals participating in ACO initiatives and other efforts cannot be participant hospitals in the CJR model. However, adjustments to account for overlaps with other innovation center models and CMS programs are discussed in section III.C.7. of this final rule.

Comment: A commenter stated that a CMS Certification Number (CCN) can include multiple hospitals. The commenter inquired, if at least one hospital under the CCN is in a selected MSA, would the entire CCN be required to participate in the CJR model. The commenter also requested if some of the hospitals in the CCN are not eligible for the CJR program, would they be required to participate because they are under the same CCN.

Response: The proposed approach indicated that CMS would base selection on the physical location of the hospital. The manner in which CMS tracks and identifies hospitals is through the CCN. In keeping with this approach, the CJR model will administer model-related activities at the CCN level including the determination of physical location. The physical location associated with the CCN at the time of the model start will be used to determine whether that CCN is located in a selected MSA. For hospitals that share a CCN across various locations, all hospitals under that CCN would be required to participate in the CJR model if the physical address associated with the CCN is in the MSA, unless otherwise excluded. Similarly, all hospitals under the same CCN, even if some are physically located in the MSA selected for participation, would not participate in in the CJR model if the physical address associated with the CCN is not in the MSA. Our analysis of the hospitals in the selected MSAs indicates that this phenomenon is not present in the selected areas.

Final Decision: After consideration of the public comments we received, we are finalizing the proposal to designate IPPS hospitals as the episode initiators. The initiation of an episode is described in § 510.100. We are also finalizing our proposal to require IPPS hospitals physically located in an area selected for participation in the CJR model, according to the address associated with the CCN, to participate in the model and bear the financial responsibility for LEJR episodes of care under the CJR model. Finally, we are finalizing our proposal that hospitals selected for the model that are active Model 1 BPCI participant hospitals as of July 1, 2015, or episode initiators for LEJR episodes in the risk-bearing phase of Model 2 or 4 of BPCI as of October 1, 2015, are excluded from participating in CJR during the time that their qualifying episodes are included in one of the BPCI models. However, LEJR episodes initiated by other providers or suppliers under BPCI Model 2 or 3 (where the surgery takes place at the participant hospital) are excluded from CJR. Otherwise qualifying LEJR episodes (that is, those that are not part of a Model 3 BPCI LEJR episode or a Model 2 physician group practice-initiated LEJR episode) at the participant hospital are included in CJR. The definition of a “participant hospital” and “CJR-regional hospital” will be codified in § 510.2, exclusions to episodes being tested due to BPCI overlap will be codified in § 510.100(b). The following chart illustrates the inclusion of episodes in CJR relative to BPCI.

4. Geographic Unit of Selection and Exclusion of Selected Hospitals

In determining which hospitals to include in the CJR model, we considered whether the model should be limited to hospitals where a high volume of LEJRs are performed, which would result in a more narrow test on the effects of an episode-based payment, or whether to include all hospitals in particular geographic areas, which would result in testing the effects of an episode-based payment approach more broadly across an accountable care community seeking to coordinate care longitudinally across settings. Selecting certain hospitals where a high volume of LEJRs are performed may allow for fewer hospitals to be selected as model participants, but still result in a sufficient number of CJR episodes to Start Printed Page 73289evaluate the success of the model. However, there would be more potential for behavioral changes that could include patient shifting and steering between hospitals in a given geographic area that could impact the test. Additionally, this approach would provide less information on testing episode payments for LEJR procedures across a wide variety of hospitals with different characteristics. Selecting geographic areas and including all IPPS hospitals in those areas not otherwise excluded due to BPCI overlap as previously described and in section III.C.7. of the proposed rule as model participants would help to minimize the risk of participant hospitals shifting higher cost cases out of the CJR model. Moreover, in selecting geographic areas we could choose certain characteristics, stratify geographic areas according to these characteristics, and randomly select geographic areas from within each stratum. Such a stratified random sampling method based on geographic area would allow us to observe the experiences of hospitals with various characteristics, such as variations in size, profit status, and episode utilization patterns, and examine whether these characteristics impact the effect of the model on patient outcomes and Medicare expenditures within episodes of care. Stratification would also substantially reduce the extent to which the selected hospitals will differ from non-selected hospitals on the characteristics used for stratification, which would improve the statistical power of the subsequent model evaluation, improving our ability to reach conclusions about the model's effects on episode costs and the quality of patient care. Therefore, given the authority in section 1115A(a)(5) of the Act, which allows the Secretary to elect to limit testing of a model to certain geographic areas, we proposed to use a stratified random sampling method to select geographic areas and require all hospitals paid under the IPPS in those areas to participate in the CJR model and be financially responsible for the cost of the episode, with certain exceptions as previously discussed and in sections III.B.3 and III.C.7. of the proposed rule.

a. Overview and Options for Geographic Area Selection

In determining the geographic unit for the geographic area selection for this model, we considered using a stratified random sampling methodology to select—(1) Certain counties based on their Core-Based Statistical Area (CBSA) status, (2) certain zip codes based on their Hospital Referral Regions (HRR) status; or (3) certain states. We address each geographic unit in turn.

We considered selecting certain counties based on their CBSA status. A CBSA is a core area containing a substantial population nucleus, together with adjacent communities having a high degree of economic and social integration within that core. Counties are designated as part of a CBSA when the county or counties or equivalent entities are associated with at least one core (urbanized area or urban cluster) of at least 10,000 in population, plus adjacent counties having a high degree of social and economic integration with the core as measured through commuting ties with the counties associated with the core. There are 929 CBSAs currently used for geographic wage adjustment purposes across Medicare payment systems.[4] The 929 CBSAs include 388 MSAs, which have an urban core population of at least 50,000, and the 541 Micropolitan Statistical Areas (μSA), which have an urban core population of at least 10,000 but less than 50,000. CBSAs may be further combined into a Combined Statistical Area (CSA) which consists of two or more adjacent CBSAs (MSAs or μSAs or both) with substantial employment interchange. Counties not classified as a CBSA are typically categorized and examined at a state level.

The choice of a geographical unit based on CBSA status could mean selection of a CBSA, an MSA, or a CSA. We proposed basing the selection on an MSA, which we will discuss later in this section.

We proposed that counties not in an MSA would not be subject to the selection process. These counties not subject to selection would include the μSA counties and the counties without a core urban area of at least 10,000. These areas are largely rural areas and have a limited number of qualifying LEJR cases. Relatively few of these areas would be able to qualify for inclusion based on the minimum number of LEJR episodes in year requirement discussed later in this section.

We considered, but ultimately decided against, using CSA designation instead of MSAs as a potential unit of selection. Under this scenario, we would look at how OMB classifies counties. We would first assess whether a county has been identified as belonging to a CSA, a unit which consists of adjacent MSAs or μSAs or both. If the county was not in a CSA, we would determine if it was in an MSA that is not part of a larger CSA. Counties not associated with a CSA or an MSA would be unclassified and excluded from selection. These unclassified areas would include the counties in a state that were either not a CBSA (no core area of at least 10,000) or associated with a μSA (core area of between 10,000 and 50,000) but unaffiliated with a CSA.

Whether to select on the basis of CSA/MSAs or just on MSAs was influenced by a number of factors. We considered the following factors:

  • CSAs, by definition, have a significantly lower degree of interchange between component parts than the interchange experienced within an MSA. Thus, we did not believe that using CSAs would be necessary in order to capture referral patterns. A case study examination of the geographic areas included in CSAs with respect to the health care markets of those areas and their respective parts helped to validate our conclusion.
  • We assessed the anticipated degree to which LEJR patients would be willing to travel for their initial hospitalization.
  • We assessed the extent to which surgeons are expected to have admitting privileges in multiple hospitals located in different MSAs.
  • We considered the degree to which we desire to include hospitals within μSAs that are part of a larger CSA.

After examining these factors, we concluded that that the anticipated risk for patient shifting and steering between MSAs within a CSA was not severe enough to warrant selecting CSAs given CMS' preference for smaller geographic units. However, because MSAs are units with significant levels of social and commercial exchange and due to the mobility of patients and providers within MSAs, we believed that selecting complete MSAs is preferable to selecting metropolitan divisions of MSAs for inclusion in the CJR model. We use the metropolitan divisions to set wage indices for its prospective payment systems (PPSs). Of the 388 MSAs, there are 11 MSAs that contain multiple metropolitan divisions. For example, the Boston-Cambridge-Newton, MA-NH MSA is divided into the following metropolitan divisions:

  • Boston, MA.
  • Cambridge-Newton-Framingham, MA.
  • Rockingham County-Strafford County, NH.Start Printed Page 73290

The Seattle-Tacoma-Bellevue, WA MSA is divided into the following metropolitan divisions:

  • Seattle-Bellevue-Everett, WA.
  • Tacoma-Lakewood, WA.

We proposed selecting entire MSAs rather than sub-divisions within an MSA.

We next considered selecting HRRs. HRRs represent regional health care markets for tertiary medical care. There are 306 HRRs with at least one city where both major cardiovascular surgical procedures and neurosurgery are performed. HRRs are defined by determining where the majority of patients were referred for major cardiovascular surgical procedures and for neurosurgery.[5] Compared to MSAs, HRRs are classified based on where the majority of beneficiaries within a zip code receive their hospital services for selected tertiary types of care. The resulting HRRs represent the degree to which people travel for tertiary care that generally requires the services of a major referral center and not the size of the referral network for more routine services, such as knee and hip arthroplasty procedures. In addition, because HRRs are defined based on referrals for cardiovascular surgical procedures and neurosurgery, they may not reflect referrals for orthopedic procedures. Therefore, we believed that MSAs as a geographic unit are preferable over HRRs for this model.

We also considered selecting states for the CJR model. However, we concluded that MSAs as a geographic unit are preferable over states for the CJR model. As stated in section III.A.4.b. of the proposed rule, we anticipate that hospitals that would otherwise be required to participate in the CJR model would be excluded from the model because their relevant LEJR episodes are already being tested in BPCI. If we were to select states as the geographic unit, there is a potential that an entire state would need to be excluded because a large proportion of hospitals in that state are episode initiators of LEJR episodes in BPCI. In contrast, if we excluded a specific MSA due to BPCI participation, as discussed in the next section, we could still select another MSA within that same state. Likewise, if we chose states as the geographic unit, we would automatically include hospitals in all rural areas within the state selected. If MSAs are selected for the geographic unit, we anticipate that fewer small rural hospitals would be included in the model. Using a unit of selection smaller than a state would allow for a more deliberate choice about the extent of inclusion of rural or small population areas. Selecting states rather than MSAs would also greatly reduce the number of independent geographic areas subject to selection under the model, which would decrease the statistical power of the model evaluation. Finally, MSAs straddle state lines where providers and Medicare beneficiaries can easily cross these boundaries for health care. Choosing states as the geographic unit would potentially divide a hospital market and set up a greater potential for patient shifting and steering to different hospitals under the model. The decision that the MSA-level analysis was more analytically appropriate was based on the specifics of this model and is not meant to imply that other levels of selection would not be appropriate in a different model such as the proposed HHVBP model.

For the reasons previously discussed, we proposed to require all IPPS hospitals to participate in the CJR model (with limited exceptions as previously discussed in section III.A.2. of the proposed rule) if located in an MSA selected through a stratified random sampling methodology (outlined in section III.A.3.b. of the proposed rule) to test and evaluate the effects of an episode-based payment approach for an LEJR episodes. We proposed to determine that a hospital is located in an area selected if the hospital is physically located within the boundary of any of the counties in that MSA where the counties are determined by the definition of the MSA as of the date the selection is made. In response to comments, we are clarifying that we will determine physical location using the address associated with the CCN of the hospital. Although MSAs are revised periodically, with additional counties added or removed from certain MSAs, we proposed to maintain the same cohort of selected hospitals throughout the 5 performance years of the model with limited exceptions as described later in this section. Thus, we proposed that, if after the start of the model, new counties are added to one of the selected MSAs or counties are removed from one of the selected MSAs, those re-assigned counties would retain the same CJR status they had at the beginning of the initiative. We believed that this approach will best maintain the consistency of the participants in the model, which is crucial for our ability to evaluate the results of the model. We retain the possibility of adding a hospital that is opened or incorporated within one of the selected counties after the selection is made and during the period of performance. (See section III.C.4. of the proposed rule for discussion of how target prices will be determined for such hospitals.) Hospitals in selected counties that do not have any LEJR cases that qualify for CJR, due to their participation in the BPCI initiative as a hospital initiator in an LEJR episode, will become subject to CJR at the time their participation in BPCI ends and their episodes become eligible for CJR. Although we considered including hospitals in a given MSA based on whether the hospitals were classified into the MSA for IPPS wage index purposes, this process would be more complicated, and we could not find any compelling reasons favoring this approach. For example, we assign hospitals to metro divisions of MSAs when those divisions exist. See our previous discussion of this issue. In addition, there is the IPPS process of geographic reclassification by which a hospital's wage index value or standardized payment amount is based on a county other than the one where the hospital is located. For the purpose of this model, it is simpler and more straightforward to use the hospital's physical location as the basis of assignment to a geographic unit. This decision would have no impact on a hospital's payment under the IPPS. We sought comment on our proposal to include participant hospitals for the CJR model based on the physical location of the hospital in one of the counties included in a selected MSA.

The following is a summary of the comments received and our responses.

Comment: Commenters generally supported MSAs as the unit of geographic selection. However, several had concerns regarding the particular circumstances of their MSAs. Some commenters stated that MSAs were too large and preferred the use of metropolitan regions for large urban areas such as New York City, while others expressed concern with the inclusion of rural portions of the MSA counties. Commenters addressing the rural providers within the selected MSAs questioned whether the inclusion of rural hospitals in the model was deliberate or whether CMS believed hospitals in rural areas should not be included in the CJR model. Other commenters expressed concern that MSAs were a smaller than ideal unit of selection and that selecting MSAs for the model would encourage practices such as funneling patients to hospitals outside a selected MSA for surgery in order to avoid inclusion in the model. Conversely, a commenter asserted that Start Printed Page 73291participation in the model would result in a competitive advantage for hospitals in a selected MSA through the use of gainsharing to recruit physicians to move referrals into a selected market. Some commenters were also concerned about patient shifting in or out of a selected MSA in areas where the MSA was part of a larger CSA, such as in the Atlanta CSA in which some, but not all, component MSAs were selected for participation in the CJR model.

Response: We first address the issue of the inclusion of the entirety of an MSA as the unit of selection rather than just the core urban area. This was a deliberate choice reflecting the fact that we seek to examine the performance of hospitals under CJR that could be considered rural, low volume, or outside the urban core. Inclusion of such hospitals in the model will give us insight on how the model functions in these areas and increase the potential generalizability of the model. The proposed rule proposed additional protections to selected classes of hospitals such as SCHs, Medicare-Dependent Hospitals (MDHs), and RCHs because we wanted to further protect these federally-recognized categories of vulnerable hospitals while including them in the model.

We chose MSAs as the unit of selection to balance the following considerations: The scope for shifting patients in or out of selected areas, our ability to observe the impact of the model in a variety of circumstances, and our preference to not use a geographic unit larger than strictly necessary to evaluate the model. We acknowledge that there are inevitably tradeoffs among these criteria. With respect to the choice of CSA versus MSA, a far greater number of commenters were concerned with the inclusion of rural providers than were concerned with their or their competitor's markets crossing the borders of MSAs within a CSA. By definition, CSAs have a lesser degree of the employment interchange than an MSA and basing the geographic unit of selection on a CSA would entail the possibility of selecting µSAs within CSAs. On balance, we believe it is appropriate to limit the extent of rural participation in CJR by confining it to rural areas within MSAs. We are sympathetic to concerns related to the experience of hospitals that are located near the borders between MSAs, but believed that those concerns did not outweigh these other considerations. In contrast, the density of populations and providers at the borders of these markets was one of the reasons that we decided to not proceed with allowing selection to be done based on metropolitan divisions for those 11 MSAs that were so sub-divided. Metropolitan divisions are very likely to have hospitals whose referral markets straddle divisions and their use as a unit would have had been problematic. After weighing the comments we continue to believe that MSAs are the most appropriate compromise position for the choice of geographic unit of selection.

Finally, we note that separate commenters stated that a hospital in a CJR selected county could be either at both a competitive advantage (for example, by providing an opportunity to attract physicians through gainsharing), or a competitive disadvantage (for example, by causing physicians to shift patients to nearby hospitals). We believe that both phenomena may occur and that the ability of a hospital to use the opportunities presented to it under the CJR model to strengthen its relationship with other providers and potentially achieve savings will vary by the hospital's specific circumstances and capabilities. We do not see a strong argument for why these types of effects necessitate a change to the geographic unit used for this model.

Comment: Some commenters contended that the CJR model has inadequate participation by small and rural providers due to the elimination of non-CBSA and µSAs from the possibility of selection for this model. The commenters wrote that CMS should include more rural providers in order to foster a model that is not overly tailored to large providers and urban areas. A commenter stated that inclusion in the model would result in rural providers being more prepared to adapt to future payment and delivery reforms. Another stated that it was important to include more small volume hospitals, and urged CMS to reconsider the implications of this exclusion and to broaden the definition of geographic areas.

Response: We appreciate commenters' input on how to incorporate rural providers in the CJR model and acknowledge commenters' concerns related to the ability of small and rural providers to effectively participate and succeed in the model. Our proposed approach to including low-volume and non-urban providers within the selected MSAs but removing from the possibility of selection counties that are not in an MSA or in an MSA with less than 400 qualifying LEJR cases is an appropriate strategy that allows for inclusion of rural providers in the model, while not oversampling such providers.

Comments related to requests for exclusion of particular hospitals are addressed in the next section, MSA Selection Methodology. Financial protections for hospitals are addressed later in section III.C.8. of this final rule.

Final Decision: After consideration of the public comments we received, we are finalizing the proposal, without modification, to utilize MSAs as the unit of selection for the model.

b. MSA Selection Methodology

We proposed to select the MSAs to include in the CJR model by stratifying all of the MSAs nationwide according to certain characteristics.

(1) Exclusion of Certain MSAs

Prior to assigning an MSA to a selection stratum, we examined whether the MSA met specific proposed exclusion criteria. MSAs were evaluated sequentially using the following 4 exclusion criteria: First, MSAs in which fewer than 400 LEJR episodes (determined as discussed in section III.B.2. of this final rule) occurred from July 1, 2013 through June 30, 2014 were removed from possible selection. The use of the 400 LEJR cases in a year was based on a simple one-sided power calculation to assess the number of episodes that would be needed to detect a 5 percent reduction in episode expenditures. Cases in hospitals paid under either the critical access hospital (CAH) methodology or the Maryland All-Payer Model are not included in the count of eligible episodes. This criterion removed 156 MSAs from possible selection.

Second, MSAs were removed from possible selection if there were fewer than 400 non-BPCI LEJR episodes in the MSA in the reference year. For the purposes of this exclusion, the number of BPCI episodes was estimated as the number of potentially eligible cases during the reference year that occurred in acute care hospitals participating in BPCI Model 1, or in phase 2 of BPCI Models 2 or 4 as of July 1, 2015 and the number of LEJRs in the reference year associated with these hospitals was examined. This criterion removed an additional 24 MSAs from potential selection.

Third, MSAs were also excluded from possible selection if the MSA was dominated by BPCI Models 1, 2, 3, or 4 episodes to such a degree that it would impair the ability of participants in either the CJR model or the BPCI models to succeed in the objectives of the initiative or impair the ability to set accurate and fair prices. We anticipate that some degree of overlap in the two models will be mutually helpful for both models. There are two steps to this exclusion. First, we looked at the number of LEJR episodes at BPCI Model 1, 2 or 4 initiating hospitals and second, Start Printed Page 73292the number of LEJR episodes among BPCI Model 3 SNF and Home Health Agency (HHA) episode initiators. First, we excluded MSAs if more than 50 percent of otherwise qualifying proposed CJR episodes were in Phase 2 of BPCI Model 2 or 4 with hospital initiators. Second, we excluded MSAs if either SNF or HHA BPCI Model 3 initiating providers accounted for more than 50 percent of LEJR referrals to that provider type. As a result of this third criterion, 4 additional MSAs were removed from possible selection. No MSAs were excluded based on SNF or HHA participation in Model 3.

Finally, MSAs were removed if, after applying the previous three criteria they remained eligible for selection, but more than 50 percent of estimated eligible episodes during the reference year were not paid under the IPPS system. The purpose of this rule was to assess the appropriateness of MSAs that contained both Maryland and non-Maryland counties. No MSAs were eliminated on the basis of this rule. Please refer to the appendix for this final rule for the status of each MSA based on these exclusion criteria, available at http://innovation.cms.gov/​initiatives/​cjr/​. After applying these four exclusions, 196 MSAs remained to be stratified for purposes of our proposed selection methodology.

The following is a summary of the comments received and our responses.

Comment: Many commenters requested that we exclude additional MSAs from the selection process. Commenters supported our exclusion of MSAs with less than a minimum number of eligible LEJR episodes and a high rate of BPCI LEJR penetration, but were concerned that the list of BPCI participating providers used in making the exclusion determination did not reflect providers entering BPCI as of October 1, 2015. Such commenters recommended that CMS recalculate BPCI participation in LEJR episodes in each MSA based on both hospital- and physician-led participants and adjust the MSA selection accordingly. Commenters also suggested adding additional selection criteria based on the overall percent of LEJR episodes associated with a BPCI episode, the percent of LEJR episodes associated with a PGP initiated BPCI episode, and the percent of LEJR episodes associated with an ACO.

Response: In response to the comments, we re-examined the exclusion rules based on an updated list of providers participating in the BPCI initiative for LEJR episodes. We also examined the potential impact on selection of MSAs that incorporating an updated list of BPCI participants would have. For the purposes of the re-examination of exclusion rule 2, which eliminates MSAs with less than 400 CJR eligible, non-BPCI episodes, we estimated the BPCI LEJR episode count as the number of potentially eligible cases during the reference year that (1) occurred in an acute care hospital participating in BPCI Model 1 that would still be active as of April 1, 2016; (2) occurred in an acute care hospital in a Phase 2 LEJR episode for BPCI Models 2 or 4 as of October 1, 2015; or (3) were associated with an operating or admitting physician on the hospital claim assigned to a PGP with an LEJR episode in Phase 2 of BPCI Model 2 as of October 1, 2015. October 1, 2015 is the final quarter for which participants in Phase 1 of BPCI could transition any episode into Phase 2. This represents a change to the exclusion rule articulated in the proposed rule, in that it updates the list of BPCI participants and also takes into account episodes associated with Model 2 PGP episode initiators. As we did for exclusion rule 2, we used the October 2015 list of BPCI participants to reassess exclusion rule 3. Rule 3 removes an MSA if more than 50 percent of patients were treated in a BPCI initiating hospital or if more than 50 percent of LEJR patients treated in a PAC setting of that type were treated in a BPCI initiating HHA or SNF.

After we made the previously stated changes, some MSAs previously eligible for selection would now be considered excluded. Additionally, two of the MSAs previously excluded would now be eligible for selection due to hospitals withdrawing from BPCI and the MSAs now having more than 400 eligible cases. Eight MSAs that were selected in the proposed rule would be classified as excluded on the basis of these updated exclusion rules.

We considered a variety of alternative approaches to address the changes in the eligibility of MSAs. First, we considered proceeding with the list of 75 MSAs as initially selected and using the exclusion rules as initially proposed. Second, we considered removing the 8 selected MSAs that would now be excluded on the basis of the updated BPCI participation numbers. Third, we considered replacing the 8 MSAs by randomly selecting new MSAs from the remaining MSAs in the relevant strata. However, we believed that it would preferable, although not required, to give the selected MSAs a consistent period of time between selection and the start of the model. Fourth, we contemplated creating a revised list of eligible MSAs and randomly selecting a new group of 75 MSAs. Given the concern of many commenters about the start date of the model, we were reluctant to create a completely new list of selected MSAs. We believe that making a new selection would be regarded unfavorably by impacted MSAs and hospitals and should be avoided if possible. In order to be responsive to concerns regarding the growth of BPCI after the publication of the proposed rule and the increase in PGP participation in BPCI, we are proceeding with the second option.

The function of the stratification approach was to ensure that our selection of MSAs covered a range of efficiency levels and population sizes and allowed us to target our sampling percentages so as to oversample in the less efficient areas. Regarding the selected MSAs now eliminated, they are distributed fairly evenly throughout the distribution of average episode payments. From the least expensive to the most expensive quartiles, the number selected and now eliminated are, in order, 2/15 (13 percent), 2/19 (11 percent), 3/30 (15 percent), and 1/22 (5 percent). We also believe that the removal of these 8 MSAs from the model will not preclude us from undertaking a rigorous statistical evaluation of the model.

Given the aforementioned information, we believe that the relatively minor reduction in statistical power due to not re-selecting MSAs is outweighed by the desire to give affected participant hospitals equal time to prepare for the model. We are removing the 8 MSAs as noted in Table 1.

Table 1—MSAs That Were Previously Selected That Are No Longer Included in CJR

CBSA titleRevised exclusion rule 2 statusRevised exclusion rule 3 status
Colorado Springs, COFailPass.
Evansville, IN-KYFailPass.
Fort Collins, COFailPass.
Las Vegas-Henderson-Paradise, NVFailFail.
Start Printed Page 73293
Medford, ORFailPass.
Richmond, VAFailPass.
Rockford, ILFailPass.
Virginia Beach-Norfolk-Newport News, VA-NCPassFail.

We next contemplated whether to apply additional MSA-level exclusion rules. We investigated a potential new rule whereby an MSA would be excluded based on the percent of the MSA's qualifying LEJR episodes associated with Phase 2 Model 2 PGP initiators. We did not believe that there was as strong of an argument for excluding MSAs on the basis of the percent of patients treated by a BPCI physician given that the hospital is the financially accountable entity in CJR. We examined two possible cut off points (>65 percent and >50 percent) to assess which MSAs would be eliminated if we were to exclude MSAs where a specific percent of an MSA's otherwise qualifying LEJR cases was attributable to a BPCI PGP. At 65 percent, no selected MSAs not otherwise excluded were impacted. 8 MSAs that were previously selected had more than 50 percent of their LEJRs performed by BPCI PGPs. Five of these 8 MSAs are already eliminated due to the revised exclusion rule 2. For markets with more than 400 non-BPCI cases but more than 50 percent BPCI PGP penetration, the number of the CJR eligible patients was between 556 and 1834 indicating that there was a sizable number of cases. Consequently, we did not find this new exclusion rule necessary.

Comment: Commenters requested modifications to the proposed exclusions of specific categories of hospitals within an MSA. While commenters stated a variety of concerns, many of them were related to the request that CMS exclude low volume hospitals from the model. Commenters made requests around specific categories of hospitals including Medicare Dependent Hospitals (MDHs), Sole Community Hospitals (SCHs), Rural Referral Centers (RRCs), hospitals that are reclassified as rural, hospitals perceived of as rural or outside of a core urban area, and larger hospitals with a low potential CJR LEJR volume due to the exclusion of their patients because their LEJR episodes were initiated by a PGP BPCI LEJR episode initiator.

Commenters provided a variety of rationales for why they believed it was undesirable or unfair to include low volume providers in the model. These reasons include, but are not limited to, observations that—

  • Low-volume providers are less likely to be proficient at taking care of these patients in an efficient cost-effective manner and they will be less likely to achieve savings;
  • Low-volume hospitals will be disproportionately impacted by outlier cases and will have less predictable cost and quality outcomes making it difficult for them to manage the model effectively. In addition, low volume providers are likely to see a greater proportion of hip fractures and non-planned procedures;
  • Low-volume hospitals will have less control over and ability to impact the behavior of other providers. The pool of collaborating providers such as orthopedic surgeons in most rural communities may be limited and small hospitals may not have the market position to successfully influence others' behavior;
  • Hospitals with a limited number of Medicare hip and knee procedures may not have sufficient incentive to invest the time and resources necessary to develop the infrastructure and partnerships required to effectively manage these episodes of care and may not find the opportunity to improve patient outcomes significant enough to engage referring physicians and PAC partners for redesign;
  • Low volume providers may be more financially vulnerable and with fewer resources to design and carry out initiatives or make effective responses to the financial incentives in the model. A commenter noted concerns with hospital margins, and the possibility for the reductions in revenue as a result of the loss of volume or loss of margin under CJR could result in additional hospital closures.

Due to these concerns, commenters requested a variety of solutions including (1) the exclusion of hospitals based on a volume cut off variously defined by volume of eligible LEJR cases, LEJR cases within specific MS-DRGs and total hospital volume, (2) making the model voluntary for low volume providers, (3) extending the protections afforded to SCH, MDH and RRC to additional categories of hospitals including hospitals electing to be treated as rural under § 412.103, and (4) the provision of additional protections or payment adjustments beyond what was included in the proposed rule.

Response: We acknowledge the fact that hospitals, particularly low volume hospitals, are concerned and would like to increase their probability of receiving reconciliation payments under CJR while minimizing the possibility of reduction in revenue. We refer readers to the following sections of this final rule: Section III.C.8. for a discussion of hospital financial protections, III.C.4. for a discussion of how we will determine target prices for hospitals with low volume, and section III.C.4. for a discussion of target prices for hip fracture patients. We believe that the modification of the treatment of hip fractures in the payment methodology should allay many concerns of small and rural providers. This change may disproportionately impact them since emergency surgeries, such as hip fractures, have a higher probability of being performed in low volume settings.

As stated in relation to comments requesting that CJR operate as a voluntary model, the inclusion of low volume hospitals in the CJR model is consistent with the goal of evaluating the impact of bundled payment and care redesign across a broad spectrum of hospitals with varying levels of infrastructure, care redesign experience, market position, and other considerations and circumstances. The design of the CJR model and the inclusion of low volume providers within the model reflects our interest in testing and evaluating the impact of a bundled payment approach for LEJR procedures in a variety of circumstances, especially among those hospitals that may not otherwise participate in such a test. The inclusion of these providers allows CMS to better appreciate and understand how the model operates as a general payment approach and its impact on a wide range of hospitals. Many LEJR surgeries are performed in low volume settings, thus, the impact of the CJR model on low volume hospitals is of great interest to the evaluation of this initiative.

We acknowledge that providers with low volumes of cases may not find it in their financial interests to make Start Printed Page 73294systematic care redesigns or engage in an active way with the CJR model. We expect that low volume providers may decide that their resources are better targeted to other efforts because they do not find the financial incentive present in the CJR sufficiently strong to cause them to shift their practice patterns. We acknowledge that low volume hospitals may achieve less savings because they did not or could not make the necessary changes to the treatment of their qualifying beneficiary population. We believe this choice is similar in nature to that made as hospitals decide their overall business strategies and where to focus their attentions.

Comment: Many commenters requested that CMS exclude hospitals where more than 50 percent of the eligible LEJRs performed at a hospital would be attributed to a PGP initiated BPCI episode and would thus not be in CJR. The majority of these commenters were concerned about low volumes of patients, which is addressed in the previous comment and response. Some were concerned about the operational complexity of identifying, tracking, and managing patients treated in CJR versus BPCI.

Response: We will not exclude IPPS hospitals in selected MSAs other than as already specified or allow IPPS hospitals to opt out of participation in CJR. As previously noted in the discussion on low volume hospitals, we consider the inclusion of low volume providers a core feature of the model that will aid us in understanding the impact of a variety of providers in various circumstances. Similarly, we do not believe it is necessary or appropriate to exclude hospitals on the basis of some of the surgeons in their hospitals being associated with a BPCI PGP. Like with more traditional low volume providers, the extent to which a hospital alters its behavior in response to the CJR model will likely be the result of a variety of factors including but not limited to the anticipated number of cases. It should be noted that the revised exclusion rule that resulted in the elimination of 8 MSAs was based on failing to meet a minimum MSA number of LEJRs and not based on either the number of LEJRs at a particular hospital or the portion of PGPs at any level of analysis. If an IPPS hospital in a selected area has some of their LEJR cases qualify as CJR episodes and some that do not due to BPCI participation, Medicare Advantage status or any other reason, the fact that CJR cases are not their full caseload will not be considered a reason for exclusion of the hospital.

With respect to challenges that hospitals may experience related to identifying eligible patients and following them over the course of their episodes, we acknowledge that concern. However, we consider the improved tracking and communication with other providers and suppliers that is likely to occur as a result of hospital efforts in CJR to be a benefit of the model that will improve the coordination of patient care and possibly improve patient outcomes.

Comment: Two commenters raised the issue of hospital systems spanning more than one MSA. They requested that CMS either allow all of the hospitals in the system to be included in CJR or allow all of the hospitals to be excluded. Commenters stated that the additional administrative burden associated with two concurrent Medicare payment methodologies would be unduly burdensome. Additionally, commenters stated that CMS should develop criteria under which all providers in health systems with a significant number of BPCI participants would be excluded from the CJR model due to operational challenges to managing the BPCI and CJR models simultaneously within a health care system.

Response: With respect to the request that all members within a health system be allowed to have all of their hospitals participate in BPCI because operating under two systems is too onerous, if a health system made the choice to enter some but not all of their locations into BPCI, they have already made the business decision to operate partly under one incentive structure and partly under another. We do not believe that the existence of CJR model as proposed should change the timelines for transitioning to Phase 2 of BPCI. We will not exclude hospitals from the model on the basis that some of the hospitals in its health system are participating in BPCI or some of the hospitals in its health system have CCNs with addresses located in a non-selected MSA.

The CJR model will require hospitals within selected geographic areas to participate (unless otherwise excluded as set forth in this final rule). The inclusion of additional voluntarily participating hospitals outside of these selected areas would constitute a major change to the model that was not considered in the proposed rule. Providers who wished to participate in a voluntary episode model had the opportunity under the BPCI initiative.

Final Decision: After consideration of the public comments we received, we are modifying the MSA exclusion rules used in determining which MSAs are eligible for selection. The following is a description of the MSA exclusion criteria used in this final rule:

In determining if an MSA was eligible for selection, we first examined whether the MSA met any of the four exclusion criteria as formulated in the proposed rule. This process resulted in a pool of 196 MSA from which we then selected 75 for inclusion in CJR via stratified random selection.

In this final rule, we revised the exclusion rules as defined later in this section, with the purpose of assessing whether any of the 75 selected MSAs would be considered not eligible for selection based on applying the new criteria.

Specifically, the second exclusion rule, which eliminates MSAs with fewer than 400 non-BPCI CJR eligible cases, is modified with the following additions (1) the determination of the count of patients associated with a BPCI Phase 2 initiating hospital is based on the participation in BPCI as of October 1, 2015 rather than July 1, 2015 and (2) the count of BPCI episodes to be removed from the count of eligible episodes takes into consideration patients who would have been attributed to a BPCI Model 2 initiating PGP in Phase 2 for an LEJR episode as of October 1, 2015. The third exclusion rule, wherein MSAs were excluded based on the percent of the MSA's LEJR population associated with either a BPCI hospital, SNF or HHA in an MSA, was changed to be based on episodes associated with participation in BPCI as of October 1, 2015 rather than July 1, 2015.

As a result of updating the list of BPCI participants to those entering the model in October 2015 and including Phase 2 PGPs in the calculation of the number of cases in the MSA, 8 MSAs out of the 75 MSAs that were previously selected are now deemed not eligible for selection and are consequently no longer required to participate in CJR. These previously selected and now excluded MSAs are shown in Table 1. The remaining 67 MSAs selected in the proposed rule will be required to participate in CJR.

(2) Selection Strata

Numerous variables were considered as potential strata for classifying MSAs included in the model. However, our proposal was intended to give priority to transparency and understandability of the strata. We proposed creating selection strata based on the following two dimensions: MSA average wage-adjusted historic LEJR episode payments and MSA population size.Start Printed Page 73295

(a) MSA Average Wage-Adjusted Historic LEJR Episode Payments

We were interested in being able to classify and divide MSAs according to their typical patterns of care associated with LEJR episodes. As a straightforward measure of LEJR patterns of care, we selected the mean MSA episode payment, as defined in the proposed rule. MSAs vary in their average episode payments. The average episode payments in an area may vary for a variety of reasons including—(1) In response to the MS-DRG case mix and thus the presence of complicating conditions; (2) readmission rates; (3) practice patterns associated with type of PAC provider(s) treating beneficiaries; (4) variations of payments within those PAC providers, and (5) the presence of any outlier payments.

The measure of both mean episode payments and median episode payments within the MSA was considered. We proposed to stratify by mean because it would provide more information on the variation in episode payments at the high end of the range of payments. We are interested in the lower payment areas for the purpose of informing decisions about potential future model expansion. However, the CJR model is expected to have the greatest impact in areas with higher average episode payments.

The average episode payments used in this analysis were calculated based on the proposed episode definition for CJR using Medicare claims accessed through the Chronic Conditions Warehouse for 3 years with admission dates from July 1, 2011 through June 30, 2014. Episode payments were wage-adjusted using the FY 2014 hospital wage index contained in the FY 2014 IPPS Final Rule, downloaded at http://www.cms.gov/​Medicare/​Medicare-Fee-for-Service-Payment/​AcuteInpatientPPS/​FY-2014-IPPS-Final-Rule-Home-Page-Items/​FY-2014-IPPS-Final-Rule-CMS-1599-F-Data-Files.html. The adjusted payment was calculated by dividing the unadjusted payment by a factor equal to the sum of 0.3 plus the multiplicative product of 0.7 and the wage index value of the hospital where the LEJR was performed. We truncated the episode payment at the 99.9th percentile of the distribution ($135,000) to limit the impact of extreme outliers.

(b) MSA Population Size

The second dimension proposed for the CJR selection strata is the number of persons in the MSA. In deciding how best to incorporate the dimensions of urban density and availability of medical resources, a variety of measures were considered, including overall population in the included counties, overall population in the core area of the MSA, population over the age of 65 in the MSA, the number of hospital beds and the number of Medicare FFS LEJR procedures in a year. The reason we decided to include this dimension in the strata definition is that these factors are believed to be associated with the availability of resources and variations in practice and referral patterns by the size of the healthcare market. When examined, these alternative measures were all very highly correlated with one another, which allowed the use of one of these measures to be able to substitute for the others in the definition of the stratum. From these alternative approaches, we choose to use MSA population. In operationalizing this measure, MSAs were classified according to their 2010 census population.

(c) Analysis of Strata

The two proposed domains, MSA population and MSA historic LEJR episode spending, were examined using a K-Means factor analysis. The purpose of this factor analysis was to inform the process of which cut points most meaningfully classify MSAs. Factor analysis attempts to identify and isolate the underlying factors that explain the data using a matrix of associations. Factor analysis is an interdependence technique. Essentially, variables are entered into the model and the factors (or clusters) are identified based on how the input variables correlate to one another. The resulting clusters of MSAs produced by this methodology suggested natural cut points for average episode payments at $25,000 and $28,500. While not intentional, these divisions correspond roughly to the 25th and 75th percentiles of the MSA distribution. Cut points based on these percentiles seemed reasonable from statistical and face validity perspectives in the sense that they created groups that included an adequate number of MSAs and a meaningful range of costs.

As a result of this analysis, we classified MSAs according to their average LEJR episode payment into four categories based the on the 25th, 50th and 75th percentiles of the distribution of the 196 potentially selectable MSAs as determined in the exclusion rules as applied in the proposed rule (80 FR 41198). This approach ranks the MSAs relative to one another and creates four equally sized groups of 49. The population distribution was divided at the median point for the MSAs eligible for potential selection as determined and defined in the proposed rule. This resulted in MSAs being divided into two equal groups of 98. The characteristics of the resulting strata are shown in Table 2.

Table 2—Summary Population and Episode Payment Statistics by MSA Group

Payment in lowest quarterPayment in 2nd lowest quarterPayment in 3rd lowest quarterPayment in highest quarterTotal eligible
MSAs deemed eligible in the proposed rule (80 FR 41198) with population less than median:
Number of Eligible MSAs3319222498
Average of Population251,899238,562268,331254,154253,554
Minimum MSA Population96,27555,274106,33196,02455,274
Maximum MSA Population425,790416,257424,858428,185428,185
Average Episode Payments ($)$22,994$25,723$27,725$30,444$26,410
Minimum Episode Payments$18,440$24,898$26,764$29,091$18,440
Maximum Episode Payments$24,846$26,505$28,679$32,544$32,544
MSAs deemed eligible in the proposed rule (80 FR 41198) with population more than median:
Number of Eligible MSAs1630272598
Average of Population1,530,0831,597,8701,732,5252,883,9661,951,987
Minimum MSA Population464,036436,712434,972439,811434,972
Maximum MSA Population4,335,3915,286,72812,828,83719,567,41019,567,410
Average Episode Payments ($)$23,192$25,933$27,694$30,291$27,082
Start Printed Page 73296
Minimum Episode Payments$16,504$25,091$26,880$28,724$16,504
Maximum Episode Payments$24,819$26,754$28,659$33,072$33,072
Total Eligible MSAs49494949
Note: Population and episode payment means are unweighted averages of the MSA values within each of the eight MSA groups.

Please refer to the addenda for this final rule for information on the non-excluded MSAs, their wage adjusted average LEJR episode spending, their population and their resultant group assignment at: http://innovation.cms.gov/​initiatives/​cjr/​.

(3) Factors Considered But Not Used in Creating Proposed Strata

In addition to the two dimensions we proposed to use for the selection groups previously discussed, a variety of possible alternative measures and dimensions were considered. Many of these variables are considered to be important but it was believed that it was important to have a fairly straightforward and easily understandable stratum definition. Simplicity, by definition, required that only the most important variables would be used. If a market characteristic under consideration was correlated with one of the chosen dimensions or it was believed that variations in the characteristic could be adequately captured by random selection within the strata, is was not prioritized for inclusion.

Some of the factors considered that we did not propose as dimensions are—

  • Measures associated with variation in practice patterns associated with LEJR episodes. In considering how to operationalize this measure, a number of alternatives were considered including total PAC LEJR payments in an MSA, percent of LEJR episodes with a SNF claim in an MSA, percent of LEJR episodes with an initial discharge to HHA, percent of LEJR episodes with an Inpatient rehabilitation facility (IRF) claim, and percent of LEJR episodes with claims for two or more types of PAC providers;
  • Measures associated with relative market share of providers with respect to LEJR episodes;
  • Healthcare supply measures of providers and suppliers in the MSA including counts of IRF beds, SNF beds, hospital beds, and number of orthopedic surgeons;
  • MSA level demographic measures such as; average income, distributions of population by age, gender or race, percent dually eligible, percent of population with specific health conditions or other demographic composition measures; and
  • Measures associated with the degree to which a market might be more capable or ready to implement care redesign activities. Examples of market level characteristics that might be associated with anticipated ease of implementation include the MSA-level EHR meaningful use levels, managed care penetration, ACO penetration and experience with other bundling efforts.

It should be noted that, while these measures were not proposed to be part of the selection strata, we acknowledge that these and other market-level factors may be important to the proper understanding of the evaluation of the impact of CJR. It is the intention that these and other measures will be considered in determining which MSAs are appropriate comparison markets for the evaluation as well as considered for possible subgroup analysis or risk adjustment purposes. The evaluation will include beneficiary, provider, and market level characteristics in how it examines the performance of this proposed model.

(4) Sample Size Calculations and the Number of Selected MSAs

Analyses of the necessary sample size to facilitate a robust statistical analysis of CJR's effects led us to conclude that we needed to include between 50 and 100 MSAs of the 384 MSAs with eligible LEJR episodes to participate in CJR and we proposed to select 75 MSAs. As previously discussed, the proposed revision of the MSA exclusion rules resulted in 8 of the previously selected MSAs now being considered excluded, leading to their removal from the model. The resulting number of selected MSAs, 67, is still within the acceptable range for an MSA count as determined by our analysis. The number and method of selection of these original 75 MSAs from the 8 proposed groups is addressed in the following section. In finalizing this approach, we are undertaking a test in as few markets as possible while still allowing us to be confident in our results and to be able to generalize from the model to the larger national context. We discuss the assumptions and modeling that went into our proposal later in this section.

In calculating the necessary size of the model, a key consideration was ensuring that the model would have sufficient power to be able to detect the desired size impact. The larger the anticipated size of the impact, the fewer MSAs we would have to sample in order to observe it. However, a model sized to be able to only detect large impacts runs the risk of not being able to draw conclusions if the size of the change is less than anticipated. The measure of interest used in estimating sample size requirements for the CJR model was wage-adjusted total episode spending. To measure wage-adjusted total episode spending, we used the 3 year data pull also used for the average regional episode spending estimation that covers LEJR episodes with admission dates from July 1, 2011 through June 30, 2014. For the purposes of the sample size calculation the impact estimate assumed we wanted to be able to detect a 2 percent reduction in wage adjusted episode spending after 1 year of experience. This amount was chosen because it is the anticipated amount of the discount we proposed to apply to target prices in CJR.

The next consideration in calculating the necessary sample size is the degree of certainty we will need for the statistical tests that will be performed. In selecting the right sample size, there are two types of errors that need to be considered “false negatives” and “false positives”. A false positive occurs if a statistical test concludes that the model was successful when it was, in fact, not. A false negative occurs if a statistical test fails to find statistically significant evidence that the model was successful, but it was, in fact, successful. In considering the minimum sample size needs of a model, a standard guideline in the statistical literature suggests calibrating statistical tests to generate no more than a 5 percent chance of a false positive and selecting the sample size to ensure no more than a 20 percent Start Printed Page 73297chance of a false negative. In contrast, the proposed sample size for this project was based on a 20 percent chance of a false positive and a 30 percent chance of a false negative after one year of episodes in order to be as conservative as was practicable. A greater degree of certainty will be available with additional years of data.

A third consideration in the sample size calculation was the appropriate unit of selection and whether it is necessary to base the calculation on the number of MSAs, the number of hospitals, or the number of episodes. As discussed later in this section, we are proposing to base the sample size calculation at the MSA level.

The CJR model is a nested comparative study, which has two key features. First, the unit of assignment (to treatment and comparison groups) is an identifiable group; such groups are not formed at random, but rather through some physical, social, geographic, or other connection among their members. Second, the units of observation are members of those groups. In such designs, the major analytic problem is that there is an expectation for a positive correlation (intra-class correlation (ICC)) among observations of members of the same group (MSA). The ICC reflects an extra component of variance attributable to the group above and beyond the variance attributable to its members. This extra variation will increase the variance of any aggregate statistic beyond what would be expected with random assignment of beneficiaries or hospitals to the treatment group.

In determining the necessary sample size, we need to take into consideration the degrees of freedom. As part of this process, we examined the number of beneficiaries, the number of hospitals, and the number of MSAs and the level of correlation in episode payments between each level. For example, while each beneficiary has their own episode expenditure level, there are commonalities between those expenditure amounts at the hospital level, based on hospital-specific practice and referral patterns. The number of degrees of freedom needed for any aggregate statistic is related to the number of groups (MSAs or hospitals), not the number of observations (beneficiary episodes). If we were to base the determination of the size of the model on beneficiary episodes where correlation exists, we would have an inflated false positive error rate and would overstate the impact of the model. We empirically examined the level of correlation between beneficiaries and hospitals and between hospitals and MSAs and determined that the correlation was high enough to be of concern and necessitate an MSA level unit of selection.

Using the previous assumptions, a power calculation was run which indicated we would need between 50 and 150 treatment MSAs to be able to reliably detect a 2 percent reduction in payments after 1 year. The lower end of this range assumed that our evaluation approach could substantially reduce variation through regression adjustment and other types of statistical modeling. We anticipated that we would have adequate statistical power based on prior research results, but wanted to ensure that we did not have to achieve the “best possible” results from such modeling in order to draw conclusions. In order to allow for some degree of flexibility we proposed the selection of 75 MSAs. We narrowed the acceptable range to between 50 and 100 MSAs rather than 50 to 150 MSAs, based on the assumption that we will be able to substantially improve our estimates through modeling, and then chose a number near the middle of this reduced range. Due to the revised exclusion rules, we are proceeding with 67 MSAs, which we believe will provide adequate statistical power.

In assessing to what degree regression adjustment and other statistical adjustments could reduce the number of MSAs needed to generate statistically reliable results, it should be noted that calculations are based on the actual Medicare payments associated with episodes. Thus, the variation in payments associated with MS-DRG case mix, or other reasons are already captured in the methodology.

(5) Method of Selecting MSA

As previously discussed, we selected 75 MSAs from our proposed 8 selection groups and subsequently reduced this number to 67. In performing the initial MSA selection, we examined and considered a number of possible approaches including equal selection in each of the eight groups, equal selection in the four payment groups, selection proportionate to the number of MSAs in each group, and a number of approaches that differentially weighted the payment categories.

After consideration, we proposed a methodology that proportionally under-weighted more efficient MSAs and over-weighted more expensive MSAs was the most appropriate approach to fulfilling the overall priorities of this model to increase efficiencies and savings for LEJR cases while maintaining or improving the overall quality of care. This approach made MSAs in the lowest spending category less likely to be selected for inclusion. We thought this appropriate because the MSAs in the lowest expenditure areas have the least room for possible improvement and are already performing relatively efficiently compared to other geographic areas, which means that experience with the model in these areas may be relatively less valuable for evaluation purposes. At the same time, we believed it was important to include some MSAs in this group in order to assess the performance of this model in this type of circumstance. We also believed it was appropriate for higher payment areas to be disproportionately included because they are most likely to have significant room for improvement in creating efficiencies. We expect more variation in practice patterns among the more expensive areas. There are multiple ways an MSA can be more relatively expensive, including through outlier cases, higher readmission rates, greater utilization of physician services, or through PAC referral patterns. A larger sample of MSAs within the higher payment areas will allow for us to observe the impact of the CJR model on areas with these various practice patterns in the baseline period.

The method of disproportionate selection between the strata used was to choose 30 percent of the MSAs in the two groups in the bottom quarter percentile of the payment distribution, 35 percent of the MSAs in the two groups in the second lowest quartile, 40 percent in the third quartile, and 45 percent in the highest episode payment quartile. This proportion resulted in the selection of the 75 originally selected MSAs out of the 196 eligible. The number of MSAs originally chosen as well as the final selection counts within the eight selection groups is shown in Table 3.Start Printed Page 73298

Table 3—Number of MSAs To Be Chosen From the Eight Selection Groups

Payment in lowest quarterPayment in 2nd lowest quarterPayment in 3rd lowest quarterPayment in highest quarterTotal eligible MSAs
Selection Proportion30%35%40%45%
Less Than Median Population (Group #)(1)(2)(3)(4)
Number Eligible MSAs per Proposed Rule (80 FR 41198)3319222498
Proportion x Number9.96.658.810.8
Number initially selected from group10791137
Number finally selected from group8681133
More Than Median Population (Group #)(5)(6)(7)(8)
Number Eligible MSAs per NPRM1630272598
Proportion x Number4.810.510.811.25
Number initially selected from group511111138
Number finally selected from group51091034
Total Eligible MSAs per Proposed Rule (80 FR 41198)49494949196
Number initially selected1518202275
Number finally selected from group1316172167

We selected the proposed MSAs for the CJR model through random selection. In the proposed method of selection, each MSA was assigned to one of the eight selection groups previously identified. Based on this sampling methodology, SAS Enterprise Guide 7.1 software was used to run a computer algorithm designed to randomly select MSAs from each strata. SAS Enterprise Guide 7.1 and the computer algorithm used to conduct selection represents an industry standard for generating advanced analytics and provides a rigorous, standardized tool by which to satisfy the requirements of randomized selection. The key SAS commands employed include a “PROC SURVEYSELECT” statement coupled with the “METHOD=SRS” option used to specify simple random sampling as the sample selection method. A random number seed was generated for each of the eight strata by using eight number seeds corresponding to birthdates and anniversary dates of parties present in the room. The random seeds for stratum one through eight were as follows: 907, 414, 525, 621, 1223, 827, 428, 524. Note that no additional stratification was used in any of the eight groupings so as to produce an equal probability of selection within each of the eight groups. For more information on this procedure and the underlying statistical methodology, please reference SAS support documentation at: http://support.sas.com/​documentation/​cdl/​en/​statug/​63033/​HTML/​default/​viewer.htm#statug_​surveyselect_​sect003.htm/​ We also considered a potential alternative approach to this random selection in which we would generate a starting number within SAS and then choose every third MSA within a group starting at this point until the relevant number of MSAs were chosen. We opted to not utilize this feature for simplicity's sake and alignment with other randomization methodologies used for CMS models.

The selection of an MSA means that all hospitals are included whose address associated with their CCN is physically located anywhere within the counties that make up the MSA. By definition, the entire county is included in an MSA and hospitals that are in the relevant counties will be impacted even if they are not part of the core urban area.

We stated in the proposed rule, should the methodology we propose in this rule change as a result of comments received during the rulemaking process, it could result in different areas being selected for the model. In such an event, we would apply the final methodology and announce the selected MSAs in the final rule. Therefore we sought comment from all interested parties in every MSA on the randomized selection methodology proposed in this section.

The following is a summary of the comments received and our responses.

Comment: Two commenters raised concerns regarding the number of MSAs selected for inclusion in the model. One noted that, given the range between 50 and 150 treatment MSAs to be able to reliably detect a 2 percent reduction in payments, CMS could drop some of the 75 selected MSAs without jeopardizing the ability to produce generalizable results from the CJR model. Another commenter suggested that the approach to the model should focus on an intense analysis within fewer markets prior to expansion into a larger representative sample.

Response: As discussed in the proposed rule, a variety of considerations were made in the determination of what would be an appropriate sample size. The initially proposed 75 MSAs represented the 25 percentage points of the acceptable range of MSAs to be included as determined by sample size calculations. We believe that using a number near the bottom of the range would represent an unnecessary risk to our ability to draw conclusions from the model in a timely manner. While we would prefer to have 75 MSAs in the model in order to increase the likelihood of being able to make definitive statements about the impact of the model at an earlier date, we believe the loss of the 8 MSAs now deemed not eligible for selection constitutes an acceptable risk.

With respect to the request to test the model in a limited pool of MSAs prior to testing it in the full set of selected MSAs, we believe that the testing of this model broadly is crucial to achieving the model's desired objectives and does not believe that proceeding in a few test MSAs prior to testing it in a broader set of MSAs would yield the same degree of information in the same time period.

Comment: A commenter was concerned that the selection strata used did not use MSA-level demographic measures in its selection process, including distributions of population by age, gender, or race; percent of population dually-eligible; percent of population with specific health conditions or other demographic composition measures. They believed these areas associated with more at-risk populations should be represented less in the selection. Another commenter did not question the selection strata but contended that the random selection happened to choose fewer areas with lower income and minority Medicare beneficiaries than they thought desirable. They specifically inquired Start Printed Page 73299after the lack of inclusion of MSAs in Alabama and Georgia.

Response: We considered but ultimately decided against including the dimension based on the demographic characteristics of an MSA incorporated in the selection strata. If we were to have done so, the purpose would have been to ensure an adequate representation along the range of these demographic considerations rather than to eliminate them from possible selection. While these factors are not explicitly part of the selection strata used, the resulting selected MSAs provide an adequate representation of a variety of circumstances including the experiences of areas with a higher degree of non-white populations, MSAs with a range in average income level, and other key characteristics. With regards to the specific concerns regarding under-representation in the MSAs selected from specific states, we note that Alabama, which has relatively high episode costs, had three of its seven eligible MSAs selected while Georgia, whose MSAs had episode payments that indicated relatively more efficient patterns of care, had two of its six eligible MSAs selected. As such, we believe that the experiences of these states and MSAs that are similar in nature to them are adequately represented in the selected MSAs.

Comment: A commenter requested clarification regarding how to interpret which MSAs are included in the model.

Response: We refer readers to Table 4 for a final list of the MSAs that are in the CJR model.

Final Decision: After consideration of the public comments we received, we are finalizing the proposal, with modification to include 67 of the originally selected 75 MSAs. We used updated BPCI participation level information in the application of the MSA exclusion rules for this final rule, resulting in the exclusion of an additional 8 MSAs that were previously selected. We note that we are posting the list of the participant hospitals in the selected MSAs on the CJR Web site at http://innovation.cms.gov/​initiatives/​CJR/​. This list will be updated throughout the model, to account for circumstances such as hospital mergers, BPCI termination, and new hospitals within the selected MSAs.

We set forth this final policy in § 510.100 and § 510.105.

Table 4—MSAs Included in the CJR Model

MSAMSA Name
10420Akron, OH
10740Albuquerque, NM
11700Asheville, NC
12020Athens-Clarke County, GA
12420Austin-Round Rock, TX
13140Beaumont-Port Arthur, TX
13900Bismarck, ND
14500Boulder, CO
15380Buffalo-Cheektowaga-Niagara Falls, NY
16020Cape Girardeau, MO-IL
16180Carson City, NV
16740Charlotte-Concord-Gastonia, NC-SC
17140Cincinnati, OH-KY-IN
17860Columbia, MO
18580Corpus Christi, TX
19500Decatur, IL
19740Denver-Aurora-Lakewood, CO
20020Dothan, AL
20500Durham-Chapel Hill, NC
22420Flint, MI
22500Florence, SC
23540Gainesville, FL
23580Gainesville, GA
24780Greenville, NC
25420Harrisburg-Carlisle, PA
26300Hot Springs, AR
26900Indianapolis-Carmel-Anderson, IN
28140Kansas City, MO-KS
28660Killeen-Temple, TX
30700Lincoln, NE
31080Los Angeles-Long Beach-Anaheim, CA
31180Lubbock, TX
31540Madison, WI
32820Memphis, TN-MS-AR
33100Miami-Fort Lauderdale-West Palm Beach, FL
33340Milwaukee-Waukesha-West Allis, WI
33700Modesto, CA
33740Monroe, LA
33860Montgomery, AL
34940Naples-Immokalee-Marco Island, FL
34980Nashville-Davidson-Murfreesboro-Franklin, TN
35300New Haven-Milford, CT
35380New Orleans-Metairie, LA
35620New York-Newark-Jersey City, NY-NJ-PA
35980Norwich-New London, CT
36260Ogden-Clearfield, UT
36420Oklahoma City, OK
36740Orlando-Kissimmee-Sanford, FL
37860Pensacola-Ferry Pass-Brent, FL
38300Pittsburgh, PA
38940Port St. Lucie, FL
38900Portland-Vancouver-Hillsboro, OR-WA
39340Provo-Orem, UT
39740Reading, PA
40980Saginaw, MI
41860San Francisco-Oakland-Hayward, CA
42660Seattle-Tacoma-Bellevue, WA
42680Sebastian-Vero Beach, FL
43780South Bend-Mishawaka, IN-MI
41180St. Louis, MO-IL
44420Staunton-Waynesboro, VA
45300Tampa-St. Petersburg-Clearwater, FL
45780Toledo, OH
45820Topeka, KS
46220Tuscaloosa, AL
46340Tyler, TX
48620Wichita, KS

B. Episode Definition for the CJR Model

1. Background

CJR model is an episode payment model, focused on incentivizing health care providers to improve the efficiency and quality of care for an episode of care as experienced by a Medicare beneficiary by bundling payment for services furnished to the beneficiary for an episode of care for a specific clinical condition over a defined period of time. Key policies of such a model include the definition of episodes of care. Episodes of care have two significant dimensions—(1) A clinical dimension that describes what clinical conditions and associated services comprise the episode; and (2) a time dimension that describes the beginning, middle, and end of an episode. We present our proposals, summarize public comments and provide our responses, and finalize the policies for these two dimensions of CJR episodes in this section.

2. Clinical Dimension of Episodes of Care

a. Definition of the Clinical Conditions Included in the Episode

As discussed previously in section I.A. of this final rule, we identified LEJR episodes, primarily hip and knee replacements, as the focus of this model. In the proposed rule, we stated our belief that a straightforward approach for hospitals and other providers to identify Medicare beneficiaries in this payment model is important for the care redesign that is required for model success, as well as to operationalize the proposed payment and other model policies.

The vast majority of LEJRs are furnished in the inpatient hospital setting, with a small fraction of partial knee replacements occurring in the hospital outpatient department (HOPD) setting. Most of the Current Procedural Terminology (CPT) codes that physicians report for LEJR are on the hospital OPPS inpatient only list. The CY 2015 OPPS inpatient only list is Addendum E of the CY 2015 Hospital Outpatient Prospective Payment—Final Rule with Comment Period, which is available on the CMS Web site at: http://www.cms.gov/​Medicare/​Medicare-Fee-for-Service-Payment/​ASCPayment/​ASC-Regulations-and-Notices-Items/​CMS-Start Printed Page 733001613-FC.html. Thus, under current FFS payment policy, Medicare pays hospitals for the facility services required for most LEJR procedures only when those procedures are furnished in the inpatient hospital setting. Therefore, in our proposal we stated our belief that an episode payment model most appropriately focuses around an inpatient hospitalization for these major surgical procedures, as there is little opportunity for shifting the procedures under this model to the outpatient setting.

We noted further that LEJRs are paid for under the IPPS through the following two Medicare Severity-Diagnosis Related Groups (MS-DRGs):

  • MS-DRG 469 (Major joint replacement or reattachment of lower extremity with Major Complications or Comorbidities (MCC)).
  • MS-DRG 470 (Major joint replacement or reattachment of lower extremity without MCC).

Multiple International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) procedure codes that describe LEJR procedures and other less common lower extremity procedures group to these MS-DRGs, with their percentage distribution within the IPPS MS-DRGs 469 and 470 for the past 4 years outlined in Table 5.

Table 5—Distribution of Hospital Claims for ICD-9-CM Procedure Codes Mapping to MS-DRGs 469 and 470

ICD-9-CM procedure codeCode descriptorFY 2014 (%)FY 2013 (%)FY 2012 (%)FY 2011 (%)
81.54Total knee replacement57585858
81.51Total hip replacement30292928
81.52Partial hip replacement12131314
81.56Total ankle replacement0000
00.85Resurfacing hip, total, acetabulum and femoral head0000
00.86Resurfacing hip, partial, femoral head0000
00.87Resurfacing hip, partial, acetabulum0000
84.27Lower leg or ankle reattachment0N/AN/AN/A
84.28Thigh reattachmentN/AN/AN/A0
Note: Percentages or claim counts with “N/A” had no claims. Percentages of 0% represent less than 0.5% of total claims.

Additionally, we noted that there are various types of claims-based information available to CMS, hospitals, and other providers, that could be used to identify beneficiaries in the model who receive LEJRs, including the MS-DRGs for the acute care hospitalization for the procedure, the ICD-9-CM procedure code on the hospital claim, or the CPT code(s) reported by the orthopedic surgeon who furnishes the surgical procedure. While we could utilize ICD-9-CM procedure codes or CPT codes to identify beneficiaries included in the model, over 85 percent of procedures that group to MS-DRGs 469 and 470 are hip or knee replacements. Additionally, the hospitals that would be participating in this model receive payment under the IPPS, which is not determined by CPT codes and is based on clinical conditions and procedures that group to MS-DRGs. Finally, our review of the other low volume procedures that group to these same MS-DRGs, aside from total or partial hip and knee replacements, did not suggest that there is significant clinical or financial heterogeneity within these two MS-DRGs such that we would need to define care for included beneficiaries by ICD-9-CM procedure codes.

Therefore, we proposed that an episode of care in the CJR model would be triggered by an admission to an acute care hospital stay (hereinafter “the anchor hospitalization”) paid under MS-DRG 469 or 470 under the IPPS during the model performance period. This approach offers operational simplicity, for providers and CMS, and is consistent with the approach taken by the BPCI initiative to identify beneficiaries whose care is included in the LEJR episode for that model. We sought public comments on this proposal to define the clinical conditions that are the target of CJR.

The following is a summary of the comments received and our responses.

Comment: Some commenters expressed support for CMS' proposal to define the clinical conditions included in the CJR model episode by discharge from an anchor hospitalization that is paid under MS-DRG 469 or 470 under the IPPS, although a commenter claimed that the cases within each MS-DRG are too heterogeneous to form the basis of a single target price as CMS proposed. The commenter added that risk adjustment could take the form of case exclusions, stratifying cases within each MS-DRG to create separate target prices, or adjusting the target prices based on principal procedure and patient characteristics. Most commenters recommended that CMS limit the model to a subset of beneficiaries that were discharged from these two MS-DRGs, effectively excluding certain cases as form of risk adjustment to reduce the heterogeneity of the cases in the model. The commenters asserted that CMS' proposal, which did not include risk adjustment beyond setting different target prices for episodes based on discharges from the two different MS-DRGs, failed to take into consideration the variability of service needs of beneficiaries discharged from these two MS-DRGs related to the specific procedure performed, the elective or urgent/emergent nature of the procedure, and the beneficiary's clinical and demographic characteristics, including underlying medical conditions and age. Several commenters recommended that CMS define the clinical conditions included in the model by discharges only from MS-DRG 470, claiming that these beneficiaries represented a more homogeneous group that had less complex health care needs. Some commenters urged CMS to define the clinical conditions in the model based on specific MS-DRG and ICD-9-CM procedure code combinations for hip and knee arthroplasty, and stated that CMS should exclude low volume procedures that also map to MS-DRGs 469 and 470 including ankle replacement; lower leg, ankle, and thigh reattachment; and hip resurfacing procedures. The commenters stated that these uncommon procedures display substantial heterogeneity in the clinical characteristics and needs of the beneficiary, as well as the associated Medicare payment for services throughout an episode. They contended that the rationale for CMS' proposal addressed hip and knee replacement in Start Printed Page 73301detail but failed to consider the different PAC patterns of other beneficiaries discharged from the same MS-DRGs but who had different surgical procedures. A commenter recommended that CMS specifically exclude episodes for conversion total joint arthroplasty procedures, which require removal of previously placed hardware followed by THA or TKA in the same operative session, arguing that these beneficiaries had more complex needs.

Many commenters recommended that CMS define the clinical conditions in the model as episodes specific to elective total hip arthroplasty (THA) and total knee arthroplasty (TKA) procedures. The commenters stated that this group of beneficiaries is more homogeneous than beneficiaries undergoing emergent joint replacement procedures for hip fractures or undergoing the other low volume procedures that map to the MS-DRGs. Given that CMS did not propose risk adjustment under the model based on procedure or patient characteristics, the commenters contended that limiting the model to these clinical conditions, that represent about 85 percent of beneficiaries discharged for the two MS-DRGs, would provide a sufficient number of cases to test LEJR episode payment and allow hospitals to create efficient, effective clinical pathways for these beneficiaries. The commenters also observed that CMS' quality measures, specifically the THA/TKA readmissions and complications measures, as well as the voluntary data collection for patient-reported outcomes, would represent only the quality of care for beneficiaries undergoing elective THA and TKA procedures. Several commenters recommended that CMS only include episodes in the model for beneficiaries discharged from MS-DRG 469 or 470 whose data would be used to determine the model's quality measures for the participant hospital.

The commenters suggested several different approaches to defining the clinical conditions included in the model as elective THA or TKA. One approach would be to eliminate from the model beneficiaries with reported ICD-9-CM procedure codes other than THA or TKA, and then further exclude some remaining beneficiaries with ICD-9-CM codes for hip fracture on their claim for the anchor hospitalization. Other commenters asserted that CMS should exclude the beneficiaries receiving the low volume procedures as well as those receiving partial hip arthroplasty (PHA) procedures. The commenters pointed out that almost all of the beneficiaries receiving PHA would have hip fractures and observed that the average Medicare episode payment for beneficiaries undergoing PHA was similar to beneficiaries discharged from MS-DRG 469 or 470 with hip fracture diagnoses, almost twice the payment for beneficiaries undergoing elective THA and TKA. Several commenters presented analyses that demonstrated that beneficiaries with hip fracture, regardless of their discharge from MS-DRG 469 or 470, when compared to beneficiaries with elective procedures, experience twice as high readmissions and PAC utilization rates, as well as higher morbidity and mortality.

The commenters in favor of excluding clinical conditions involving hip fractures from the model stated that the number of hip fracture cases treated by individual hospitals can vary significantly on an annual basis, both due to random variation and practice or population changes. Moreover, different hospitals provide care for different percentages of beneficiaries with hip fracture and, according to some commenters, academic medical centers and small hospitals care for disproportionate percentages of these cases for reasons of medical complexity and the urgent nature of the procedure, respectively, because beneficiaries who fall and experience a hip fracture are commonly transported to their local hospital for emergent treatment. Furthermore, in addition to the variation a hospital itself may experience regarding the percentage of hip fracture cases, which could lead to the hospital-specific historical data used for a portion of the target price to not be reflective of the health care needs of the hospital's episode population in a given performance year, some commenters observed that the increasing percentage of the target price contributed by regional data exacerbated their concerns. Hospitals in a region that care for a disproportionately high percentage of hip fracture patients compared to the regional average would be disadvantaged due to the more intense service needs of hip fracture patients, whereas hospitals caring for a disproportionately low percentage of hip fracture patients compared to the regional average would be advantaged. The commenters contended that excluding clinical conditions involving hip fractures from the CJR model would ensure homogeneity in the beneficiaries in the model such that hospitals would be treated fairly with respect to episode pricing based on the hospital-specific and regional historical CJR episode data for only those beneficiaries undergoing elective THA and TKA.

Response: We appreciate the analyses and suggestions provided by the commenters regarding the most appropriate approach to defining the clinical conditions included in the CJR model. As discussed in section III.C.4.b. of this final rule, we have decided to risk stratify the target price for each MS-DRG-anchored episode based on a beneficiary's hip fracture status. This policy allows us to maintain beneficiaries who receive LEJR procedures due to hip fractures in the CJR model, while acknowledging their typically greater health care needs by providing a target price that is based on payment for services furnished in the historical CJR episode data for Medicare beneficiaries with hip fractures in order to account for a significant amount of beneficiary-driven episode expenditure variation. While beneficiaries with hip fractures may present a more costly population due to greater health care needs, and CJR participant hospitals may vary in their percentages of such beneficiaries, we believe that beneficiaries with hip fracture have the potential to benefit substantially from the care pathways and improved care coordination among providers and suppliers that is incentivized by an episode payment model. In addition, we believe there are opportunities for increased efficiency in the care of beneficiaries with hip fracture who receive LEJR procedures with respect to appropriate PAC utilization and care coordination and management of chronic conditions that may be affected by the LEJR procedure or post-surgical care. Thus, we are finalizing our proposal to include LEJR procedures that result from hip fracture treatment in the clinical conditions that are part of the CJR model episodes, rather than limiting the model conditions to only elective THA and TKA.

We are also finalizing our proposal to include clinical conditions represented by discharge from both MS-DRG 469 and 470 in the CJR model. We believe that providing separate prices for episodes anchored by the two different MS-DRGs accounts for the differences in typical health care needs of the two groups of beneficiaries, specifically the higher IPPS payment for the anchor hospitalization for beneficiaries discharged under MS-DRG 469, as well as the pattern of service utilization for this group of beneficiaries in the 90 days following discharge.

Additionally, we are finalizing our proposal to include any lower extremity joint procedure that results in discharge from MS-DRG 469 or 470 in the CJR Start Printed Page 73302model, including ankle replacement; lower leg, ankle, and thigh reattachment; and hip resurfacing procedures. While the model beneficiaries with these less common clinical conditions are likely to be a small number at any specific participant hospital, they too may benefit from care redesign resulting in improved care coordination and quality that are goals of the CJR model. These beneficiaries share the experience of undergoing major surgical procedures involving the lower extremity with the majority of CJR model beneficiaries undergoing THA or TKA, and they too are likely to require PAC services and care coordination and management of chronic medical conditions to optimize their return to function. We expect that the Medicare actual episode payments for these clinical conditions may be highly variable given the small numbers and variable clinical characteristics of these beneficiaries such that historical episode data may have little predictive power regarding the actual episode payment for the beneficiaries in a model performance year. We do not believe this small number of beneficiaries will put participant hospitals at undue financial risk and further note that our payment policies as discussed in section III.C.3.c. and III.C.8. of this final rule provide a pricing adjustment for high payment episodes and limit hospital financial responsibilities to provide participant hospitals with additional protections.

We note that our final policy to include all clinical conditions that result in a discharge from MS-DRGs 469 or 470 in the CJR model allows us to continue to rely on MS-DRGs to define the clinical conditions included in the LEJR episode being widely tested under the CJR model, consistent with the BPCI methodology to define clinical conditions included in 48 different episodes based on the MS-DRGs for the anchor hospitalization. This approach provides greater certainty from the perspective of participant hospitals or CMS regarding the clinical conditions included in episodes, since the discharge MS-DRG is the defining parameter, and includes the greatest number of beneficiaries with similar clinical conditions in the CJR model test.

Comment: Several commenters urged CMS to include in the CJR model LEJR procedures where the procedure that would result in a beneficiary's discharge from MS-DRG 469 or 470 if furnished in the inpatient hospital setting is furnished in the HOPD, ambulatory surgical center (ASC), or other dedicated facility that is not an acute care facility. The commenters explained that elective procedures are commonly furnished in the HOPD, ASC, or other dedicated facilities that are not acute care facilities for certain beneficiaries covered by commercial insurance, while Medicare covers and pays for the procedures only when they are furnished in the inpatient hospital settings. The commenters disputed CMS' assertion in the proposed rule that there is little opportunity for shifting these procedures to the outpatient setting. They urged CMS to permit these LEJR procedures to be furnished to Medicare beneficiaries in other settings under the CJR model to improve episode efficiency. The commenters contended that physicians should be able to select the most appropriate inpatient hospital or outpatient setting based on the beneficiary's clinical condition.

Response: We appreciate the interest of the commenters in providing LEJR procedures under the CJR model to Medicare beneficiaries in alternative outpatient settings as a further opportunity to test strategies to provide high quality, efficient episode care for beneficiaries undergoing LEJR procedures. As we discussed in the proposed rule, the vast majority of LEJR procedures are furnished to Medicare beneficiaries in the inpatient hospital setting, with a small fraction of partial knee replacements occurring in the hospital outpatient department (HOPD). Most of the CPT codes that physicians report for LEJR procedures are on the hospital OPPS inpatient only list. Thus, under current Medicare program policy, Medicare generally pays hospitals for the facility services required for LEJR only when those procedures are furnished in the inpatient hospital setting. When we stated our belief in the proposed rule that an episode payment model such as the CJR model most appropriately focuses around an inpatient hospitalization for these major surgical procedures, as there is little opportunity for shifting the procedures under the model to the outpatient setting, we meant that this would be true under current Medicare policy. Because Medicare generally does not pay hospitals if procedures that would be assigned to MS-DRG 469 or 470 when furnished to inpatients are performed on hospital outpatients, these procedures would not be able to be shifted under the CJR model to the outpatient setting.

Because most LEJR procedures are on the OPPS inpatient list and CMS has, therefore, determined that Medicare beneficiaries require an inpatient hospitalization for payment of these procedures to hospitals, we are not changing the current inpatient only list designation of these LEJR procedures for the CJR model. CJR is an episode payment model, not a model designed to test different sites of services for procedures that CMS has thus far determined may not be safely performed on Medicare beneficiaries in the outpatient setting. Therefore, we are finalizing our proposal that the CJR model will continue to focus around an inpatient hospitalization for these major surgical procedures that result in a discharge from MS-DRG 469 or 470, and a procedure furnished in the outpatient setting will not be included in the model.

Comment: Several commenters maintained that because the procedures that result in discharge from MS-DRG 469 and 470 that define the clinical conditions included in the CJR model are on the OPPS inpatient only list, CMS should commit to keeping these procedures on the inpatient only list for the 5-year performance period of the model. The commenters pointed out that CMS has previously proposed, but not finalized, the removal of TKA procedures from the inpatient only list. The commenters stated that if any additional procedures that would otherwise result in discharge from one of the two MS-DRGs in the CJR model were to be removed from the inpatient only list during a year when the CJR model is being tested, the beneficiaries who would be included in the model performance year due to a procedure in the inpatient hospital setting would be sicker and more complex than those included in the historical CJR episodes used to set target prices. Therefore, the commenters reasoned that in order to establish target prices that reflect the health care needs and medical complexity of the CJR model beneficiaries in a model performance year, CMS should not remove any LEJR procedures from the OPPS inpatient only list until after the CJR model ends.

Response: We share the commenters' interest in ensuring that the historical CJR episodes that are used to set the target prices for CJR model episodes during a performance year reflect the health care needs and medical complexity of beneficiaries who are comparable to those actually included in the CJR model. If we were to remove an LEJR procedure from the OPPS inpatient only list at any point during the 5-year model test, we agree with the commenters that we would need to consider the effects of such a change on the CJR model pricing methodology, taking into consideration the characteristics of the beneficiaries expected to be in the model due to a Start Printed Page 73303procedure furnished in the inpatient hospital setting after the change to the inpatient only list. If we concluded that changes in our pricing methodology were necessary because the beneficiaries in the historical CJR episodes used to set target prices would no longer be similar to those in the model performance year, we would propose such changes through notice and comment rulemaking.

Comment: Several commenters claimed that different states were testing different LEJR episode payment models. A commenter provided the example of Tennessee mandatory Medicaid bundles that utilize a different episode definition than proposed for the CJR model. The commenters encouraged CMS to move toward standard episode definitions for mandatory models, noting that each of the inconsistent mandatory models is being tested under the Innovation Center's statutory authority. The commenters contended that different episode payment models lead to excessive burden and greater cost for health care providers.

Response: We appreciate the perspective of the commenters on the challenges related to testing mandatory bundled payments with different episode definitions in the same community. We note, however, that the CJR model and various state episode payment models are all in various stages of testing and have used different strategies to arrive at the episode definitions for each model. By definition, models being tested have not yet produced evidence of improved quality and/or cost savings, so we lack the necessary evaluation results from various approaches to consider standardizing episode definitions. We believe there is value in testing different episode definitions given the current state of knowledge about bundled payment. We also believe that, regardless of the specific definitions for episodes that address the same clinical conditions in various different payment models, episode payment models share a common focus on improving the quality of care and increasing the efficiency of care through a variety of well-established strategies, such as increased communication among health care providers along the continuum of acute and PAC and improved care coordination and care management to promote beneficiary engagement that leads to adherence to treatment plans and, correspondingly, reductions in hospital readmissions and complications. As we gain more experience with episode payment models and examine their results, we will consider the potential benefits of standardizing episode definitions to the extent possible.

Summary of Final Decisions: After consideration of the public comments we received, we are finalizing our proposal to define the clinical conditions included in the CJR model by admission to an IPPS hospital that results in a discharge from MS-DRG 469 or 470.

The final policies for defining the clinical conditions are set forth in § 510.100 and § 510.200.

b. Definition of Related Services Included in the Episode

For purposes of this model, as in BPCI, given the frequent comorbidities experienced by Medicare beneficiaries and the generally elective nature of LEJR, we are interested in testing inclusive episodes to incentivize comprehensive, coordinated patient-centered care for the beneficiary throughout the episode. We proposed to exclude only those Medicare items and services furnished during the episode that are unrelated to LEJR procedures based on clinical justification. During our experience with BPCI implementation, we reviewed a number of narrow episode definitions for LEJR episodes that were recommended by BPCI participants and other interested parties during the design phase for this project. We concluded that these narrow definitions commonly exclude many services that may be linked to the LEJR, as LEJR beneficiaries, on average, are at higher risk for more clinical problems than Medicare beneficiaries who have not recently undergone such procedures.

Therefore, we proposed that all CJR episodes, beginning with the admission for the anchor hospitalization under MS-DRG 469 or 470 through the end of the proposed episode, include all items and services paid under Medicare Part A or Part B with the exception of certain exclusions that would be excluded because they are unrelated to the episode. The items and services ultimately included in the episode after the exclusions are applied are called related items and services. As discussed in sections III.C.4. and III.C.6. of this final rule, Medicare spending for related items and services would be included in the historical data used to set target prices, as well as in the calculation of actual episode spending that would be compared against the target price to assess the performance of participant hospitals. In contrast, Medicare spending for unrelated items and services (excluded from the episode definition) would not be included in the historical data used to set target prices or in the calculation of actual episode spending.

We proposed that related items and services included in CJR episodes would be the following items and services paid under Medicare Part A or Part B, after the exclusions are applied:

  • Physicians' services.
  • Inpatient hospital services (including readmissions), with certain exceptions discussed later in this section.
  • Inpatient psychiatric facility (IPF) services.
  • Long Term Care Hospital (LTCH) services.
  • IRF services.
  • SNF services.
  • HHA services.
  • Hospital outpatient services.
  • Independent outpatient therapy services.
  • Clinical laboratory services.
  • Durable medical equipment (DME).
  • Part B drugs.
  • Hospice.

We noted that under our proposed definition of related services included in the episode, the episode could include certain per-member-per-month model payments, as discussed in section III.C.7.d. of this final rule.

We proposed to exclude from CJR drugs that are paid outside of the MS-DRG, specifically hemophilia clotting factors (§ 412.115), identified through HCPCS code, diagnosis code, and revenue center on IPPS claims. Hemophilia clotting factors, in contrast to other drugs that are administered during an inpatient hospitalization and paid through the MS-DRG, are paid separately by Medicare in recognition that clotting factors are costly and essential to appropriate care for certain beneficiaries. Thus, in the proposed rule we stated our belief that there are no efficiencies to be gained in the variable use of these high cost drugs when particular beneficiaries receive LEJR procedures who have significantly different medical needs for clotting factors under an episode payment model, so we proposed to exclude these high cost drugs from the actual historical episode expenditure data used to set target prices and from the hospital's actual episode spending that is reconciled to the target price. Similarly, we proposed to exclude IPPS new technology add-on payments for drugs, technologies, and services from CJR episodes, excluding them from both the actual historical episode expenditure data used to set target prices and from the hospital's actual episode spending that is reconciled to the target price. This proposal would apply to both the anchor hospitalization Start Printed Page 73304and any related readmissions during the episode. New technology add-on payments are made separately and in addition to the MS-DRG payment under the IPPS for specific new drugs, technologies, and services that substantially improve the diagnosis or treatment of Medicare beneficiaries and would be inadequately paid otherwise under the MS-DRG system. Medicare pays a marginal cost factor of 50 percent for the costs to hospitals of the new drugs, technologies, or services. We did not believe it would be appropriate for the CJR model to potentially hamper beneficiaries' access to new technologies that are receiving new technology add-on payments or to burden hospitals who choose to use these new drugs, technologies, or services with concern about these payments counting toward actual episode expenditures. In addition, because new drugs, technologies, or services approved for the add-on payments vary unpredictably over time in their application to specific clinical conditions, in the proposed rule we stated our belief that we should exclude IPPS new technology add-on payments from CJR episodes.

We followed a number of general principles in determining other proposed excluded services from the CJR episodes in order to promote coordinated, high-quality, patient-centered care. Based on the broad nature of these episodes, we proposed to identify excluded (unrelated) services rather than included (related) services based on the rationale that all Part A and Part B services furnished during the episode are related to the episode, unless they are unrelated based on clinical justification as described in more detail later in this section. In developing our proposals for exclusions for this model, we stated our belief that no Part A services, other than certain excluded hospital readmissions during the episode as described in this section, furnished post-hospital discharge during the episode should be excluded, as post-hospital discharge Part A services are typically intended to be comprehensive in nature. We also stated our belief that no claims for services with diagnosis codes that are directly related to the LEJR procedure itself (for example, loosening of the joint prosthesis) based on clinical judgment, and taking into consideration coding guidelines, should be excluded. Furthermore, we stated our belief that no claims for diagnoses that are related to the quality and safety of care furnished during the episode, especially the anchor hospitalization under MS-DRG 469 or 470, should be excluded, such as direct complications of post-surgical care during the anchor hospitalization. Examples of diagnoses that would not be excluded on this basis include surgical site infection and venous thromboembolism. Finally, in the proposed rule we stated our belief that no claims for services for diagnoses that are related to preexisting chronic conditions such as diabetes, which may be affected by care furnished during the episode, should be excluded. However, severe exacerbations of chronic conditions (for example, some surgical readmissions) that are unlikely to be affected by care furnished during the episode should be excluded; thus, when a beneficiary is admitted to the hospital during the episode for these circumstances, we would not consider it to be a related readmission for purposes of CJR. We also stated our belief that services for clinical conditions that represent acute clinical conditions not arising from an existing chronic clinical condition or complication of LEJR surgery occurring during an episode of care, which would not be covered by the previous principles about included services, should be excluded.

To operationalize these principles for CJR, we proposed to exclude unrelated inpatient hospital admissions during the episode by identifying MS-DRGs for exclusion. We proposed to exclude unrelated Part B services based on the ICD-9-CM diagnosis code (or their International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) equivalents when ICD-10-CM codes are implemented) that is the principal diagnosis code reported on claims for services furnished during the episode. More specifically, we proposed to exclude specific inpatient hospital admissions and services consistent with the LEJR episode definition (also triggered by MS-DRGs 469 and 470) that is currently used in BPCI Model 2. We note that the list of exclusions was initially developed over 2 years ago for BPCI through a collaborative effort of CMS staff, including physicians from medical and surgical specialties, coding experts, claims processing experts, and health services researchers. The list has been shared with thousands of entities and individuals participating in one or more phases of BPCI, and has undergone refinement over that time in response to stakeholder input about specific diagnoses or MS-DRGs for exclusion, resulting in only minimal changes over the last 2 years. Thus, the BPCI list of exclusions for LEJR procedures has been vetted broadly in the health care community; refined based on input from a wide variety of providers, researchers and other stakeholders; and successfully operationalized in the BPCI models. We proposed its use in CJR based on our confidence related to our several of years of experience that this definition is reasonable and workable for LEJR episodes, for both providers and CMS.

With respect to the proposed inpatient hospital admission exclusions for this model, we proposed that all medical MS-DRGs for readmissions be included in CJR episodes as related services, with the exception of oncology and trauma medical MS-DRGs. We proposed that admissions for oncology and trauma medical MS-DRGs be excluded from CJR episodes. Readmissions for medical MS-DRGs are generally linked to the hospitalization for the LEJR procedure as a complication of the illness that led to the surgery, a complication of treatment or interactions with the health care system, or a chronic illness that may have been affected by the course of care. We refer readers to section III.D. of this final rule for background and discussion of the complication rate measure proposed for CJR that includes common medical complications resulting from the previously stated circumstances following LEJR procedures and that may result in related hospital readmissions. For readmissions for medical MS-DRGs, the selection of the primary diagnosis code is not clear-cut, so in the proposed rule we stated our belief that all should be included because providers should focus on comprehensive care for beneficiaries during episodes. We proposed to include all disease-related surgical MS-DRGs for readmissions, such as hip/knee revision, in CJR episodes. We also proposed to include readmissions for all body system-related surgical MS-DRGs as they are generally related to complications of the LEJR procedures. An example of a readmission of this type would be for an inferior vena cava filter placement for treatment of thromboembolic complications of the LEJR. We proposed to exclude hospital admissions for chronic disease surgical MS-DRGs, such as prostatectomy (removal of the prostate gland), as they are unrelated to the clinical condition that led to the LEJR and they would not have been precipitated by the LEJR. Finally, we proposed that hospital admissions for acute disease surgical MS-DRGs, such as appendectomy, be excluded because they are highly unlikely to be related to, or precipitated by, LEJR procedures and would not be affected by LEJR episode care redesign.Start Printed Page 73305

With respect to the LEJR proposed diagnosis code exclusions for Part B services for this model, we proposed that ICD-9-CM codes be excluded or included as a category and as identified by code ranges. We proposed that disease-related diagnoses, such as osteoarthritis of the hip or knee, are included. We also proposed that body system-related diagnoses are included because they relate to complications that may arise from interactions with the health care system. An example of this would be pressure pre-ulcer skin changes. Additionally, we proposed that all common symptom diagnoses are included because providers have significant discretion to select these as principal diagnosis codes. We proposed that acute disease diagnoses, such as severe head injury, are excluded. Finally, we proposed that chronic disease diagnoses be included or excluded based on specific clinical and coding judgment as described previously with respect to the original development of the exclusions for LEJR episodes under BPCI, taking into consideration whether the condition was likely to have been affected by the LEJR procedure and recovery period and whether substantial services were likely to have been provided for the chronic condition during the episode. Thus, chronic kidney disease and cirrhosis would be included in the episode, but glaucoma and chemotherapy would be excluded.

Proposed exclusions from CJR episodes were based on care for unrelated clinical conditions represented by MS-DRGs for readmissions during the episode and ICD-9 CM codes for Part B services furnished during the episode after discharge from the anchor hospitalization. The complete lists of proposed excluded MS-DRGs for readmissions and proposed excluded ICD-9-CM codes for Part B services are posted on the CMS Web site at http://innovation.cms.gov/​initiatives/​cjr/​.

In the proposed rule, we noted that as CMS moves to implement ICD-10-CM we would make the CJR exclusions that would map to the final ICD-9-CM exclusions for CJR available in the ICD-10-CM format as well. We proposed that all Part A and B-covered items and services that would not be excluded based on the exclusions list are included in the episode. Furthermore, we proposed to update the exclusions list without rulemaking on an annual basis, at a minimum, to reflect annual changes to ICD-CM coding and annual changes to the MS-DRGs under the IPPS, as well as to address any other issues that are brought to our attention by the public throughout the course of the model test.

We would first develop potential exclusions list revisions of MS-DRGs for readmissions and ICD-9-CM (or ICD-10-CM, as applicable) diagnosis codes for Part B services based on our assessment against the following standards:

  • We would not exclude any items or services that are—

++ Directly related to the LEJR procedure itself (such as loosening of the joint prosthesis) or the quality or safety of LEJR care (such as post-surgical wound infection or venous thromboembolism); and

++ For chronic conditions that may be affected by the LEJR procedure or post-surgical care (such as diabetes). By this we mean that where a beneficiary's underlying chronic condition would be affected by the LEJR procedure, or where the beneficiary's LEJR or post-LEJR care must be managed differently as a result of the chronic condition, then those items and services would be related and would be included in the episode.

  • We would exclude items and services for—

++ Chronic conditions that are generally not affected by the LEJR procedure or post-surgical care (such as removal of the prostate). By this we mean that where a beneficiary's underlying chronic condition would not be affected by the LEJR procedure, or where the beneficiary's LEJR or post-LEJR care need not be managed differently as a result of the chronic condition, then those items and services would not be related and would not be included in the episode; and

++ Acute clinical conditions not arising from existing episode-related chronic clinical conditions or complications of LEJR surgery from the episode (such as appendectomy).

We proposed to post the potential revised exclusions, which could include additions to or deletions from the exclusions list, to the CMS Web site to allow for public input on our planned application of these standards, and then adopt changes to the exclusions list with posting to the CMS Web site of the final revised exclusions list after our consideration of the public input.

We sought comment on our proposals for identifying excluded readmissions and Part B-covered items and services, as well as our proposed process for updating the exclusions list.

The following is a summary of the comments received and our responses.

Comment: Several commenters recommended that CMS clarify the proposal that named “independent outpatient therapy services” in the episode definition list of related Part A and Part B services included in the episode. The commenters pointed out that while this list specified “independent outpatient therapy services,” which would appear to only represent services furnished by therapists in private practices included in CMS data under certain supplier specialty codes, the commenters believe that CMS should refer to the services as outpatient therapy services in order to include all outpatient physical therapy, occupational therapy, and speech-language pathology therapy services in the definition of related Part A and Part B services included in the episode. The commenters noted that in the proposed rule discussion of CJR collaborators CMS referred to financial arrangements with outpatient therapy providers, a category of providers that was not defined in the proposed rule and has not otherwise been previously defined in the Medicare program. Therefore, the commenters recommended that CMS define outpatient therapy providers in regulation in the CJR final rule as a physician, supplier, or provider furnishing outpatient physician therapy services, outpatient occupational therapy services, or outpatient speech-language pathology services. The commenters suggested that CMS should then clarify that services furnished by these outpatient therapy providers (outpatient therapy services) would be included in the episode definition, thereby including these payments in the CJR historical episode data used to set target prices and in the calculation of actual episode spending that would be compared against the target price.

Response: We agree with the commenters' suggestion that we define outpatient therapy providers in regulation to ensure consistent and accurate reference to certain providers and services under the CJR model, and that we should include services furnished by outpatient therapy providers as related services in the CJR model after the exclusions are applied. Therefore, we are adding the following new definition to § 510.2: Provider of outpatient therapy services means a provider or supplier furnishing—(1) Outpatient physical therapy services as defined in 410.60 of this chapter, or (2) outpatient occupational therapy services as defined in 410.59 of this chapter, or (3) outpatient speech-language pathology services as defined in 410.62 of this chapter. We are also revising § 510.200(b)(10) to remove the word “independent” preceding outpatient therapy services.Start Printed Page 73306

Comment: Several commenters recommended that CMS add to the list of related services included in CJR model episodes drugs covered under Medicare Part D. The commenters asserted that Part D-covered drugs make important contributions to beneficiary health, especially for beneficiaries with chronic medical conditions and, therefore, should be included in a broadly defined episode payment model such as the CJR model to provide opportunities for improved quality and efficiency of care for beneficiaries.

Response: We appreciate the interest expressed by the commenters in including drugs covered under Part D in the LEJR episode definition used for the CJR model. However, while we agree with the commenters that the appropriate use of Part D-covered drugs can play an important role in improving a beneficiary's health, we will not be expanding our list of Part A and Part B items and services related to the episode to add Part D-covered drugs. We proposed to require all beneficiaries included in the CJR model to have both Part A and Part B coverage throughout the duration of the episode in order to ensure we had comprehensive episode payment data to calculate actual episode spending to be compared against the target price. However, enrollment in Part D is voluntary and a substantial percentage of Medicare beneficiaries do not have Part D coverage, so we would lack comprehensive payment information for all beneficiaries in the model in order to determine an episode target price and calculate actual episode spending. In addition, beneficiary-specific information about Part D drug spending that could be attributed to episodes would not be available in a timeframe consistent with the time periods for claims used to set target prices and the timeline for reconciliation where actual episode spending is aggregated and compared against the target price. Finally, given that the CJR model is testing LEJR episodes, we believe there is limited opportunity to shift spending from Part B to Part D to reduce actual episode spending, even though we have not included Part D payments in the episode definition. Most beneficiaries with chronic conditions would be taking similar drugs before and during the episode, and, other than pain medications, Part D-covered drugs are not commonly used to manage the direct post-surgical and PAC rehabilitation needs of most LEJR beneficiaries, who rarely experience significant complications from the surgery. Therefore, we are finalizing our proposal to not include all Part D-covered drugs from the list of related items and services included in CJR episodes.

Comment: Several commenters recommended that CMS exclude Inpatient Psychiatric Facility (IPF) services from CJR episodes because they would be unlikely to be related to the LEJR procedure. The commenters suggested that the services are always medically necessary with no opportunities for efficiency and would be more likely to be associated with injury that led to the need for LEJR procedure, rather than related to the surgical procedure or recovery. Several commenters stated that CMS should exclude these services from the episode definition because they were excluded under LEJR episodes in BPCI. Another commenter suggested that CMS exclude IPF services furnished more than 14 days after surgery because after than point, the commenter believes these services would be unlikely to be related to the surgery or recovery.

Response: We are clarifying that under BPCI, IPF services furnished following discharge from the LEJR episode anchor hospitalization but during the episode are included in the LEJR episode definition, unless they fall into one of the excluded MS-DRGs. Thus, we include inpatient psychiatric services whether paid under the IPPS or the IPF PPS in LEJR episodes under BPCI according to the same policy that would exclude readmissions paid under either payment system based on the same exclusion list. We see no reason under the CJR model not to apply the standard we proposed to define related and unrelated Part A and Part B services with respect to CJR episodes. Therefore, we believe the list of excluded MS-DRGs identifies those IPF admissions during the episode that would be clinically unrelated to the LEJR episode so we exclude them from the episode definition, whereas IPF services any time during a CJR episode that result in discharge from an MS-DRG that is not excluded would be related and included in the CJR model episode definition. We disagree with the commenter that all IPF services furnished more than 14 days after surgery are unlikely to be related to the LEJR procedure or complications of the procedure or to a chronic condition that must be managed differently as a result of the procedure. Regardless of the time IPF services are furnished following discharge from the anchor hospitalization, we believe the MS-DRG exclusions identify those circumstances when IPF services are unrelated to the CJR model episode. Therefore, consistent with the BPCI policy, we are finalizing our proposal to include IPF services in the CJR model episode definition when they are assigned to an MS-DRG that is not excluded from episode definition.

Comment: Several commenters commended CMS for proposing to include hospice services in the episode definition for the CJR model, which provides recognition of hospice services as an essential element of the health care continuum. They stated that they looked forward to CMS sharing data resulting from the model that provides insight into the impact of incorporating hospice as part of a bundled care model and coordinated approach to post-hospital care. However, the commenters asserted that generally hospice services would be unrelated to the LEJR episode because they would most commonly address a serious and unanticipated complication of surgery or the hospitalization, discovery during or immediately after the surgery of a previously undetected terminal prognosis, or an unrelated accident following the procedure. While acknowledging that some hospice services would be related to the LEJR episode under uncommon circumstances, the commenters encouraged CMS to include in the final rule the process that would be used to identify included and excluded hospice services from CJR episodes. The commenters urged CMS to further describe its rationale for including hospice services in the episode definition, and supply data that relates to hospice services and the CJR model. Finally the commenters recommended that CMS establish a data acquisition system on hospice use in the final model. Some commenters expressed confusion about CMS' proposal to include hospice services in the episode definition and inquired about whether CMS intended to include all hospice services or to exclude certain hospice services as unrelated to the LEJR episode according to the beneficiary's diagnosis.

A number of other commenters recommended that CMS exclude all hospice services from the CJR episode definition, except for the post-episode spending calculation that analyzes all Part A and Part B spending for model beneficiaries, both for consistency with BPCI and to ensure no incentives for underutilization of the hospice benefit were created by the CJR model. The commenters asserted that all hospice services were unrelated to the LEJR episode, and encouraged CMS to exclude hospice services in order to ensure timely access to hospice for CJR model beneficiaries.Start Printed Page 73307

Response: We appreciate the interest of the commenters in ensuring continuing beneficiary access to hospice services under the CJR model. We note that while we exclude all hospice services under BPCI, our proposal for the CJR model would exclude no hospice services. Specifically, we proposed no exclusions of Part A services furnished during the 90 day period after discharge from the anchor hospitalization other than certain hospital readmissions identified by excluded MS-DRGs. We understand that CJR model beneficiaries could receive hospice services during an episode under several different types of clinical circumstances. For example, the beneficiary could be enrolled in hospice prior to the LEJR episode, experience a pathologic hip fracture, and require THA to stabilize the beneficiary's hip. Alternatively, the beneficiary could have an LEJR procedure and enter into hospice at some point during the episode in the 90 days following discharge from the anchor hospitalization, either after experiencing a surgical complication leading to a terminal prognosis or based on a new diagnosis of a terminal stage of an illness. We note that given the pre-surgical screening that patients must undergo before an LEJR procedure, it would be rare for a new diagnosis that would render the patient terminally ill to occur within 3 months after the LEJR procedure that was not already identified during the pre-surgical screening process.

Medicare hospice care is palliative care for individuals with a prognosis of living 6 months or less if the terminal illness runs its normal course. As referenced in § 418.22(b)(1), to be eligible for Medicare hospice services, the patient's attending physician (if any) and the hospice medical director must certify that the individual is “terminally ill,” as defined in section 1861(dd)(3)(A) of the Act and our regulations at § 418.3 that is, the individual's prognosis is for a life expectancy of 6 months or less if the terminal illness runs its normal course. When an individual is terminally ill, many health problems are brought on by underlying condition(s), as bodily systems are interdependent. Section 1861(dd)(1) of the Act establishes the services that are to be rendered by a Medicare certified hospice program and those services include: nursing care; physical therapy; occupational therapy; speech-language pathology therapy; medical social services; home health aide services (now called hospice aide services); physician services; homemaker services; medical supplies (including drugs and biologics); medical appliances; counseling services (including dietary counseling); short-term inpatient care (including both respite care and care necessary for pain control and acute or chronic symptom management) in a hospital, nursing facility, or hospice inpatient facility; continuous home care during periods of crisis and only as necessary to maintain the terminally ill individual at home; and any other item or service which is specified in the plan of care and for which payment may otherwise be made under Medicare, in accordance with Title XVIII of the Act. The services offered under the Medicare hospice benefit must be available, as needed, to beneficiaries 24 hours a day, 7 days a week (section 1861(dd)(2)(A)(i) of the Act).

The regulations at § 418.54(c) stipulate that the comprehensive hospice assessment must identify the patient's physical, psychosocial, emotional, and spiritual needs related to the terminal illness and related conditions, and address those needs in order to promote the hospice patient's well-being, comfort, and dignity. The comprehensive assessment must take into consideration the following factors: The nature and condition causing admission (including the presence or lack of objective data and subjective complaints); complications and risk factors that affect care planning; functional status; imminence of death; and severity of symptoms (§ 418.54(c)). Additionally, the hospice CoPs at § 418.56(c) require that the hospice must provide all reasonable and necessary services for the palliation and management of the terminal illness, related conditions and interventions to manage pain and symptoms. Therapy and interventions must be assessed and managed in terms of providing palliation and comfort without undue symptom burden for the hospice patient or family. In the December 16, 1983 Hospice final rule (48 FR 56010 through 56011), regarding what is related versus unrelated to the terminal illness, we stated: “. . . we believe that the unique physical condition of each terminally ill individual makes it necessary for these decisions to be made on a case-by-case basis. It is our general view that hospices are required to provide virtually all the care that is needed by terminally ill patients.”

Thus, hospice services furnished to CJR model beneficiaries should be included in the episode definition for the CJR model, regardless of the specific diagnosis of the beneficiary, because hospices are to provide virtually all care that is needed by terminally ill patients. If a CJR beneficiary was receiving hospice services during an episode, either because the beneficiary was enrolled in hospice prior to surgery and continued in hospice following surgery or the beneficiary enrolled in hospice following surgery that initiated the CJR model episode, we believe that hospice services would encompass care related to the LEJR episode and should, therefore, be included in the episode definition. As previously noted, given the comprehensive nature of the hospice benefit and the fact that body systems are interdependent at end of life, virtually all care needed by the terminally-ill individual would be related to the terminal prognosis and thus the responsibility of the hospice. As previously noted, hospices are required, per the Hospice CoPs at § 418.56(c), to provide all reasonable and necessary services for the palliation and management of the terminal illness, related conditions, and interventions to manage pain and symptoms. For patients that underwent LEJR procedures as part of the CJR model that have also elected the Medicare hospice benefit, hospice services would need to adapt and respond to the care needs of the CJR beneficiary following surgery. As in the case of other medically necessary services that would improve a beneficiary's quality of care and quality of life, we expect that CJR model beneficiaries will receive clinically appropriate referrals to hospice in a timely manner. Furthermore, we also believe hospice services could contribute to episode efficiency through improved comprehensive care coordination and management for CJR model beneficiaries that have a terminal prognosis. As previously stated, hospices are required to provide comprehensive care coordination and management per the hospice CoPs at 418.56. As discussed in sections III.F.3. and 5. of this final rule, we will be monitoring for access to care and delayed care and will take actions as described if problems are found. Therefore, we are finalizing our proposal to include hospice services in the CJR model episode definition.

With regard to the commenters' request for data regarding hospice use and the CJR model, we note that the evaluation approach described in IV.D. of this final rule will yield utilization information on CJR beneficiaries' episodes for specific types of providers and services. As discussed in section IV.E. of this final rule, we plan to evaluate the CJR model on an annual basis and release internal periodic summaries to offer useful insight, with Start Printed Page 73308a final analysis after the end of the 5-year performance period. Finally, we plan to make available to participant hospitals upon their request periodic summary claims data reports or raw claims data, including payment information, using type of service categories that including hospice. We refer readers to section III.E.2. of this final rule for a more detailed discussion of the plans for sharing data under the CJR model.

Comment: Several commenters requested that CMS exclude prosthetic limbs, orthopedic braces, and customized durable medical equipment (DME) from the related services included in CJR model episodes. The commenters stated that these uncommonly furnished items were at risk of not being provided to CJR model beneficiaries, and provided historical example of access problems during implementation of the SNF PPS that eventually resulted in some HCPCS codes for these items being exempted from SNF consolidated billing. Another commenter requested clarification about included services with respect to the definition of DME. The commenter expressed its belief that there would be no need for verification by CMS or its contractors about coverage of DME as CMS would be making a single episode payment to hospitals. The commenter sought clarification that devices that would usually be paid for under the MS-DRG payment should be able to be used in the CJR beneficiary's home.

Response: While some commenters recommended that we exclude altogether certain prosthetics, braces, and customized DME from the episode definition under the CJR model, we believe that our Part B ICD-9-CM (or equivalent ICD-10-CM) diagnosis code exclusions will allow these items to be excluded when they are unrelated to the episode., both in determining historical CJR episode payments used to set the target price and in calculating actual episode spending during the model performance years Just as for other Part B services, when the primary ICD-9-CM (or equivalent ICD-10-CM) diagnosis code on the claim for the item is not excluded, the prosthetics, orthopedic braces, and customized DME will be included in the CJR episode. Because we will identify unrelated items when they are furnished, and the Medicare payment for those items will not be included in calculating the actual episode spending, we believe that CJR model beneficiaries will continue to have access to these items when they are furnished for unrelated diagnoses on the Part B ICD-10-CM diagnosis code exclusions list. With regard to the commenter who discussed a single payment by CMS to hospitals for the episode, we want to emphasize that this is a retrospective payment model and, thus, payments for all covered items and services will continue to be made under the usual Medicare program rules to all providers and suppliers furnishing services to CJR model beneficiaries, unless we have specifically waived certain Medicare program rules under the CJR model. We refer readers to section III.C.11. of this final rule for further discussion of waivers of Medicare program rules, but note that we have waived no existing requirements or conditions about DME. All existing program rules for coverage and payment of DME continue to apply. Therefore, we are finalizing our proposal to include DME in the CJR model episode definition, after application of the exclusions.

Comment: A number of commenters commended CMS on the proposal to exclude IPPS new technology add-on payments from the CJR model episode definition, as well as hemophilia clotting agents furnished to hospital inpatients. The commenters believe these policies will ensure access to these important treatments for CJR model beneficiaries who would benefit from them. Several commenters suggested that CMS also exclude from the CJR model episode definition OPPS transitional pass-through payments for devices, which are paid separately for a limited period of time based on their increased cost over existing technologies and evidence that they are a substantial clinical improvement, for consistency with CMS' proposed treatment of IPPS new technology add-on payments which accomplish the same objective for hospital inpatients. Other commenters recommended that CMS exclude other innovative technologies from the episode definition by establishing a review process to see if their cost should be removed from CJR episode spending to ensure that the financial incentives under the CJR episode payment model did not discourage appropriate use of new technologies for CJR model beneficiaries who would benefit from them. These commenters stated that such a policy would ensure that beneficiaries in the CJR model have access to beneficial new technologies that otherwise might be limited because of participant hospitals' concerns over providing items and services that would increase actual episode spending. A commenter, arguing in support of CMS' proposal to exclude IPPS new technology add-on payments from the episode definition, suggested that CMS analyze Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) data to see if customized joints correlated with HCAHPS scores under the model.

Response: We agree with the commenters that CJR model beneficiaries should have access to beneficial new technologies while they are in CJR episodes. We do not believe it would be appropriate for the CJR model to potentially hamper beneficiaries' access to new technologies that are receiving IPPS new technology add-on payments or to burden hospitals who choose to use these new drugs, technologies, or services with concern about these payments counting toward actual episode expenditures. We also agree with the commenters' recommendation that we should exclude OPPS pass-through payments for medical devices from the episode definition for the same reasons we proposed to exclude IPPS new technology add-on payments. In both of these cases, through the established OPPS and IPPS review processes, we have determined that these technologies have a substantial cost but also lead to substantial clinical improvement for beneficiaries. Therefore, we are finalizing our proposal to exclude from the CJR episode definition IPPS new technology add-on payments and hemophilia clotting factors paid separately during an inpatient hospitalization. In addition, we are modifying our proposal and will exclude OPPS transitional pass-through payments for medical devices from the CJR model episode definition and price determinations.

We will not establish a new process to review innovative technologies and make individual determinations regarding their exclusion from the CJR model episode definition, as recommended by some commenters. Because the CJR model is a retrospective reconciliation model that pays all providers and suppliers under the regular Medicare program throughout the episode of care, we believe it is more appropriate to rely on the existing processes under the Medicare program to make determinations about separate payment for new technology items and services. If those existing processes identify new technologies that would qualify for add-on payments under the IPPS or transitional pass-through payment under the OPPS, we will exclude them from the CJR model episode definition to ensure that access to new technology items and services for beneficiaries is not influenced by their care being include in the CJR model. We note that the evaluation Start Printed Page 73309approach for the model as discussed in section IV. of this final rule will analyze a variety of information about the model to draw conclusions about its effects on quality and cost but is not designed to examine patient experience as related to specific items or services furnished during the episode.

Comment: Most commenters expressed support for CMS' proposed episode definition that would exclude certain readmissions based on a list of MS-DRGs, as well as certain Part B services based on the principal diagnosis on the claim, consistent with the episode definition for LEJR episodes under BPCI that has been used for several years. The commenters acknowledged that most services would be included in the episode definition under the proposal, thus creating broadly defined episodes that should lead to comprehensive care for beneficiaries following LEJR procedures. A number of commenters characterized the proposed episode definition as clinically reasonable and agreed with the proposed lists of services that would be excluded. A commenter claimed that the proposed episode definition would encourage the integration of post-fracture care coordination, such as could be provided through a fracture liaison service, with acute care for CJR model beneficiaries with hip fractures, leading to improved outcomes. However, some commenters expressed general concern about CMS' proposal to hold participant hospitals financially accountable for these broadly defined episodes, especially as CMS did not propose to risk adjust target prices for the episodes to reflect beneficiaries' chronic conditions.

Several commenters suggested that CMS adopt an episode definition for the CJR model that is flexible and condition-specific. A commenter questioned the role of the beneficiary's health care provider in evaluating relatedness to the episode under the proposal and recommended that CMS permit the beneficiary's health care provider to make determinations of relatedness of services to the episode on a case-by case basis specific to a beneficiary's unique clinical condition. A few commenters suggested that CMS' proposed episode definition was more consistent with a total cost of care model by including beneficiaries with chronic conditions and excluding so few services. These commenters stated that if CMS finalizes such a broad definition, risk adjustment would be necessary in order to ensure fair payment to participant hospitals. Some commenters contended that CMS should include in the episode definition only services that are directly related to the procedure and complications for which the hospital could be held accountable. In the view of some commenters, CMS should exclude all chronic conditions from the episode definition, especially when the LEJR episode is unavoidable, such as in trauma cases. Examples provided by commenters of chronic conditions that should be excluded include diabetes and renal failure. Other commenters recommended that CMS only exclude care for unrelated chronic conditions and acute medical conditions such as urinary tract infection and dehydration occurring later than 30 days following discharge from the anchor hospitalization or otherwise shorten the episode duration of the model to 30 days. They claimed that holding the participant hospital accountable through the episode definition for chronic conditions two months after surgery is unfair. A commenter recommended that CMS include all readmissions for the first 30 days following discharge from the anchor hospitalization and thereafter only those hospital readmissions for the subsequent 60 days that are directly related to the LEJR procedures. Overall, a number of commenters expressed concern that unless CMS narrowed the proposed CJR model episode definition to exclude more services or diagnoses or shortened the episode duration, hospitals may be more cautious about treating patients with complex medical status, especially if CMS also does not risk adjust the target prices for the episode based on beneficiary characteristics and specific procedures.

A commenter stated that the proposed episode definition was not sufficiently broad for frail patients, especially those with multiple illnesses who may have had a hip fracture. The commenter contended that providers should be paid to provide comprehensive care and treat the whole person, who can have many different types of interrelated health care needs when he or she is acutely ill due to a hip fracture in the face of serious underlying chronic conditions. The commenter stated that the CJR model would contribute to the fracturing of comprehensive care for vulnerable beneficiaries by excluding some services from the episode definition, even if those services are for clinical conditions that appear to be clinically unrelated to the LEJR episode, and claimed that the solution to this challenge is moving people with complex medical needs into a patient-centered medical home or comprehensive ACO. The commenter stressed that any existing medical home or ACO arrangements that apply to CJR model beneficiaries should be respected by the participant hospital managing the CJR episode, so as to not disrupt or otherwise interfere with comprehensive care for beneficiaries with complex medical needs.

Response: We appreciate the support of many commenters for our proposed overall approach of identifying excluded services by MS-DRGs for hospital readmissions and ICD-9-CM (or equivalent ICD-10-CM) diagnosis codes for Part B services for LEJR episodes that are broadly inclusive of related services. Because the methodology for setting episode prices as discussed in section III.C. of this final rule requires the construction of historical CJR episodes upon which to base target prices that are then compared with actual episode payment during each performance year of the model, we must use a standard episode definition for the CJR model to ensure comparability of services included in the episode in the historical CJR episode data and the model performance year. Thus, we are unable to adopt the suggestions of commenters that the CJR model episode definition be flexible or that health care providers make service-by-service determinations of relatedness for individual beneficiaries.

As discussed in the proposed rule and confirmed by the commenters, beneficiaries undergoing LEJR procedures have frequent comorbidities where their management may be affected by the surgery and post-operative recovery period. We do not believe it would be appropriate given the frequent comorbidities experienced by Medicare beneficiaries and the generally elective nature of LEJR to utilize a narrow episode definition for CJR that includes only those services directly related to the LEJR procedure or the quality or safety of the LEJR care, as we are interested in testing inclusive episodes to incentivize comprehensive, coordinated patient-centered care for the beneficiary throughout the episode. The care for many chronic conditions and the development of acute medical conditions may be affected by the LEJR procedure or post-surgical care throughout the post-surgical recovery period that extends significantly beyond 30 days following hospital discharge, a point in time where beneficiaries are usually still receiving PAC services, often including SNF services, and have not returned to their level of presurgical function. Therefore, we do not believe it would be appropriate to define services for chronic conditions and acute medical conditions as related to the CJR Start Printed Page 73310model episode for 30 days post-discharge from the anchor hospitalization, and unrelated for the remaining 60 days in the episode. We believe that care for chronic medical conditions affected by the LEJR procedure or post-surgical care is related to the episode for the full episode duration because the care for these conditions is likely to be affected by the procedure and associated recovery for 90 days post-hospital discharge or even longer as the beneficiary recovers function over the course of the episode and returns to the community. We note that we have finalized several waivers of Medicare program rules as discussed in section III.C.11. of this final rule specifically to assist participant hospitals in efficient and effective care coordination and care management for CJR beneficiaries with significant, ongoing health needs, including chronic medical conditions whose care may be affected by the LEJR procedure and post-surgical recovery. Thus, we will exclude only those Medicare Part A and B-covered items and services furnished during the episode that are unrelated to LEJR procedures based on clinical justification, and the exclusions will apply throughout the episode duration. Finally, we believe that the payment policies of the model as described in sections III.C.3.c. and III.C.8. of this final rule to adjust pricing for high payment episodes and to provide stop-loss limits provide sufficient protections for participant hospitals from excessive financial responsibility for high payment cases that may result from the broad episode definition adopted for the model. We expect that participant hospitals, with responsibility for the quality and cost performance of CJR model episodes, will work closely with all providers, suppliers, and organizations engaged in the care of model beneficiaries, in order to ensure that efficient, coordinated care is furnished to the beneficiary.

We appreciate the concerns expressed by commenters about holding participant hospitals financially responsible for broad LEJR episodes extending 90 days post-discharge from the anchor hospitalization. We note that we are finalizing 90 days post-discharge from the anchor hospitalization as we proposed for the reasons discussed later in this section. Additionally, we refer readers to section III.C.4.b. of this final rule for the final policy that will risk stratify the target prices based on the presence or absence of a hip fracture for CJR model beneficiaries. We believe that this risk stratification policy addresses the commenters' concerns that beneficiaries with chronic conditions are likely to need more costly care throughout the CJR model episode that would have been inadequately paid under our proposal because these beneficiaries are those most likely to be present in the population receiving LEJR procedures emergently due to a hip fracture. Beneficiaries with chronic conditions are more likely to initiate CJR episodes due to hip fracture than beneficiaries without chronic condition who more likely undergo elective THA or TKA, so the typically higher historical spending for chronically ill beneficiaries will be reflected in the historical CJR episodes used to risk stratify target prices for hip fracture patients. In contrast, beneficiaries undergoing elective THA or TKA are less likely to have chronic conditions, so their typically lower historical spending will be reflected in the historical CJR episodes used to risk stratify target prices for LEJR patients without hip fracture. Thus, risk stratification of target prices based on a beneficiary's hip fracture status should account for patient-specific expenditure variation both directly resulting from more intense care due to the hip fracture itself, as well as indirectly resulting from the higher prevalence of chronic conditions that must be treated and managed in beneficiaries with hip fracture. We also believe that risk stratification based on a model beneficiary's hip fracture status will help to ensure that participant hospitals continue to treat these medically complex patients because target prices for these episodes will reflect the more costly care that these beneficiaries are likely to require based on historical experience.

Additionally, while we agree with the commenter that the ongoing and acute health care needs of medically complex beneficiaries may be addressed through a patient-centered medical home or ACO, many of these vulnerable beneficiaries currently are not included in such models or programs. In the case of other beneficiaries who are included in medical home or ACO models or programs, they may have specific, new care management needs arising from an LEJR procedure that may be best managed by the participant hospital that has substantial expertise in coordinating and managing care throughout LEJR episodes because of the hospital's participation in the CJR model, while the ACO or patient-centered medical home may have less specific expertise in managing beneficiaries recovering from major orthopedic surgery. We expect that participant hospitals, accountable for LEJR episode quality and cost performance under this model, will work closely with all providers and other organizations with which a model beneficiary has established relationships, toward the mutual goal of high quality, well-coordinated care that maximizes the rate of a beneficiary's return of function following surgery.

We are finalizing our proposal to include all Medicare Part A and B items and services in the CJR model episode definition, except for excluded services identified by the CJR model exclusions list, with modification to additionally exclude OPPS transitional pass-through payments for devices.

Comment: Many commenters expressed support for CMS' proposed approach to identifying excluded services by MS-DRGs for readmissions and ICD-9-CM diagnosis codes on Part B claims. Some commenters suggested that CMS consider additional coding sources beyond ICD-9-CM diagnosis codes to identify exclusions by adding ICD-9-CM procedure codes and HCPCS and/or CPT codes to the list of Part B exclusions.

Response: We appreciate the commenters' support for our proposal. We note that we have successfully used our current approach to identify excluded services for 48 clinical episodes under BPCI Models 2, 3, and 4 for several years. We will consider whether supplementing our current approach to identifying excluded services with additional coding strategies could help us more accurately identify unrelated services as we review future stakeholder input about the CJR model episode definition. We would need to also take into consideration the current coding requirements for different Part A and Part B services in assessing the potential benefit of supplementing our existing approaches to identifying excluded services. We would address any changes to the current CJR model approach to identifying excluded services through rulemaking. Therefore, we are finalizing our proposal to identify CJR model excluded services by MS-DRGs for readmissions and ICD-9-CM (or equivalent ICD-10-CM) diagnosis codes for Part B services.

Comment: A number of commenters provided their perspective on certain specific proposed related services and exclusion. Several commenters expressed support for CMS' proposal to exclude readmissions for trauma medical and oncology MS-DRGs from the CJR episode definition. The commenters agreed with CMS that readmissions during LEJR episodes for Start Printed Page 73311the clinical conditions that would result in discharge from trauma medical or oncology MS-DRGs would be clinically unrelated to the LEJR episode. A commenter recommended that CMS exclude rheumatoid arthritis care from the LEJR episode definition. While the commenter pointed out that rheumatoid arthritis can result in the need for LEJR procedures, the commenter observed that including treatment for rheumatoid arthritis in the episode would result in the accompanying high payments for this care being included in actual episode spending. The commenter stated that the high costs of treatment could either affect a beneficiary's treatment for rheumatoid arthritis during the CJR model episode or reduce the beneficiary's access to a medically necessary joint replacement. Several commenters recommended that CMS exclude services for which beneficiary claims data are not made available, specifically those subject to the regulations governing the confidentiality of alcohol and drug abuse patient records (42 Code of Federal Regulations (CFR) part 2). Other commenters suggested that CMS exclude elective surgery during the CJR model episode, providing examples of cataract surgery, hernia repair, gallbladder procedures, and transurethral resection of the prostate. A commenter requested that CMS add the ICD-9-CM procedure code for chemotherapy administration to the Part B exclusions list, because CMS proposed to consider chemotherapy to be unrelated and, therefore, excluded from the CJR episode definition.

Several commenters requested further justification of CMS's proposals to include all body system-related surgical MS-DRGs and medical MS-DRGs except oncology and trauma medical MS-DRGs in the CJR episode definition. Several commenters requested further rationale for CMS' proposal to include all PAC services in the episode following an excluded readmission. Another commenter requested clarification on the inclusion of communication, cognitive, and swallowing-related diagnoses in the LEJR episode and CMS' intent in bundling services the commenter believes to be unrelated. The commenter also requested information about how providers could submit clinical justification when an exclusion of therapy services from the CJR model episode is needed. Finally, several commenters expressed support for excluding patients from the model with acute disease diagnoses such as head injury, based on their conclusion that CMS proposed to exclude these beneficiaries due to CMS' proposed exclusion of Part B claims reporting acute disease diagnoses, such as severe head injury.

Response: We appreciate the specific requests by the commenters for clarification and modification of our proposed list of exclusions from the CJR model episode definition. We agree with the commenters who supported our proposal to exclude readmissions resulting in discharges from oncology and trauma medical MS-DRGs. While we believe that readmissions for medical MS-DRGs are generally linked to the hospitalization for the LEJR procedure as a complication of the illness that led to the surgery, a complication of treatment or interactions with the health care system, or a chronic illness that may have been affected by the course of care, we agree with the commenters that hospitalizations resulting in discharge from oncology and trauma medical MS-DRGs are not related to the hospitalization for the LEJR procedure.

We do not believe that Part B claims including ICD-9-CM diagnosis codes for rheumatoid arthritis should be excluded from CJR model episodes. This chronic condition is likely to be affected by care during the procedure and recovery period and, therefore, we would consider claims reporting these diagnoses codes to be related to the LEJR episode. With regard to the commenter's concerns about delays in timely treatment as a result of high treatment costs and reduced access to joint replacement procedures for beneficiaries with rheumatoid arthritis, we refer readers to sections III.F.3. and 5. of this final rule for discussion of our plans to monitor for access to care and delayed care due to the potential of the CJR model to direct patients away from more expensive services at the expense of outcomes and quality. We will also not exclude claims for substance abuse and mental health services that are not available in beneficiary claims data because these services are clinically related to LEJR episodes. Claims for substance abuse and mental health services include care for clinical conditions that are related to the CJR episode because these conditions may be affected by the LEJR procedure or post-surgical care. With regard to the commenters' requests that we exclude elective procedures such as cataract surgery, hernia repair, gallbladder procedures, and transurethral resection of the prostate from the CJR model episode definition, while we believe these procedures will be uncommon during the post-surgical recovery period for CJR model beneficiaries that extends 90 days following discharge from the anchor hospitalization, we will not exclude them as unrelated because all of the procedures may be related to care furnished during the post-surgical recovery period. Our exclusion methodology does not allow us to identify those procedures that are truly elective; that is, the condition was present and surgery was planned prior to the LEJR procedure and scheduled during the 90-cay post-hospital discharge period.

While we agree with the commenter that chemotherapy services should be excluded from the CJR model episode, our exclusion methodology for Part B services does not rely upon ICD-9-CM procedure codes but instead upon ICD-9-CM (or equivalent ICD-10-CM) diagnosis codes reported on Part B claims. We note that the Part B payment systems, including those for physicians' services, Part B drugs, and institutional services, reject claims that do not report valid ICD-9-CM diagnosis codes. Therefore, we believe that our proposal to base Part B exclusions only on ICD-9 diagnosis codes and not additionally upon ICD-9 procedure codes should allow us to identify and exclude from the CJR episodes all Part B claims for chemotherapy administration services. Providers and suppliers do not report ICD-9-CM (or equivalent ICD-10-CM) procedure codes on Part B claims because they are paid for their chemotherapy and other services on the basis of the CPT or HCPCS codes that describe those services. However, these Part B claims must also include ICD-9-CM (or equivalent ICD-10-CM) diagnosis codes. CMS requires ICD-9-CM (or equivalent ICD-10-CM) procedure codes to be reported only on Part A claims, which are excluded from the CJR model on the basis of readmission MS-DRG rather than ICD-9 (or equivalent ICD-10) codes, so adding ICD-9-CM (or equivalent ICD-10-CM) procedure codes to the Part B exclusions list is not necessary.

As we stated in the proposed rule, for readmissions to medical MS-DRGs the selection of the primary diagnosis code is not clear-cut so we believe they should all be included in the episode definition so that providers focus on comprehensive care to beneficiaries in episodes. We reiterate our belief that readmissions to medical MS-DRGs are generally linked to the hospitalization for the procedure as a complication of the illness that led to the surgery, a complication of treatment or interactions with the health care system, or a chronic illness that may have been affected by the course of care. Moreover, Start Printed Page 73312we believe that all body-system related surgical MS-DRGs for readmissions are also related to the LEJR episode because these readmissions are generally related to complications of the LEJR procedure. Such surgeries result from the treatment of systemic conditions that arise from the LEJR procedure or its complications. Examples include placement of an inferior vena cava filer or a percutaneous coronary intervention for treatment of thromboembolic complications of the LEJR procedure.

We did not propose to exclude any PAC services in the 90-day post-hospital discharge period, even when those PAC services follow an excluded readmission. As Part A services are generally intended to be comprehensive in nature and because the beneficiary in a CJR model episode would still be in the post-operative recovery period following LEJR surgery, we believe any PAC services provided during the episode would be related to the LEJR procedure. Regardless of the reason for the hospitalization immediately preceding the initiation of PAC services, the PAC provider would need to address the beneficiary's post-surgical recovery from the LEJR procedure, even if the PAC services immediately followed an unrelated readmission to the hospital.

We did not propose to exclude claims for Part B services for communication, cognitive, or swallowing-related diagnoses from the CJR model episode definition because we believe these diagnoses are due either to chronic conditions whose care may be affected by the LEJR procedure or post-surgical care or to complications of the procedure, such as stroke, that result in these diagnoses. Therefore, we consider all Part B claims reporting these diagnoses in the principal diagnosis field to be related to the CJR episode. Providers are unable to submit clinical justification or other special requests for services to be designated as unrelated to the episode if one of these diagnoses is in the principal diagnosis field on claims. The CJR model is testing LEJR episode payment and we need consistency in the scope of the episode for the model. We will include all related Part A and Part B services as identified in this final rule in the calculation of episode target prices based on historical CJR episode data and in the calculation of actual episode spending for a model performance year.

Finally, in response to the commenters who supported the exclusion of beneficiaries with acute disease diagnoses, such as head injury, from the CJR model, we want to clarify that we did not propose to exclude these beneficiaries from the model. Instead, we proposed to exclude Part B claims reporting acute disease diagnoses from the episode because we consider these services to be unrelated under the episode definition. Therefore, we will not include claims for Part B services reporting excluded acute disease ICD-9-CM (or equivalent ICD-10-CM) diagnosis codes in calculating target prices based on historical CJR episodes or in calculating actual episode spending that will be compared to the episode's target price in the CJR model.

We are finalizing our proposal to exclude the specific list of MS-DRGs for readmissions and ICD-9-CM (or equivalent ICD-10-CM) diagnosis codes that is posted on the CMS Web site at: http://innovation.cms.gov/​initiatives/​cjr/​.

Comment: A commenter requested that CMS clarify how it will address hospital-acquired conditions that should never occur, when these conditions are part of CMS' Hospital-Acquired Condition Reduction Program and experienced by CJR model beneficiaries. The commenter explained that under current Medicare program policy, Medicare will not pay the higher MS-DRG arising from a specified list of non-reimbursable hospital-acquired conditions. The commenter pointed out that CMS proposed to not exclude claims for diagnoses related to the quality and safety of care furnished during the episode in the CJR model episode definition, but CMS' list of non-reimbursable hospital-acquired conditions includes surgical site infections after certain orthopedic procedures. In addition to clarifying how never events will be addressed in setting payments under the CJR model, the commenter recommended that CMS incorporate an analysis of never events and their incidence into the reconciliation process and review whether to expand the list of never events for elective surgeries.

Another commenter recommended that the CJR episode include a warranty for complications associated with surgery and other treatment, that is, if complications occur, they should be treated at no additional cost to the patient or Medicare.

Response: We appreciate the commenter's request for clarification about treatment of IPPS claims that include hospital-acquired conditions under the CJR model. Our model policy as discussed in section III.C.4. of this final rule bases the CJR target prices on historical CJR episodes that reflect discharge MS-DRGs and paid claim amounts for those beneficiaries who would have begun episodes by admission to an IPPS hospital that resulted in a discharge from MS-DRG 469 and 470. To the extent that Medicare does not pay the higher MS-DRG amount due to a hospital-acquired condition that was not present on admission, the lower payment for the hospitalization due to the hospital-acquired condition would be used in setting the episode target price for the MS-DRG that anchored the episode. This same would hold true for related readmissions during the episode. When calculating actual episode spending during a performance year, we would use, once again, the paid claim amount that, in the case of a hospital-acquired condition that was not present on admission, would be at the level of the lower paying MS-DRG for the anchor hospitalization or related readmission, as applicable. We further note that if a CJR beneficiary experiences a hospital-acquired condition that was not present on admission during an anchor hospitalization and has no other comorbid conditions other than the HAC that would result in assignment of MS-DRG 469, the beneficiary's episode would be considered an MS-DRG 470-anchored episode (initiated by the MS-DRG for LEJR procedures without complications). Therefore, the hospital-acquired condition penalty would not itself inflate the target price such that CMS would pay back the hospital-acquired condition penalty through a reconciliation payment.

Our proposal not to exclude claims for diagnoses related to the quality and safety of care during the episode is the basis for our excluded list of MS-DRGs for readmissions and ICD-9-CM (or equivalent ICD-10-CM) diagnosis codes for Part B services and, therefore, this list would not apply to the anchor hospitalization itself where hospital-acquired conditions that were not present on admission could be reported.

As discussed in sections III.C.5. and 6. of this final rule, the model evaluation will examine changes in utilization, as well as outcomes and quality, in order to assess the impact of the CJR model on the aims of improved care quality and efficiency as well as well as reduced health care costs. We refer readers to section IV. of this final rule for further information on the planned evaluation. We have an ongoing process to review claims data regarding potential candidates for additions to the list of hospital-acquired conditions, so we do not believe there is a need to specifically identify CJR episodes for analysis because the IPPS claims included in CJR episodes would already be considered in the ongoing process used by CMS in the Hospital-Acquired Condition Reduction Program.Start Printed Page 73313

In response to the commenter who recommended for the CJR model that if complications due the LEJR procedure occur, they should be treated at no additional cost to the patient or Medicare, we note that because the CJR model uses a retrospective payment approach, we will rely on the existing Medicare program policies under the Hospital-Acquired Condition Reduction Program that define the specific circumstances in which Medicare will not make additional payment for a condition occurring after surgery. When these circumstances occur for CJR model beneficiaries in episodes, the existing Medicare program policies apply and Medicare would not provide additional payment. We do not believe it would be appropriate to establish policies specific to the CJR model regarding Medicare nonpayment for other complications, and we further note that some complications may not be preventable. The final pay-for-performance methodology for the CJR model as discussed in section III.C.5. of this final rule provides strong financial incentives for participant hospitals to coordinate and manage care to reduce complications, as the THA/TKA Complications measure (NQF #1550) contributes half of the available points for the hospital's composite quality score that determines the hospital's eligibility for reconciliation payments and quality incentive payments.

Comment: Several commenters opposed CMS' proposal to make changes to CJR model exclusions through an annual, at a minimum, update outside of rulemaking. Most commenters recommended that CMS update the exclusions annually through rulemaking, at least for routine annual updates. Other commenters stated that they did see value in CMS making possible additions and deletion to the exclusions list on a quarterly basis, especially early in the model. If following a quarterly process outside of rulemaking, these commenters urged CMS to seek stakeholder comment and input on candidate revisions through the CMS Web site and list serves to ensure broad input. The commenters encouraged CMS to adopt a transparent process for revisions to the episode definition in considering other exclusions. A number of commenters recommended that CMS explore other exclusions for the future, such as those inpatient hospital admissions or outpatient procedures planned for the beneficiary prior to the episode, ongoing care for patients' chronic conditions, and PAC following an excluded hospital readmission.

Response: We appreciate the interest of the commenters' in ensuring that any changes to the CJR model episode definition involve a transparent process with opportunity for broad stakeholder input. We continue to believe that updating the exclusions annually, at a minimum, outside of rulemaking, is most appropriate for this 5-year model, allowing for more frequent updates than through rulemaking as necessary to accommodate timely ICD-CM annual coding changes and the transition to ICD-10-CM and annual IPPS MS-DRG changes, as well as to address significant issues raised by participant hospitals and other stakeholders.

Commenters who supported an exclusions list update process outside of rulemaking did not suggest specific revisions to our proposed criteria for updating the exclusions, namely that:

  • We would not exclude any items or services that are—

++ Directly related to the LEJR procedure itself (such as loosening of the joint prosthesis) or the quality or safety of LEJR care (such as post-surgical wound infection or venous thromboembolism); and

++ For chronic conditions that may be affected by the LEJR procedure or post-surgical care (such as diabetes). By this we mean that where a beneficiary's underlying chronic condition would be affected by the LEJR procedure, or where the beneficiary's LEJR or post-LEJR care must be managed differently as a result of the chronic condition, then those items and services would be related and would be included in the episode.

  • We would exclude items and services for—

++ Chronic conditions that are generally not affected by the LEJR procedure or post-surgical care (such as removal of the prostate). By this we mean that where a beneficiary's underlying chronic condition would not be affected by the LEJR procedure, or where the beneficiary's LEJR or post-LEJR care need not be managed differently as a result of the chronic condition, then those items and services would not be related and would not be included in the episode; and

++ Acute clinical conditions not arising from existing episode-related chronic clinical conditions or complications of LEJR surgery from the episode (such as appendectomy).

Thus, we continue to believe these criteria provide the appropriate clinical review framework for updates to the CJR model exclusions. Finally, we believe that our proposed process to post the potential revised exclusions, which could include additions to or deletions from the exclusions list, to the CMS Web site to allow for public input on our planned application of these standards, and then adopt changes to the exclusions list with posting to the CMS Web site of the final revised exclusions list after our consideration of the public input, is consistent with the recommendation of commenters that we use a transparent process reflective of robust opportunity for public input. Conducting this update process outside of rulemaking based on the criteria set forth in this final rule will allow us the greatest flexibility to update the exclusions as changes to the MS-DRGs and ICD diagnosis codes, upon which our exclusions rely, are released. This process will also allow us to respond quickly to any episode definition issues that arise during implementation of the model across the broad array of participant hospitals in the selected MSAs. We would widely publicize the opportunity for review and public input through the CMS Web site and listservs. We also note that any changes to our overall approach to identifying excluded services or to our criteria for evaluating services for exclusion would be addressed through rulemaking. Therefore, we are finalizing our proposal to update the exclusions list annually, at a minimum, using the process as described.

Comment: Several commenters referred to the impending change from ICD-9-CM to ICD-10-CM coding on claims and identified that this change would have implications for the Part B exclusions list. A commenters stated that CMS would need to define the excluded ICD-10-CM codes prior to implementation of the CJR model and recommended that CMS also provide the ICD-10-CM diagnosis code list that would identify included Part B services.

Response: We appreciate the commenters' interest in the list of CJR model exclusions that are identified based on ICD-10-CM codes. In the proposed rule, we stated that as we move to implement ICD-10-CM we would develop the CJR exclusions that would map to the final ICD-9-CM exclusions for CJR available in the ICD-10-CM format as well.

With ICD-10-CM implementation beginning in October 2015, we are making available the final CJR model Part B exclusions list in ICD-10-CM format as additional worksheet tabs to the final exclusions list posted on the CMS Web site at: http://innovation.cms.gov/​initiatives/​cjr/​. This is the same list of exclusions that will be used for LEJR episodes under BPCI. This list will be applied to claims for services furnished on or after October 1, 2015 and that Start Printed Page 73314report ICD-10-CM codes. For ease of understanding by the public, our objective was to present the ICD-10-CM excluded codes as ranges of excluded ICD-10-CM categories, just as we present the ICD-9-CM excluded codes as ICD-9-CM ranges.

To develop the ICD-10-CM exclusions list, we began with the list of final CJR ICD-9-CM code ranges. From that list of ranges, we generated an expanded list of all excluded ICD-9-CM codes. We then compared the list of excluded ICD-9-CM codes against both the ICD-9-CM-to-ICD-10-CM and ICD-10-CM-to-ICD-9-CM General Equivalence Mappings (GEMs) available at: https://www.cms.gov/​Medicare/​Coding/​ICD10/​2015-ICD-10-CM-and-GEMs.html. Comparing against both GEM files was necessary because there were matches in the ICD-9-CM-to-ICD-10-CM GEM that did not appear in the ICD-10-CM-to-ICD-9-CM GEM and vice versa. For example —

  • In the ICD-9-CM-to-ICD-10-CM GEM file, ICD-9-CM code 85110 (Cortex (cerebral) contusion with open intracranial wound, unspecified state of consciousness) maps to ICD-10-CM code S0190XA (Unspecified open wound of unspecified part of head, initial encounter), but there is not a corresponding map from S019XA to 85110 in the ICD-10-CM-to-ICD-9-CM GEM.
  • In the ICD-10-CM-to-ICD-9-CM GEM file, ICD-10-CM code A0101 (Typhoid meningitis) maps to ICD-9-CM code 020 (Typhoid), but there is not a corresponding map from 020 to A0101 in the ICD-9-CM to-ICD-10-CM GEM.

After compiling the results from both GEM files, we created a list of every billable ICD-10-CM code and whether each billable ICD-10-CM code matched to an excluded ICD-9-CM code. We then moved from the list of individual codes to a list of ICD-10-CM three-digit categories (for example, ICD-10-CM code A0101 (Typhoid meningitis) is in ICD-10-CM category A01 (Typhoid and paratyphoid fevers)) to present the final CJR exclusions. We excluded ICD-10-CM categories in which 100 percent of billable ICD-10-CM codes matched to an excluded ICD-9-CM code. There are 574 such categories, and we consider these CD-10-CM categories excluded based on a direct mapping from ICD-9-CM (see the “Excluded Part B ICD10 Direct” worksheet tab in the final exclusions list file). We did not exclude ICD-10-CM categories in which no billable ICD-10-CM codes matched to an excluded ICD-9-CM code. There are 1,258 categories, and we consider these categories not excluded based on a direct mapping from ICD-9-CM. For those 71 categories in which only some billable ICD-10-CM codes in the category matched to an excluded ICD-9-CM code after mapping, we excluded 48 ICD-10-CM categories where all of the ICD-10-CM codes in the category met one or more of our two final criteria for updating the excluded codes on the exclusions list as described previously in this section (see the “Excluded Part B ICD10 Medical” worksheet tab in the final exclusions list file). Specifically, the 48 ICD-10-CM categories that are excluded on this basis include ICD-10-CM codes that meet one or more of the following two criteria:

  • Chronic conditions that are generally not affected by the LEJR procedure or post-surgical care (such as removal of the prostate). By this we mean that where a beneficiary's underlying chronic condition would not be affected by the LEJR procedure, or where the beneficiary's LEJR or post-LEJR care need not be managed differently as a result of the chronic condition, then those items and services would not be related and would not be included in the episode.
  • Acute clinical conditions not arising from existing episode-related chronic clinical conditions or complications of LEJR surgery from the episode (such as appendectomy).

We did not exclude the 23 other ICD-10-CM categories in which only some billable ICD-10-CM codes in the category matched to an excluded ICD-9-CM code after mapping because the ICD-10-CM codes in these categories met one or more of the following criteria:

  • Directly related to the LEJR procedure itself (such as loosening of the joint prosthesis) or the quality or safety of LEJR care (such as post-surgical wound infection or venous thromboembolism).
  • For chronic conditions that may be affected by the LEJR procedure or post-surgical care (such as diabetes). By this we mean that where a beneficiary's underlying chronic condition would be affected by the LEJR procedure, or where the beneficiary's LEJR or post-LEJR care must be managed differently as a result of the chronic condition, then those items and services would be related and would be included in the episode.

When constructing prices for CJR, we will exclude Part B services from target prices and from performance year episodes based on the final excluded ICD-9-CM code ranges and final excluded ICD-10-CM code categories as appropriate, based on the applicable version of ICD diagnosis coding at the time the services was furnished.

In addition, we have addressed changes to the CJR model exclusion list that result from revisions for the FY 2016 IPPS. From FY 2015 to FY 2016, there were few changes to IPPS MS-DRGs that appear on the MS-DRG excluded readmissions list for the CJR model. Specifically, the FY 2016 IPPS update contains changes to existing MS-DRGs 237 and 238, Major Cardiovascular Procedures with MCC and without MCC, respectively, which are on the exclusions list for CJR episodes. For discharges after October 1, 2015, inpatient stays that previously would have been assigned to MS-DRG 237 or 238 will be assigned to one of the following MS-DRGs:

  • 268 Aortic and Heart Assist Procedures Except Pulsation Balloon with MCC.
  • 269 Aortic and Heart Assist Procedures Except Pulsation Balloon without MCC.
  • 270 Other Major Cardiovascular Procedures with MCC.
  • 271 Other Major Cardiovascular Procedures with CC.
  • 272 Other Major Cardiovascular Procedures without CC/MCC.

We also note that the list of excluded readmissions posted with the proposed rule inadvertently omitted MS-DRGs 490 and 491, which were eliminated in the FY 2015 IPPS Final Rule and from which MS-DRGs 518, 519, and 520 were created in FY 2015. We are adding MS-DRGs 490 and 491 to the list of excluded readmissions posted with this final rule as we will exclude readmissions in MS-DRGs 490 and 491 for the purposes of calculating CJR target prices.

Additional information on the new MS-DRGs is provided in the FY 2016 IPPS final rule (80 FR 49371 through 49390, available at: https://www.cms.gov/​Medicare/​Medicare-Fee-for-Service-Payment/​AcuteInpatientPPS/​FY2016-IPPS-Final-Rule-Home-Page.html). When constructing prices for CJR, we will exclude readmissions for MS-DRGs 237 and 238 in historical data. We will also exclude readmissions for MS-DRGs 268, 269, 270, 271, and 272 from performance year episodes.

Summary of Final Decisions: After consideration of the public comments we received, we are adding the following new definition for the CJR model: “Provider of outpatient therapy services” means a provider or supplier furnishing—(1) Outpatient physical therapy services as defined in § 410.60 of this chapter, or (2) outpatient occupational therapy services as defined in § 410.59 of this chapter, or (3) outpatient speech-language pathology Start Printed Page 73315services as defined in § 410.62 of this chapter.

We are finalizing our proposal, with modification to remove the term “independent” preceding outpatient therapy services, that related items and services included in CJR episodes, defined by all of the clinical conditions requiring an admission to an IPPS hospital that results in a discharge from MS-DRG 469 or 470 would be the following items and services paid under Medicare Part A or Part B, after the final exclusions are applied:

  • Physicians' services.
  • Inpatient hospital services (including readmissions), with certain exceptions, as discussed later in this section.
  • IPF services.
  • LTCH services.
  • IRF services.
  • SNF services.
  • HHA services.
  • Hospital outpatient services.
  • Outpatient therapy services.
  • Clinical laboratory services.
  • DME.
  • Part B drugs.
  • Hospice.

Medicare spending for related items and services will be included in the historical data used to set episode target prices, as well as in the calculation of actual episode spending that would be compared against the target price to assess the performance of participant hospitals. In contrast, Medicare spending for unrelated items and services (excluded from the episode definition) will not be included in the historical data used to set target prices or in the calculation of actual episode spending.

Additionally, we are finalizing our proposal to exclude inpatient hospital readmissions based on the list of excluded MS-DRGs and Part B services that report an excluded ICD-9-CM (or equivalent ICD-10-CM) diagnosis code as the principal diagnosis based on the list posted on the CMS Web site at: http://innovation.cms.gov/​initiatives/​cjr/​. As we proposed, we will exclude IPPS new technology add-on payments for drugs, technology, and services and hemophilia clotting factors paid separately during an inpatient hospitalization from the CJR model episode definition. We are modifying our proposal and, under our final policy, we will also exclude OPPS transitional pass-through payments for devices. We are also finalizing our proposal to update the exclusions list without rulemaking on an annual basis, at a minimum, to reflect annual changes to ICD-CM coding and annual changes to the MS-DRGs under the IPPS, as well as to address any other issues that are brought to our attention by the public throughout the course of the model test.

We will first develop potential exclusions list revisions of MS-DRGs for readmissions and ICD-9-CM (or equivalent ICD-10-CM) diagnosis codes for Part B services based on our assessment against the following standards:

  • We would not exclude any items or services that are—

++ Directly related to the LEJR procedure itself (such as loosening of the joint prosthesis) or the quality or safety of LEJR care (such as post-surgical wound infection or venous thromboembolism); and

++ For chronic conditions that may be affected by the LEJR procedure or post-surgical care (such as diabetes). By this we mean that where a beneficiary's underlying chronic condition would be affected by the LEJR procedure, or where the beneficiary's LEJR or post-LEJR care must be managed differently as a result of the chronic condition, then those items and services would be related and would be included in the episode.

  • We would exclude items and services for—

++ Chronic conditions that are generally not affected by the LEJR procedure or post-surgical care (such as removal of the prostate). By this we mean that where a beneficiary's underlying chronic condition would not be affected by the LEJR procedure, or where the beneficiary's LEJR or post-LEJR care need not be managed differently as a result of the chronic condition, then those items and services would not be related and would not be included in the episode; and

++ Acute clinical conditions not arising from existing episode-related chronic clinical conditions or complications of LEJR surgery from the episode (such as appendectomy).

We will post the potential revised exclusions, which could include additions to or deletions from the exclusions list, to the CMS Web site to allow for public input on our planned application of these standards, and then adopt changes to the exclusions list with posting to the CMS Web site of the final revised exclusions list after our consideration of the public input. Through the process for public input on potential revised exclusions and then posting of the final revised exclusions, we will also provide information to the public about when the revisions would take effect and to which episodes they would apply. These parameters could vary, depending on the relationship of exclusion list changes to annual ICD-CM or MS-DRG changes or to other issues brought to our attention by the public. While these revised exclusions may correspond to the time when we provide new target prices for a performance year, depending on the timing of when they would take effect and to which episodes they would apply, we would recalculate target prices as necessary.

The final definitions are set forth in § 510.2 which has been revised to remove proposed (b)(3) for inpatient hospital readmission services because hospital readmissions are already referenced in (b)(2). The remaining provisions under § 510.2(b) have been renumbered accordingly. The final policies for included services, excluded services, and updating the lists of excluded services are set forth in § 510.200(b), (d), and (e). We note that § 510.200(d)(3) has been renumbered to § 510.200(d)(4) and § 510.200(d)(3) added to state, “Transitional pass-through payments for medical devices as defined in § 419.66 of this chapter.” In addition, § 510.200(b)(10) has been modified to read “Outpatient therapy services.”

3. Duration of Episodes of Care

a. Beginning the Episode and Beneficiary Care Inclusion Criteria

While we proposed to identify LEJR episodes by an acute care hospitalization for MS-DRG 469 and 470, we recognize that the beneficiary's care for an underlying chronic condition, such as osteoarthritis, which ultimately leads to the surgical procedure, typically begins months to years prior to the surgical procedure. Because of the clinical variability leading up to the joint replacement surgery and the challenge of identifying unrelated services given the multiple chronic conditions experienced by many beneficiaries, we did not propose to begin the episode prior to the anchor hospitalization (that is, the admission that results in a discharge under MS-DRG 469 or 470). In the proposed rule, we stated our belief that the opportunities for care redesign and improved efficiency prior to the inpatient hospitalization are limited for an episode payment model of this type that focuses on a surgical procedure and the associated recovery once the decision to pursue surgery has been made, rather than an episode model that focuses on decision-making and management of a clinical condition itself (such as osteoarthritis).

We proposed to begin the episode with an inpatient anchor hospitalization for MS-DRG 469 or MS-DRG 470 in Start Printed Page 73316accordance with the methodology described. This proposal to begin the episode upon admission for the anchor hospitalization is consistent with LEJR episode initiation under Model 2 of BPCI. While we did not propose to begin the episode prior to the inpatient hospital admission, we noted that our proposed episode definition includes all services that are already included in the IPPS payment based on established Medicare policies, such as diagnostic services (including clinical diagnostic laboratory tests) and nondiagnostic outpatient services related to a beneficiary's hospital admission provided to a beneficiary by the admitting hospital, or by an entity wholly owned or wholly operated by the admitting hospital (or by another entity under arrangements with the admitting hospital), within 3 days prior to and including the date of the beneficiary's admission. For more information on the 3-Day Payment Window payment policies, see CMS Pub. 100-04, Chapter 3, section 40.3 and Chapter 4, section 10.12.

We proposed that the defined population of Medicare beneficiaries whose care will be included in CJR meet the following criteria upon admission to the anchor hospitalization. We noted that these criteria are also consistent with Model 2 of BPCI, as well as most other Innovation Center models that do not target a specific subpopulation of beneficiaries. We proposed that the LEJR episodes for all beneficiaries in the defined population will be included in CJR (although we proposed that certain episodes may be canceled for purposes of determining actual episode payments for reasons discussed later in this final rule), and we refer readers to section III.F.2. of this final rule for further discussion of beneficiary notification and a beneficiary's ongoing right under CJR to obtain health services from any individual or organization qualified to participate in the Medicare program.

  • The beneficiary is enrolled in Medicare Part A and Part B throughout the duration of the episode.
  • The beneficiary's eligibility for Medicare is not on the basis of End Stage Renal Disease (ESRD).
  • The beneficiary must not be enrolled in any managed care plan (for example, Medicare Advantage, Health Care Prepayment Plans, cost-based health maintenance organizations).
  • The beneficiary must not be covered under a United Mine Workers of America health plan, which provides healthcare benefits for retired mine workers.
  • Medicare must be the primary payer.

Our proposal for inclusion of beneficiaries in CJR is as broad as feasible, representing all those LEJR episodes for which we believe we have comprehensive historical Medicare payment data that allow us to appropriately include Medicare payment for all related services during the episode in order to set appropriate episode target prices. For beneficiaries whose care we proposed to exclude from the model, we are unable to capture or appropriately attribute to the episode the related Medicare payments because of Medicare's payment methodology. For example, if a beneficiary is enrolled in a Medicare Advantage plan, Medicare makes capitated payments (and providers do not submit complete claims data to CMS), so we would not have a way to identify and attribute the portion of those payments related to an LEJR episode. More information on setting bundled payment target prices for episodes under CJR is available in section III.C.4.b. of this final rule. Including the broadest feasible array of Medicare beneficiaries' admissions in the model would provide CMS with the most robust information about the effects of this model on expenditures and quality for beneficiaries of the widest variety of ages and comorbidities, and allow the participant hospitals the greatest opportunity to benefit financially from systematic episode care redesign because most Medicare beneficiaries undergoing an LEJR procedure will be included in the model and, therefore, subject to the policies we proposed.

We sought comment on our proposal on when to begin the CJR episode, as well as to identify the care included for beneficiaries.

The following is a summary of the comments received and our responses.

Comment: Most commenters agreed that the episode should begin with the hospital admission for the LEJR procedure. Some of these commenters noted that it would not be appropriate to include the period prior to the hospital admission as it could include unrelated care and introduce variability.

Several orthopedic surgeons commented that physician treatment and care management begin prior to surgery, with the physician continuing to manage care during surgery, following surgery, and throughout the entire PAC period. These commenters were concerned that beginning the episode with the hospital admission would result in beneficiaries choosing and initiating care plans designed with their treating physicians and later, when hospitalized, the beneficiaries would receive conflicting care plans and, ultimately, experience adverse outcomes.

Many commenters recommended starting the episode earlier than the hospital admission. Some commenters recommended starting the episode once the decision to pursue surgery is made, and some recommended specific timeframes that ranged between four to eight weeks prior to the surgery. Some commenters provided examples of presurgical services that they have found improve patient outcome and satisfaction, improve care quality, and reduce costs, such as comprehensive patient evaluations to assess a beneficiary's overall condition and chronic comorbid conditions; pre‐surgical counseling for non‐medical pain management; home safety reviews; post‐discharge planning; patient and caregiver education; weight loss programs; and physical therapy. Some commenters requested that CMS consider additional program rule waivers for the CJR model, beyond those specifically proposed, to facilitate the provision of various preoperative services and incentives that are not allowed or payable under current Medicare rules.

A few commenters were concerned that starting the episode with the hospital admission may lead to participants shifting costs to just prior to the start of the episode to receive payments for those services in addition to the bundle. To minimize gaming, they recommended starting the episode once the surgery has been elected and prior to the hospital admission, which is consistent with many private sector models.

Response: We appreciate the interest expressed by the commenters in starting comprehensive care coordination prior to the hospital admission, and we recognize that the beneficiary's care which ultimately leads to the LEJR surgery, including the physician-patient relationship, often begins long before the surgical procedure. We also appreciate concerns about providers unbundling services and shifting costs to just prior to the episode, between the time the surgery has been elected and the hospital admission. However, beginning the episode too far in advance of the LEJR surgery would make it difficult to avoid bundling unrelated items, and starting the episode prior to the hospital admission is more likely to encompass costs that vary widely among beneficiaries, which would make the episode more difficult to price appropriately.

We appreciate commenters' suggestions of pre-surgical services and Start Printed Page 73317programs that could support the continuum of care for CJR beneficiaries. However, identifying a specific set of related presurgical services to include in the episode, as recommended by some commenters, would be of little value in the model because many of the services that are typically necessary or the standard of care prior to surgery are often included in the IPPS payment under the three day payment window payment policies and are therefore already included in the CJR episode. We note that some of the related services suggested by commenters that are not typically included in the three-day payment window are intended to more broadly manage the clinical condition(s) that may have led to the LEJR, and as discussed previously in this section, the CJR model is designed to focus on the surgical procedure and the associated recovery. We also note that some of these suggested services would be applicable to a subset of CJR beneficiaries and, therefore, do not present a significant opportunity for improving efficiency and redesigning care management for the typical beneficiary receiving an LEJR.

We believe that using the date of admission as the start of the episode is appropriate as hospitals are unlikely to shift related services earlier than when is clinically indicated. With respect to expanding the waivers to presurgical services that are not currently covered or payable, we have finalized several waivers of Medicare program rules as discussed in section III.C.11. of this final rule specifically to assist participant hospitals in efficient and effective care coordination and care management for CJR beneficiaries, and we do not believe it would be consistent with the model design or otherwise necessary for the model test to implement waivers for the preoperative period. While we appreciate commenters' interest in providing additional presurgical services that may enhance care coordination and care management, the waivers of Medicare program rules are only available if the beneficiary is in the episode at the time a service under the waiver is furnished. We believe that allowing waivers in the preoperative period prior to the anchor hospitalization, based on an expectation that a beneficiary will be in a CJR Model episode, would not be appropriate as there is no guarantee that the beneficiary will actually initiate a CJR Model episode and qualify for services furnished under a waiver.

For purposes of the CJR model, we continue to believe that beginning the episode with the anchor hospitalization is most appropriate due to the clinical variability leading up to the joint replacement surgery and the challenge of distinguishing between related and unrelated services. We also believe that beginning the episode with the anchor hospitalization, and not prior to admission, would be easier to administer and provide more consistent episodes for testing the CJR Model. Therefore, we are finalizing our proposal to begin the episode with admission to an inpatient anchor hospitalization for MS-DRG 469 or MS-DRG 470 in accordance with the methodology described.

Comment: Commenters generally supported the proposed beneficiary inclusion criteria as reasonable and consistent with other programs. Some commenters suggested we exclude additional populations from CJR, namely beneficiaries with serious conditions or acute diseases, such as traumatic brain injury, spinal cord injuries, multiple-limb trauma, amputations, moderate to severe strokes, severe neuromuscular and musculoskeletal conditions, HIV infection, and cancer. A commenter recommended that we design a separate model to address the needs of patients with chronic conditions. A few commenters recommended excluding all patients on hospice.

Many commenters recommended that if we did not exclude high risk cases, we must develop more robust risk adjustment to account for socioeconomic, clinical, or other risk factors that are out of the hospital's control and impact patients' health and recovery. Some commenters were concerned that without accurate risk adjustment, hospitals will have an incentive to avoid higher-risk LEJR candidates. A commenter cited a study that found significant differences in Medicare spending per beneficiary during the 90-day episode based on various patient characteristics, such as type of LEJR surgery; emergency versus scheduled surgery; hip fractures versus degenerative conditions; patients age 85 or older; patients with multiple comorbidities, and patients who were dual eligible. The commenter asserted that robust risk adjustment based on the risk profile of each hospital's patients is essential for the CJR model because individual hospitals will not have enough enrollment to spread their risk. A few commenters recommended that at least the initial implementation of the Model should exclude vulnerable populations with complicated or intensive care needs until the CJR model demonstrates sufficient quality outcomes and has developed accurate risk adjustments and patient safeguards to ensure high-quality care for populations that the commenters believe could face serious care disadvantages in the CJR model.

Response: Many beneficiaries undergoing procedures that result in discharge from MS-DRG 469 and 470 have underlying conditions that may affect care throughout the episode or that may be influenced by the surgical procedure that initiates the episode. We believe it is important to include these beneficiaries in the model so that they can benefit from care coordination and management throughout the episode, and including the broadest feasible array of Medicare beneficiaries in the CJR model provides participant hospitals with greater incentive to redesign episode care. We also believe that patients in hospice would benefit from the improved comprehensive care coordination incentivized by the CJR model, and we refer readers to the related discussion in section III.B.2. of this final rule regarding our policy to include hospice claims in the episode.

We refer readers to section III.C.4.b. of this final rule for the final policy that will risk stratify the target prices based on the presence or absence of a hip fracture for CJR model beneficiaries. We believe that this risk stratification policy addresses many of the commenters' concerns that beneficiaries with serious conditions, acute diseases, and chronic conditions are likely to need more costly care throughout the CJR model episode that would have been inadequately paid under our proposal because these beneficiaries are those most likely to be present in the population receiving LEJR procedures emergently due to a hip fracture.

Comment: Several commenters recommended that CMS exclude beneficiaries who opted out of data sharing. These commenters asserted that it would be virtually impossible to manage risk and improve outcomes without claims data.

Response: As discussed in section III.E. of this final rule, we have decided not to finalize our proposal to allow beneficiaries the opportunity to decline having their data shared. We refer readers to section III.E. of this final rule for additional discussion of data sharing.

Comment: Some commenters suggested that CMS limit the CJR model to beneficiaries that live within a limited distance from participant hospitals so that the hospital would not be penalized for inadequately managing the PAC of medically complex patients from remote or distant locations.

Response: We expect that in some limited circumstances, participant Start Printed Page 73318hospitals will have limited ability to coordinate care. However, following the care coordination that takes place in the hospital, we believe that much of the subsequent coordination for PAC can be accomplished through telecommunications that do not require the patient to remain within geographic proximity of the hospital. Moreover, the design of the model does not preclude hospitals from coordinating care with local providers outside of their immediate referral area. We also note that we have finalized several waivers of Medicare program rules, as discussed in section III.C.11. of this final rule, to facilitate efficient and effective care coordination for beneficiaries in remote or distant locations outside the immediate community. Therefore, we will not exclude beneficiaries who are referred to participant hospitals from other areas.

Comment: A commenter requested CMS to consider including beneficiaries enrolled in MA plans in the model as they are likely to be healthier and their inclusion will help hospitals maintain costs within their targets. The commenter recognizes that the CJR payment methodology makes it difficult to identify and attribute payment related to the LEJR episode. However, the commenter asserts that participant hospitals in states with a high percentage of beneficiaries enrolled in MA plans are more likely to care for CJR patients with a higher than average risk profile, which could make it more difficult for a hospital to maintain costs within the target rate.

Response: We appreciate the commenter's interest in increasing the population of beneficiaries included in the CJR model, and we recognize that participant hospitals with higher risk CJR beneficiaries may find it more challenging to maintain actual aggregate episode payments within their target price. However, as discussed previously in this section, Medicare makes capitated payments for beneficiaries enrolled in MA plans, and providers do not submit complete claims data to CMS. Therefore, we are finalizing our proposal not to include beneficiaries enrolled in MA plans because we are unable to capture or appropriately attribute to the episode the related Medicare payments.

Comment: A couple of commenters requested that CMS exclude episodes where the LEJR surgery was furnished either by an opt-out physician, because the principal procedure is not paid by Medicare, or by a non-participating physician who does not accept assignment. They requested that if such episodes are to be included, CMS should establish policies under which participant hospitals can provide reconciliation payments to and receive alignment payments from opt-out physicians as well as non-participating physicians.

Response: Consistent with the BPCI policy, we do not believe it would be appropriate to exclude beneficiaries from the CJR model if a physician who opted out of Medicare pursuant to § 405.420 or a non-participating physician performs the LEJR surgery during the anchor hospitalization. We would expect that beneficiaries undergoing LEJR procedures, regardless of the Medicare participation or opt-out status of the operating surgeon, would have similar needs for care coordination and management throughout the episode period that extends 90 days post-hospital discharge, and we see no reason that hospitals should not have the same quality and cost performance responsibility for these episodes. We note that less than 15 percent of episode spending, on average, would be expected to be paid for physicians' services, with more than 80 percent of the episode payment made for inpatient hospital and PAC services. Thus, for a beneficiary who otherwise meets the CJR model's inclusion criteria, a CJR model episode would begin at the time of the beneficiary's admission for the anchor hospitalization, regardless of whether an opt-out physician or non-participating physician performs the LEJR surgery during that stay.

We refer readers to section III.C.3. of this final rule for discussion of the effect on reconciliation payments on services furnished by non-participating and opt-out physicians and to section III.C.10.a. of this final rule for discussion of issues related to gainsharing payments and alignment payments.

Summary of Final Decisions: After consideration of the public comments we received, we are finalizing our proposal to begin the episode with admission for an inpatient anchor hospitalization for MS-DRG 469 or MS-DRG 470 in accordance with the methodology described. We also are finalizing our proposal as to the criteria for beneficiary inclusion in the model as follows:

  • The beneficiary is enrolled in Medicare Part A and Part B throughout the duration of the episode.
  • The beneficiary's eligibility for Medicare is not on the basis of ESRD.
  • The beneficiary must not be enrolled in any managed care plan (for example, Medicare Advantage, Health Care Prepayment Plans, cost based health maintenance organizations).
  • The beneficiary must not be covered under a United Mine Workers of America health plan, which provides healthcare benefits for retired mine workers.
  • Medicare must be the primary payer.

The final policies for beginning an episode are set forth in § 510.210(a). The final policies for beneficiary inclusion are set forth in § 510.205.

b. Middle of the Episode

We proposed that once the episode begins for a beneficiary whose care is included, the episode continues until the end as described in the next section of this final rule, unless the episode is canceled because the beneficiary no longer meets the same inclusion criteria proposed for the beginning of the episode at any point during the episode. When an episode is canceled, we proposed that the services furnished to beneficiaries prior to and following the episode cancellation will continue to be paid by Medicare as usual but we will not calculate actual episode spending that would otherwise under CJR be reconciled against the target price for the beneficiary's care (see section III.C.6. of the proposed rule). As discussed in section III.C.11.a. of the proposed rule, if the beneficiary is in the episode at the time the service under the waiver is furnished, the waiver is available, even if the episode is later canceled.

In the proposed rule, we stated our belief that it would be appropriate to cancel the episode when a beneficiary's status changes during the episode such that they no longer meet the criteria for inclusion because the episode target price reflects full payment for the episode, yet we would not have full Medicare episode payment data for the beneficiary to reconcile against the target price.

In addition, we proposed that the following circumstances would also cancel the episode:

  • The beneficiary is readmitted to an acute care hospital during the episode and discharged under MS-DRG 469 or 470 (in this case, the first episode would be canceled and a new LEJR episode would begin for the beneficiary).
  • The beneficiary dies during the anchor hospitalization.
  • The beneficiary initiates an LEJR episode under BPCI Models 1, 2, 3 or 4.

In the case of beneficiary death during the anchor hospitalization, we stated our belief that it would be appropriate to cancel the episode as there are limited efficiencies that could be expected during the anchor hospitalization itself. In the case of beneficiary readmission during the first Start Printed Page 73319CJR episode for another LEJR (typically a planned staged second procedure), we stated our belief that it would not be appropriate to include two episodes in the model with some time periods overlapping, as that could result in attribution of the Medicare payment for 2 periods of PAC to a single procedure.

We sought comment on our proposals to cancel episodes once they have begun but prior to their end.

The following is a summary of the comments received and our responses.

Comment: Commenters were generally supportive of our proposals for canceling the episode, though many recommended additional circumstances for canceling the episode, such as adverse events which are beyond the hospital's control. Many commenters, including MedPAC, recommended that CMS cancel the episode if the beneficiary dies at any time during the episode, arguing that such cases could be extremely low or high cost and spending is, therefore, not typical. These commenters recommended that all episodes that end in patient death should be excluded from the calculations of the target price and reconciliation amounts, not just those episodes where patients die during the initial hospitalization as CMS proposed, as this type of episode of care could skew the data. Given that hospitals are held financially responsible for the entire 90-day episode, a few commenters suggested excluding all episodes with death for consistency and administrative simplicity. A commenter observed that a deceased beneficiary no longer meets all of the beneficiary inclusion criteria, and on that basis recommended that CMS cancel the episode when the patient dies. A commenter suggested also canceling episodes for any beneficiaries that die during the 30 day post-episode monitoring period. Some commenters suggested that other circumstances should cancel an episode, such as a beneficiary geographic move, change in beneficiary residence from a home to a facility, and loss of the beneficiary to follow up care.

Response: While beneficiary deaths during LEJR episodes are uncommon, we expect them to vary unpredictably across hospitals and, therefore, we agree that it would be appropriate to cancel episodes under these circumstances. We also agree that canceling all episodes during which a beneficiary dies is consistent with the otherwise applicable episode duration as the episode would not extend to 90 days hospital post-discharge. However, we would include episodes where the patient dies during the 30 days post-episode as this would not affect the variability of episode spending, and it would be appropriate to monitor for beneficiary death during the immediate post-episode period.

We expect some limited circumstances where participant hospitals will have limited ability to coordinate care. However, we believe that participant hospitals will be incentivized to seek creative solutions that do not rely on in-person services, and we are finalizing our proposal that all other beneficiary episodes would remain in the CJR model, regardless of where the beneficiary is located. Payment for beneficiaries in these circumstances will be reflected in the target prices based on historical utilization.

Comment: Commenters urged CMS to hold beneficiaries and providers financially harmless for care received as part of a CJR episode if the episode is later canceled. A few commenters supported the continued application of Medicare program waivers if an episode is canceled when a beneficiary's status changes, and a few commenters were unclear if waivers apply to beneficiaries who are retrospectively identified as ineligible for CJR program waivers due to changes in coverage status.

Response: As discussed previously in this section, we proposed that if the beneficiary is in the episode at the time the service under the program rule waiver is furnished, the waiver is available, even if the episode is later canceled. If the beneficiary is not in the episode at the time the service under the waiver is furnished, financial liability for these services would be determined in accordance with the policies outlined in the Medicare Claims Processing Manual (Pub. L. 100-04), Chapter 30. As we gain experience with CJR, we may revisit this issue in future rulemaking. We refer readers to section III.C.11. of this final rule for additional discussion and our finalized policy to apply waivers of programs rules if the beneficiary is in the episode at the time the service under the waiver is furnished, even if the episode is later canceled.

Comment: A commenter was concerned that initiation of a BPCI episode would cancel a CJR episode, when the CJR episode begins first. The commenter also requested clarification whether a BPCI episode for a different clinical condition, such as cardiac procedures, would cancel a CJR LEJR episode.

Response: We proposed and are finalizing our policy that a CJR episode would be canceled when a beneficiary initiates an LEJR episode under BPCI Models 1, 2, 3, or 4. A CJR beneficiary initiating a different clinical episode under BPCI Models 1, 2, 3, or 4 would remain in a CJR episode. We refer readers to section III.C.7.b. of this final rule for additional discussion of CJR beneficiary overlap with BPCI episodes.

Summary of Final Decisions: After consideration of the public comments we received, we are finalizing our proposal to cancel episodes once they have begun but prior to their end if the beneficiary no longer meets the same inclusion criteria proposed for the beginning of the episode at any point during the episode. We also are finalizing our proposal that the following circumstances would also cancel an episode:

  • The beneficiary is readmitted to a participant hospital during the episode and discharged under MS-DRG 469 or 470.
  • The beneficiary initiates an LEJR episode under BPCI Models 1, 2, 3 or 4.

We are modifying our proposal for canceling an episode when a beneficiary dies during an anchor hospitalization. Under our final policy, the following circumstance would also cancel an episode:

  • The beneficiary dies at any time during the episode.

The final policies for cancellation of an episode are set forth in § 510.210(b). We note that § 510.210(b)(4) has been revised to state that an episode is canceled if the beneficiary dies during the episode.

c. End of the Episode

LEJR procedures are typically major inpatient surgical procedures with significant associated morbidity and a prolonged recovery period that often is marked by significant PAC needs, potential complications of surgery, and more intense management of chronic conditions that may be destabilized by the surgery. In light of the course of recovery from LEJRs for Medicare beneficiaries, we proposed that an episode in the CJR model end 90 days after discharge from the acute care hospital in which the anchor hospitalization (for MS-DRG 469 or 470) took place. Hereinafter, we refer to the proposed CJR model episode duration as the “90-day post-discharge” episode. To the extent that a Medicare payment for included services spans a period of care that extends beyond the episode duration, we proposed that these payments would be prorated so that only the portion attributable to care during the fixed duration of the episode is attributed to the episode spending.

We noted that for the vast majority of beneficiaries undergoing a hip or knee joint replacement, a 90-day post-Start Printed Page 73320discharge episode duration encompasses the full transition from acute care and PAC to recovery and return to activities. We stated our belief that the 90-day post-discharge episode duration encourages acute care hospitals, physicians, and PAC providers to promote coordinated, quality care as the patient transitions from the inpatient to outpatient settings and the community.

In proposing the 90-day post-discharge duration for LEJR episodes in CJR, we took into consideration the literature regarding the clinical experiences of patients who have undergone THA or TKA procedures. In 2007-2008, the 30-day all-cause readmission rate for primary THA among Medicare beneficiaries was 8.5 percent, while the 90-day all-cause readmission rate was 11.9 percent, indicating that while the rate of readmission begins to taper after 30 days, readmissions continue to accrue throughout this 90 day window.[6] In single center studies, Schairer et al found unplanned 30-day hospital readmission rates were 3.5 percent and 3.4 percent and unplanned 90-day hospital admission rates were 4.5 percent and 6 percent for primary THA and TKA, respectively, demonstrating that the risk of readmission remains significantly elevated from 30 through 90 days post-hospital discharge.[7] [8] Further exploring the reasons for unplanned admission for TKAs within 90 days of a knee replacement procedure, Schairer et al found that 75 percent were caused by surgical causes such as arthrofibrosis and surgical site infection. Additional information on the common reasons for hospital readmission following TKA or THA can be obtained from The American College of Surgeons National Surgical Quality Improvement Program.[9] These data identified the top 10 reasons for readmission within 30 days of a hip or knee arthroplasty:

  • Surgical site infections (18.8 percent).
  • Prosthesis issues (7.5 percent).
  • Venous thromboembolism (6.3 percent).
  • Bleeding (6.3 percent).
  • Orthopedic related (5.1 percent).
  • Pulmonary (3.2 percent).
  • Cardiac (2.4 percent).
  • CNS or CVA (2.4 percent).
  • Ileus or Obstruction (2.3 percent).
  • Sepsis (2.1 percent).

In addition, the authors concluded that “readmissions after surgery were associated with new post-discharge complications related to the procedure and not exacerbation of prior index hospitalization complications, suggesting that readmissions after surgery are a measure of post-discharge complications.” Finally, with regard to the potential for readmission for joint replacement revision within a 90-day post-discharge episode, in a twelve-year study on Medicare patients conducted by Katz, et al., the risk of revision after THA remained elevated at approximately 2 percent per year for the first eighteen months and then 1 percent per year for the remainder of the follow-up period.[10] This study suggests that a longer episode, as opposed to a shorter episode, is more likely to simulate the increased risk of revision LEJR patients face.

In order to address the complication rates associated with elective primary total hip or knee arthroplasty, we developed an administrative claims-based measure (for a detailed description of the measure see section III.D. of the proposed rule). During the development of the Hospital-level Risk-Standardized Complication Rate (RSCR) following elective primary THA or TKA or both, complications of elective primary total hip or knee replacement were identified to occur within specific timeframes.[11] For example, analyses done during the development of the measure as well as Technical Expert Panel opinion found that—(1) Mechanical complications and periprosthetic joint infection/wound infection are still attributable to the procedure for the 90 days following admission for surgery; (2) death, surgical site bleeding, and pulmonary embolism are still likely attributable to the hospital performing the procedure for up to 30 days; and (3) medical complications of acute myocardial infarction (AMI), pneumonia, and sepsis/septicemia/shock are more likely to be attributable to the procedure for up to 7 days.

Other factors further supporting a 90-day post-discharge episode duration are the elevated risk of readmission throughout this time period, as well as the fact that treatment for pneumonia is considered by American Thoracic Society guidelines to be “health care-associated” if it occurs up to 90 days following an acute care hospitalization of at least 2 days.[12] According to the American Academy of Orthopedic Surgeons, patients undergoing total hip replacement should be able to resume most normal light activities of daily living within 3 to 6 weeks following surgery.[13] In a small randomized controlled trial of two approaches to hip arthroplasty, average time to ambulation without any assistive device was 22-28 days.[14] According to a 2011 systematic review of studies evaluating physical functioning following THA, patients have recovered to about 80 percent of the levels of controls by 8 months after surgery.[15]

We also refer readers to a study by the Assistant Secretary for Planning and Evaluation (ASPE) in the U.S. Department of Health and Human Services that assessed the mean payments for acute care, PAC, and physician services grouped in the MS-DRG 470.[16] In this study, CMS payment for services following an MS-DRG 470 hospitalization were concentrated within the first 30 days following discharge, with plateauing of payments between 60- or 90-days post-discharge.

Start Printed Page 73321

Finally, payment and length of stay analyses found the average length of stay in PAC during a 90-day post-discharge episode for MS-DRG 470 to be 47.3 days, indicating that a longer period post-discharge of 90 days is reasonable as a proposal to end the episode of care.[17] We noted that these analyses did not include any time between hospital discharge and the start of PAC.

Table 6—Cost and Length of Stay Statistics for MS-DRG 470 for Various Episode Durations

Statistics for DRG 470 (2006 data)30-day episode60-day episode90-day episode
Mean Medicare spending per hospital discharge (acute+PAC+physician)$18,838$20,343$21,125
Mean payment for anchor hospitalization$10,463$10,463$10,463
Mean payment for PAC$6,835$8,339$9,122
Mean payment for physicians (during anchor hospitalization)$1,540$1,540$1,540
Mean payment for readmission (includes all PAC users, even if no readmission occurs during the episode)$550$929$1,242
Mean length of stay (LOS) for PAC25.5 days39.6 days47.3 days
Note: Data are per PAC user (88% of beneficiaries hospitalized under MS-DRG 470 are discharged to PAC). PAC users are defined as beneficiaries discharged to SNF, IRF, or LTCH within 5 days of discharge from the index acute hospitalization, or discharged to HHA or hospital outpatient therapy within 14 days of discharge from the index acute hospitalization. Mean LOS for PAC does not include any gap between hospital discharge date and start of PAC.

Other tests of bundled payment models for hip and knee replacement have used 90-day post-discharge episodes.[18] We also noted that despite BPCI Model 2 allowing participants a choice between 30-, 60-, or 90-day post-discharge episodes, over 86 percent of participants have chosen the 90-day post-discharge episode duration for the Start Printed Page 73322LEJR episode. Furthermore, a 90-day post-discharge episode duration aligns with the 90-day global period included in the Medicare Physician Fee Schedule (MPFS) payment for the surgical procedure.

We also considered proposing a 60-day post-discharge episode duration, but the full transition of care following LEJR would exceed this window for some beneficiaries, especially those who are discharged to an institutional PAC provider initially and then transition to home health or outpatient therapy services for continued rehabilitation. According to a report from ASPE on Medicare beneficiaries receiving PAC following major joint replacement in 2006, 13 percent first receive SNF services and then receive HHA services—with a total mean episode duration of 56.8 days.[19] An additional 9.2 percent receive HHA services first and then receive outpatient therapy services—with a total mean episode duration of 78.7 days. Finally, 6.7 percent receive IRF services first and then HHA services (total mean length of stay 55.3 days), and 4.8 percent receive SNF services first and then outpatient therapy services (total mean length of stay 71.5 days). The remainder only receives one type of PAC.

Therefore, in order to be inclusive of most possible durations of recovery, and services furnished to reach recovery, we proposed the 90-day post-discharge episode duration for CJR. We stated our belief that beneficiaries will benefit from aggressive management and care coordination throughout this episode duration, and hospitals will have opportunities under CJR to achieve efficiencies from care redesign during the 90-day post-discharge episode period.

We sought comment on our proposal to end the episode 90 days after the date of discharge from the anchor hospitalization, as well as on the alternative we considered of ending the CJR episode 60 days after the date of discharge.

The following is a summary of the comments received and our responses.

Comment: Most commenters supported the 90-day post-discharge episode duration. Many of these commenters provided rationales for supporting the 90-day duration (as compared to 60 days or other shorter durations), such as: It is a clinically appropriate length to manage an LEJR to recovery; it creates strong incentives for collaboration for multiple providers across the care continuum that improves care transitions and care coordination; it will promote better long-term results; it aligns with quality measures; and it is the most popular timeframe selected for BPCI Model 2. Some of these commenters asserted that a shorter duration is not sufficiently long to capture the vast majority of issues arising directly from LEJR procedures and could put beneficiary care at risk by encouraging providers to reduce utilization inappropriately or shift utilization outside of an episode.

A few commenters supported a 90-day episode duration, but recommended that we revise the 90-day post-discharge episode duration to begin from the date of surgery instead of discharge, thereby aligning the episode with the MPFS global surgical period and billing policies. A commenter who appeared to believe that CMS proposed to begin the CJR episode immediately after discharge from the anchor hospitalization and extend the episode 90 days post-hospital discharge, rather than upon admission for the anchor hospitalization as CMS actually proposed, asserted that beginning the episode after hospital discharge would make it difficult to understand and account for patient acuity changes within the episode in the post-discharge period as the hospital length-of-stay is related to the PAC acuity of the beneficiary following hospital discharge, especially if the beneficiary has comorbidities. In other words, the commenter believed that beneficiaries with comorbidities would be more likely to have longer anchor hospitalizations and associated higher intensity of PAC services, yet CMS would not understand these relationships if the anchor hospitalization was not included in the episode.

Several commenters supported a 60-day post-discharge episode duration because LEJR patients are nearly fully recovered within 60 days. Some commenters asserted that PAC services associated with LEJR rarely occur after 60 days post-discharge; some commenters cited data that the majority of services for patients with LEJR surgery occur within two months of discharge with only a 6.2 percent change in the total cost of an episode between a 60-day episode and a 90-day episode. Some of these commenters asserted that a 60-day episode would be sufficient to evaluate quality and cost, and a longer duration would increase the financial risk for hospitals without providing significant value to CMS. Some commenters asserted that a 90-day duration increases the risk that unrelated random events that occur well after surgery will disadvantage the hospitals by unfairly impacting participants' performance.

Some commenters recommended a hybrid approach, with every service within the first 30 days post-discharge assumed to be related unless specifically excluded, and services in days 31-90 included only if they meet specified criteria for relatedness.

Some commenters recommended that the episode end prior to 60 days post-discharge. A commenter recommended an episode length of 45 to 60 days, asserting that hospital admissions past the 45 to 60 day window would be for chronic medical admissions that are unrelated to the LEJR procedure. A few commenters recommended that we limit the episode to 30 days citing various rationales, such as: A SNF stay must commence within 30 days of a hospitalization; 30 days better aligns with other quality improvement initiatives such as readmissions; analyses by Medicare Payment Advisory Commission (MedPAC) and the Congressional Budget Office that found that the majority of a bundled payment's episode costs are incurred during the first 30 days; and hospitals may find it difficult to manage follow-up care after 30 days if patients have more than one residence. Several commenters asserted that multiple factors can exacerbate comorbidities in the period beyond 30 days post-operatively, and a model of longer duration that broadly defines related services could result in participant hospitals being more cautious about selecting patients for LEJR and complex patients being discouraged from seeking LEJR procedures in a participant hospital. A few of these commenters noted that Tennessee and Arkansas only include 30 days post-discharge for unrelated chronic conditions in their bundled payment episodes. A commenter shared its experience that, while nearly all patients are diligent about keeping 14-day and 30-day post-operative appointments, those with good outcomes are less likely to return for appointments at 90 days and beyond, resulting in potentially skewed outcomes as patients with complications are much more likely to keep a follow-up appointment at 90 days.

Some commenters recommended giving participant hospitals the flexibility to define the episode duration, either as a duration for all of a participant hospital's LEJR episodes, or to choose a duration based on a patient's clinical condition and comorbidities. A couple of commenters Start Printed Page 73323recommended that if CMS offers participants the option to choose the duration, consistent with BPCI, CMS should lower the discount percentage for those willing to take the longer episodes. A commenter disagreed with CMS' cited rationale of the operational simplicity of a single duration for all LEJR episodes by noting that BPCI Model 2 operationalized a variety of different bundles and gave participants the choice of three durations for 48 different clinical episodes.

Other commenters suggested even longer episode durations. A commenter recommended increasing the episode duration to 150 days post-discharge to promote better long-term results and reduce the likelihood of delaying care beyond the end of the episode, specifically urging CMS to adopt a longer episode period for certain clinically-complex subpopulations with predictably longer recovery timeframes. For outcome and quality measurement purposes, some commenters recommended that participant hospitals be held accountable for a longer period, with suggestions of six months, a year, and even two to three years. A commenter recommended increasing the episode duration to two years to better manage the improvements for the entirety of the treatment. A commenter recommended increasing the episode duration to five years to account for the late effects of sub-optimal implant selection.

Response: We appreciate the support of many commenters for the proposed 90-day post-hospital discharge CJR model episode duration. We agree with the commenters that this relatively long episode duration should capture the great majority of health care services that are related to the episode, as well as the beneficiary's return to function and short- and medium-term health outcomes. We believe this episode duration provides participant hospitals with a substantial period of time in which to work to improve the quality and efficiency of LEJR episode performance for beneficiaries who undergo LEJR surgery at their hospital. We have substantial BPCI Model 2 experience with Awardees engaged in testing 90-day LEJR episodes, and note that the vast majority of Awardees have selected the 90-day episode duration, compared to the 30-day and 60-day alternative durations that are available in the model. Our goal is to incentivize efficient high quality care that returns people to the community, and we believe that a 90-day post-discharge duration reflects a full continuum of clinical services and transition of care following LEJR procedures for the average beneficiary, at which time the patient's functional recovery is relatively complete and the patient is able to resume most normal activities of daily living.

Due to the concentration of Medicare spending in the earlier part of the episode, we also believe that a 90-day episode duration only nominally increases the hospital's financial risk when compared to 30 or 60 days. While we understand that uncommon events during the 90-day episode may occur for an individual beneficiary, resulting in an unanticipated or unavoidable need for costly health care services, we believe that our episode definition that excludes unrelated items and services and our payment policies, namely the adjustment for high payment episodes and stop-loss policies discussed in sections III.C.3. and III.C.8. of this final rule, provide sufficient protections for participant hospitals from undue financial responsibility for the care of unrelated clinical conditions as well as for unusual circumstances. We also believe that shorter episode durations may incur a higher clinical risk for beneficiaries if participants delay services beyond the episode, and the risk to beneficiaries of this response by providers to episode payment that can be minimized by the longer 90-day episode duration that we proposed. We refer readers to sections III.F.3. and 5. of this final rule for discussion of our plans to monitor for access to care and delayed care.

In response to those commenters requesting a hybrid approach where CMS would include a broader set of related services in the 30 days following discharge from the anchor hospitalization and a more limited set of related services from days 31 to 90 because of the closer clinical link of a beneficiary's clinical conditions in the first 30 days to the events during the anchor hospitalization itself, we emphasize that the CJR model is an episode payment model where many Medicare beneficiaries who receive PAC services as part of their post-operative recovery from surgery will also have underlying health conditions that may be affected by the surgery itself and care throughout the recovery period and that require attentive, flexible management if good health outcomes are to be achieved. Because PAC services are designed to be comprehensive in nature, we believe that the same Part A and Part B services should be included throughout the episode duration because PAC providers should broadly address the beneficiary's health care needs in high quality, efficient episodes, even though the anchor hospitalization itself may be more remote from the beneficiary's health needs as the time from hospital-discharge increases. As discussed in section III.A.3. of this final rule, we have identified hospitals as the financially responsible organization for the episode, although episode quality and cost performance will clearly be related in part to the quality and efficiency of care furnished by other providers and suppliers treating the beneficiary throughout the episode. We expect that participant hospitals will develop the care pathways and partnerships with other providers and suppliers necessary for the hospital to be successful in this responsibility, and this model provides a variety of tools that should be helpful to participant hospitals, such as waivers of Medicare program rules, the opportunity to engage in certain financial arrangements, and the ability to offer certain beneficiary incentives (as discussed in sections III.C.11. and III.C.10. of this final rule, respectively).

We appreciate the interest of some commenters in significantly longer episodes than the 90 days post-hospital discharge period we proposed, in order to include the longer recovery period that some beneficiaries may require as well as to account for longer term health outcomes, because the timing or frequency of joint replacement revisions may be related to implant selection, surgical technique, or other aspects of the primary joint replacement procedure. However, as previously noted, we believe that a 90-day post-discharge duration reflects a full continuum of clinical services and transition of care following LEJR procedures for the average CJR beneficiary, and we do not believe it would be an appropriate test of the model to extend the CJR episode duration beyond 90 days post-hospital discharge to reflect the longer recovery needed by some beneficiaries. Moreover, as noted previously in this section, the CJR model focuses on the surgical procedure and the associated recovery, and at this time, we are not testing a model of longer term outcomes. Therefore, we are not going to incorporate a longer time period in the episode, and will not include periods beyond then, other than to monitor the 30-day post-episode period. The 30-day post-episode period is discussed in section III.C.8.d. of this final rule, where we describe the CJR model policy that holds participant hospitals financially responsible for significantly increased Medicare Parts A and B spending in the 30 days immediately following the end of the episode. We note that the Start Printed Page 73324evaluation described in section IV. of this final rule will focus on a variety of key topics including potential unintended consequences such as cost shifting beyond the CJR model episode period and stinting on medically necessary and appropriate care. As such, CMS anticipates the examination of claims submitted beyond the 90-day episode will be incorporated in the evaluation strategy. Finally, we maintain that allowing for multiple durations would be administratively complex for a model of this scope as it would be akin to implementing multiple models concurrently, each with its own customized payment calculations, risk adjustments, and other elements. We do not believe a variable approach such as is used in BPCI, which is a voluntary model, is appropriate for this large test of LEJR episode payment for all IPPS hospitals in the selected MSAs, as it would greatly increase the administrative complexity of the CJR model. We also believe that a standard duration for all episodes is important for this test of LEJR episode payment in providing us with a larger sample of episodes of the same duration from which we can learn.

Regarding the request to align the CJR model episode duration with the MPFS by beginning the 90-day duration on the date of surgery, rather than on the date of discharge from the hospital, we do not agree with this suggestion. We believe that the 90-day global surgical period for LEJR procedures under the MPFS lends support for an episode duration under the CJR model that is similar, because beneficiaries have a significant post-operative recovery period throughout which close care coordination and management among treating providers is important to beneficiary return to function. The MPFS global payment policy sets an expectation that the operating surgeon plays a significant role in caring for beneficiaries in the typical case that extends up to 90 days following surgery. However, using this same 90-day accounting methodology under the CJR episode would lead to model episodes including variable post-discharge lengths because the duration of the anchor hospitalization, which can vary substantially, would count toward the 90 days. We are interested in testing under the CJR model an episode duration that is most likely to cover the time for the beneficiary's full recovery and return to the community so we believe that including a standard length of 90 days post-hospital discharge is the best way to ensure that each CJR beneficiary's episode includes the same length of post-hospital discharge recovery in the episode. We do not believe the minor 90-day definitional differences between this model and the MPFS global billing policies for LEJR procedures should create significant problems for physicians collaborating with participant hospitals in the episode care of CJR model beneficiaries.

In response to the commenter concerned that starting the bundle after hospital discharge would make it difficult to account for patient acuity changes post-discharge under the CJR model, we want to emphasize that the CJR model episode actually begins on the day of admission for the anchor hospitalization and extends 90 days post-hospital discharge, with the day of hospital discharge counting as the first day in the 90-day post-hospital discharge period. Thus, the episode includes the full anchor hospital length-of-stay that may affect changes in patient acuity in the post-discharge period. We note that according to this episode duration definition, episodes for individual beneficiaries will have a variable total length that depends on the length of the anchor hospitalization. For example, the average length-of-stay for MS-DRG 470 is 3 days, so the average CJR model episode length for an individual beneficiary would be 92 days. The average length-of-stay for MS-DRG 569 is 6 days, so the average CJR model episode length for an individual beneficiary would be 95 days. Despite their variable total length, all CJR model episodes will include the complete anchor hospitalization and 90 days post-hospital discharge and, therefore, will include all related items and services furnished to the beneficiary throughout the episode, including those provided to address beneficiary acuity changes during the hospitalization and post-discharge period.

Summary of Final Decisions: After consideration of the public comments we received, we are finalizing our proposal to end the episode 90 days after discharge from the anchor hospitalization. We are revising the definition of Episode of care to clarify that the day of discharge itself counts as the first day of the post-discharge period and adding the same clarification to § 510.210(a)

The final definitions policies for ending an episode are set forth in § 510.2 and

C. Methodology for Setting Episode Prices and Paying Model Participants Under the CJR Model

1. Background

As described in section II.B. of the proposed rule, we proposed to use the CJR episode payment model to incentivize participant hospitals to work with other health care providers and suppliers to improve quality of care for Medicare beneficiaries undergoing LEJR procedures and post-operative recovery, while enhancing the efficiency with which that care is provided. We proposed to apply this incentive by paying participant hospitals or holding them responsible for repaying Medicare based on their CJR episode quality and Medicare expenditure performance. The following sections describe our final decisions for the—

  • Performance years covered by the model, the retrospective methodology that will be applied, and the application of two-sided risk beginning in the second year of the model;
  • Adjustments that will be made to payments included in the episode;
  • Episode price setting methodology;
  • Use of quality performance in the payment methodology;
  • Process for reconciliation;
  • Adjustments for overlaps with other CMMI models and CMS programs;
  • Limits and adjustments on hospitals' financial responsibility;
  • Appeal procedures for reconciliation;
  • Financial arrangements and beneficiary incentives; and
  • Waivers of Medicare program rules.

2. Performance Years, Retrospective Episode Payment, and Two-Sided Risk Model

a. Performance Period

We proposed that the CJR model would have 5 performance years. The performance years would align with calendar years, beginning January 1, 2016. Table 7 includes details on which episodes would be included in each of the 5 performance years.

Start Printed Page 73325

Table 7—Proposed Performance Years for CJR Model

Performance yearCalendar yearEpisodes included in performance year
12016Episodes that start on or after January 1, 2016, and end on or before December 31, 2016.
22017Episodes that end between January 1, 2017, and December 31, 2017, inclusive.
32018Episodes that end between January 1, 2018, and December 31, 2018, inclusive.
42019Episodes that end between January 1, 2019, and December 31, 2019, inclusive.
52020Episodes that end between January 1, 2020, and December 31, 2020, inclusive.

Under our proposal, all episodes tested in this model would have begun on or after January 1, 2016 and ended on or before December 31, 2020. We noted that this definition would result in performance year 1 being shorter than the later performance years in terms of the length of time over which an anchor hospitalization could occur under the model. We also noted that some episodes that began in a given calendar year may be captured in the following performance year due to the episodes ending after December 31st (for example, episode beginning in December 2016 and ending in March 2017 would be part of performance year 2). We stated our belief that 5 years would be sufficient time to test the CJR model and gather sufficient data to evaluate whether it improves the efficiency and quality of care for an LEJR episode of care. Further, having fewer than 5 performance years may not provide sufficient time or data for evaluation. The 5-year performance period is consistent with the performance period used for other CMMI models (for example, the Pioneer Accountable Care Organization (ACO) Model).

The following is a summary of the comments received and our responses.

Comment: Several commenters supported our proposal for a 5-year performance period as well as our proposed start date of January 1, 2016. However, a substantial number of commenters expressed concerns over the proposed start date and requested that we delay implementation of the model. Most of these commenters expressed concerns about the ability of participants to successfully participate in the model, given the proposed timeframes. Commenters noted that participants would need additional time for activities such as developing a new infrastructure with respect to provider networks, which would include identifying and establishing contracts with collaborators as well as determining appropriate incentives and gainsharing structures; identifying and developing new care pathways and performance metrics; and developing as well as modifying accounting and IT systems. In particular, a number of commenters expressed concern with the proposed start date in light of the requirement that hospitals begin to assume risk in the second year of the model, which is discussed further later in this section. Moreover, given variation in hospital preparedness, these kinds of issues could be particularly acute for certain kinds of hospitals, for example, smaller hospitals or those with more limited resources. Also, as discussed in section III.E of this final rule, commenters noted that their ability to implement the previously stated changes would be impeded by not having received baseline and episode-level data from CMS until after the proposed start date. Commenters indicated that these data would be essential to identifying opportunities and strategies for quality and efficiency improvement, and that the model should be delayed until after they have had a chance to review and understand their own episode data.

We also received comments suggesting that implementation of the model is premature and that it should be delayed until certain actions or events have occurred, for example, until certain quality measures have been developed, data required under the Improving Medicare Post Acute Care Transformation Act of 2014 (Pub. L. 113-185, enacted October 6, 2014) (IMPACT Act) have been collected or analyzed, or CMS has considered the results of other bundled payment models such as BPCI. For example, several commenters requested a phased implementation of the CJR model, due to the limited evaluation results that have been publicly released to date for BPCI, and to allow for testing and monitoring of the CJR model prior to full implementation. Another commenter asserted that a phased-in approach to implementing CJR is appropriate, given that while episode-based payment models have shown potential to reduce cost, rigorous studies and evaluation data on episode-based payment models are limited. Some commenters expressed the view that CMS' timeline ignores multiple competing mandates that hospitals and other providers have, for example, ICD-10-CM implementation as well as EHR Meaningful Use and other quality-related programs.

In addition, we received a comment urging a delayed start date due to concerns on how requirements with respect to the civil monetary penalty (CMP) law (sections 1128A(a)(5), (b)(1) and (b)(2) of the Act), the Federal Anti-kickback statute (section 1128B(b)(1) and (2) of the Act), or the physician self-referral prohibition (section 1877 of the Act) would apply under the model. For example, a commenter noted that the proposed rule offered insufficient protection from certain statutory and regulatory risks associated with developing coordinated care arrangements among providers and that significant ambiguity and challenges existed with respect to compliance with these requirements.

Commenters also stated that in contrast to our proposed start date for the CJR model, CMS allowed voluntary BPCI participants, who were more likely to be well positioned to participate in an episode-based payment model, at least one year to consider their episode data, yet many of them likely found the program and timing demands challenging. Further, mandating the program, especially for unprepared participants, could result in even greater challenges, and increase the chance of failure and disruption of health care services for Medicare beneficiaries.

Some commenters offered examples of how, in their view, implementing the model by the proposed start date could result in unintended consequences such as reduced access for beneficiaries or lower care quality. For example, commenters suggested that the proposed timeframe could cause hospitals to make care redesign choices that reduced access for beneficiaries or certain kinds of beneficiaries such as those who posed greater risk or that care quality could be compromised because participants would have had insufficient time to implement new care practices.

Given these concerns, commenters generally requested that we delay the start date by a specific period of time, for example, by three months, six months, nine months or a year, with most commenters requesting a delay of nine months to a year. Some Start Printed Page 73326commenters recommended delay periods of two years or more. In some cases, commenters tied their proposed delay period to an event, for example, some period of time subsequent to having received baseline and episode-level data from CMS. Some commenters requested that only the mandatory aspect of the model be delayed, allowing providers willing to participate the opportunity to do so or, in the event of a delayed start date, providers be permitted to voluntarily opt-in to the model prior to the date of implementation. As such, providers who had begun to prepare for the model could begin to generate cost savings while driving improvements in quality and patient experience for LEJR patients.

Response: We appreciate the comments we received in support of our proposed performance period and start date. We also appreciate and are persuaded by comments expressing concerns that our proposed start date does not provide sufficient time for participants to implement the kinds of changes needed to successfully participate in the model, particularly given that baseline data would not be available until after our proposed start date of January 1, 2016. Accordingly, this final rule will delay the start date of the model to April 1, 2016. Also, as indicated in section III.E.4 of this rule, we intend to make participating hospitals' baseline data available upon request in advance of the April 1, 2016 start date, which will allow participants the opportunity to assess their baseline data as they consider changes to their care practices in advance of the model's start date. Also, as discussed in section III.C.8. of this final rule, we are reducing the potential risk to participants in Year 2 by lowering the stop-loss limit from 10 percent to 5 percent (and from 20 percent to 10 percent in Year 3). We believe that these changes will both facilitate participants' abilities to be successful under this model and allow for a more gradual transition to full financial responsibility under the model.

Table 8 includes details on which episodes would be included in each of the 5 performance years under this delay.

Table 8—Performance Years for CJR Model

Performance yearCalendar yearEpisodes included in performance year
12016Episodes that start on or after April 1, 2016, and end on or before December 31, 2016.
22017Episodes that end between January 1, 2017, and December 31, 2017, inclusive.
32018Episodes that end between January 1, 2018, and December 31, 2018, inclusive.
42019Episodes that end between January 1, 2019, and December 31, 2019, inclusive.
52020Episodes that end between January 1, 2020, and December 31, 2020, inclusive.

Under this revised schedule, all episodes tested in this model will have begun on or after April 1, 2016 and ended on or before December 31, 2020. Additional discussion on how this revised performance year schedule affects the use of quality measures for the model and the timeline for the reconciliation process is included in sections III.C.5. and III.C.6. of this final rule.

We do not agree that a longer delay is needed. Hospital participants will not be financially responsible for repayment to Medicare until the second performance year of the model. In addition, as discussed in section III.C.8. of this final rule, we have further limited financial risk to hospitals in performance years 2 and 3 by lowering stop-loss limits; specifically, from 10 percent to 5 percent in Year 2, and from 20 percent to 10 percent in Year 3. Finally, while we note that commenters are correct that voluntary BPCI participants received claims data prior to taking on risk under the BPCI model, and in some cases had more than a year to prepare for participation in BPCI, we believe that providing claims data to CJR participants in early 2016 and beginning the model April 1, 2016 is appropriate for several reasons. First, we note that under BPCI, voluntary participants in Phase I had the option of receiving claims data for multiple episodes, up to the 48 clinical episodes included in the BPCI initiative. The CJR model will only include one type of episode, and as such we believe it is reasonable for hospitals to begin to analyze data and identify care patterns and opportunities for care redesign for this episode in our stated implementation timeline. We also note that due to the gradual implementation of downside risk, we expect that hospitals would spend the first performance year of the model analyzing data, identifying care pathways, forming clinical and financial relationships with other providers and suppliers, and assessing opportunities for savings under the model, utilizing the quarterly claims data we provide to them. This is similar to the approach we took to allow hospitals to participate in Phase I of BPCI prior to entering Phase II (the risk-bearing phase). As noted in this section, participant hospitals would also be eligible to receive reconciliation payments for performance year 1 if actual spending is below the target price. We believe that our implementation timeline is reasonable, given the financial opportunity for hospitals to earn reconciliation payments for performance year 1 and the gradual transition to financial responsibility.

We are also not persuaded by commenters that implementation of the model is premature or that it should be delayed until results for BPCI or other episode-based payment models are available. While we anticipate that the BPCI model will offer valuable information that should assist CMS in developing bundling payment models, the CJR model will offer additional insights that are not available under the BPCI model; in particular, insights with respect to bundling payment models on a mandatory rather than voluntary basis. Thus, we will be able to observe how a bundling payment model might work with participants that would otherwise not participate in such a model. As such, we expect the results from this model should produce data that are more broadly representative than what might be achieved under a voluntary model. Also, this model tests a different target pricing approach than the one used in BPCI. BPCI uses a purely participant-specific pricing approach, rewarding participants for improving based on their historical performance. While this may incentivize historically less efficient participants to improve, there may not be as much incentive for already efficient participants. The regional target pricing approach for this model, though, would consider a participant hospital's performance relative to its regional peers. As part of this test, we will learn whether our alternative pricing approach in this model will better incentivize participants who are already delivering high quality and efficient care while Start Printed Page 73327still incentivizing historically less efficient providers to improve. We would not be able to test such a regional pricing approach under a purely voluntary model because it is likely that only the already high quality and efficient providers would sign up.

We would note that we have released final evaluation results from the ACE demonstration, which determined that the demonstration led to reduced episode spending with no adverse impact on quality of care. Further, we note that the significant level of voluntary participation in BPCI, as well as high participation in LEJR episodes in particular in all BPCI models, signify the potential for financial opportunity for both hospitals and CMS to achieve savings and improve quality of care through an episode-based payment model targeting LEJR procedures. As further evaluation results for BPCI and other models are available, we will make such information available to the public, and if necessary, could incorporate lessons learned into the CJR model. In addition, in section III.F. of this final rule, we detail our plans to monitor care to ensure beneficiaries' access to quality and timely health care is maintained under the CJR model.

While we acknowledge the benefits of having more rigorous evidence to support the success of episode-based payment models, we believe that the aforementioned findings and encouraging preliminary evaluation data from our prior and current bundled payment models and demonstrations support our plan to more broadly test the model's effectiveness at this time. Moreover, the mission of the Innovation Center is to test models of care that reduce spending while maintaining or improving the quality of care furnished to Medicare, Medicaid and CHIP beneficiaries. Testing this model will provide additional information for CMS and providers on successful payment structures and care redesign strategies.

We also disagree that the model should be delayed simply because other similar efforts are currently ongoing. Rather, we would note that it is not uncommon for CMS to test multiple similar models concurrently rather than sequentially. For example, CMS currently has multiple primary care-focused models in testing, the Comprehensive Primary Care Initiative (CPCI) and the Multi-Payer Advanced Primary Care Practice (MAPCP) models. In addition, CMS has a permanent ACO program (the Medicare Shared Savings Program), as well as multiple other ACO models in the testing phase. We believe our decision to test the CJR model at this time is consistent with the approach taken for other models and programs to test payment models that may be similar in design but are targeted at different groups of providers. Such an approach provides CMS with additional information on the potential success of various model and program aspects and design features.

Likewise, we do not agree that the model should be delayed until certain other actions have occurred (for example, after additional quality measures have been developed or data required under the IMPACT Act have been analyzed) or because of the multiple competing mandates faced by hospitals and other providers. Since the Medicare program's inception, providers have and will continue to contend with constantly evolving statutory and administrative requirements that often require them to make concurrent changes in their practices and procedures. We do not believe the CJR model is dissimilar to those requirements.

As stated previously, some commenters urged a delayed start date due to concerns on how requirements with respect to the CMP law (sections 1128A(a)(5), (b)(1) and (b)(2) of the Act), the Federal Anti-kickback statute (section 1128B(b)(1) and (2) of the Act), or the physician self-referral prohibition (section 1877 of the Act) would apply under the model. In response, we would note that for programmatic reasons discussed elsewhere in this final rule and to give providers additional time to ensure compliance with applicable laws, we are delaying the start date of the model to April 1, 2016.

Also as discussed earlier in this section, some commenters pointed to the potential for unintended consequences that could result from our proposed start date, including impediments to beneficiary access and reduced quality of care. As discussed in section III.D of this final rule, we are including quality measures for purposes of evaluating hospitals' performance both individually and in aggregate across the model. Also, as discussed in section III.F of this final rule, we are making final policies and actions to monitor both care access and quality. We believe these features will help ensure that beneficiary access to high quality care is not compromised under the model.

Final Decision: We are modifying our proposed policy on the model performance years and establishing April 1, 2016 as the start date for the model. Accordingly, we are replacing “January 1, 2016” in § 510.200(a) with “April 1, 2016.”

b. Retrospective Payment Methodology

As described in section III.B. of the proposed rule, we proposed that an episode in the CJR model begins with the admission for an anchor hospitalization and ends 90 days post-discharge from the anchor hospitalization, including all related services covered under Medicare Parts A and B during this timeframe, with limited exclusions and adjustments, as described in sections III.B., III.C.3., and III.C.7. of the proposed rule. The episodes would be attributed to the participant hospital where the anchor hospitalization occurred.

We proposed to apply the CJR episode payment methodology retrospectively. Under this proposal, all providers and suppliers caring for Medicare beneficiaries in CJR episodes would continue to bill and be paid as usual under the applicable Medicare payment system. After the completion of a CJR performance year, Medicare claims for services furnished to beneficiaries in that year that were included in the model would be grouped into episodes and aggregated, and participant hospitals' CJR episode quality and actual payment performance would be assessed and compared against episode quality thresholds and target prices, as described in sections III.C.5. and III.C.4. of the proposed rule, respectively. After the participant hospitals' actual episode performance in quality and spending are compared against the previous episode quality thresholds and target prices, we would determine if Medicare would make a payment to the hospital (reconciliation payment), or if the hospital owes money to Medicare (resulting in Medicare repayment). The possibility for hospitals to receive reconciliation payments or be subject to repayment (note: participant hospitals would not be subject to repayment for performance year 1) was further discussed in section III.C.2.c. of the proposed rule.

We considered an alternative option of paying for episodes prospectively by paying one lump sum amount to the hospital for the expected costs of the 90-day episode. However, we believed such an option would be challenging to implement at this time given the payment infrastructure changes for both hospitals and Medicare that would need to be developed to pay and manage prospective CJR episode payments. We noted that a retrospective episode payment approach is currently being utilized under BPCI Model 2. Also, we expressed our belief that a retrospective payment approach can accomplish the objective of testing episode payment in Start Printed Page 73328a broad group of hospitals, including financial incentives to streamline care delivery around that episode, without requiring core billing and payment changes by providers and suppliers, which would create substantial administrative burden. However, we sought comment on potential ways to implement a prospective payment approach for CJR in future performance years of the model.

The following is a summary of the comments received and our responses.

Comment: Commenters submitted mixed responses on our proposed retrospective payment methodology. Many comments we received expressed support for our proposed retrospective model. Some of these commenters indicated that, since it would build upon existing payment system infrastructures and processes, a retrospective model would be most administratively feasible and straightforward as well as involve fewer infrastructure changes and logistical challenges than would be required under a prospective model. A commenter noted that a retrospective model would allow providers to gain experience with a bundling payment model without altering existing revenue cycle practices. Further, the availability of fee-for-service payments under a retrospective model would maintain a predictable cash flow for participants in the model.

Some commenters expressed support for the proposed retrospective methodology provided that certain conditions existed. For example, a commenter expressed support for this methodology provided that payment reconciliation could be available on a quarterly basis. Another commenter supported the retrospective methodology provided that beneficiaries had access to any provider they chose and were not limited to those with whom a hospital had a contractual arrangement.

Response: We appreciate the comments we received that were in support of our proposed retrospective payment methodology, and concur with commenters' views on some of the benefits of this model. As discussed further in section III.C.6. of this final rule, we are making final our proposed reconciliation on an annual basis. Also, as further discussed in section III.F.2. of this final rule, because hospitals in selected geographic areas will be required to participate in the model, individual beneficiaries will not be able to opt out of the CJR model. However, the payment model does not limit a beneficiary's ability to choose among Medicare providers and suppliers or the range of services that are available to them. Beneficiaries may continue to choose any Medicare enrolled provider or supplier, or any physician or practitioner who has opted out of Medicare, with the same costs, copayments and responsibilities as they have with other Medicare services. Also, although the proposed model would allow participant hospitals to enter into sharing arrangements with certain providers and suppliers and these preferred providers and suppliers may be recommended to beneficiaries as long as those recommendations are made within the constraints of current law, hospitals may not restrict beneficiaries to any list of preferred or recommended providers and suppliers that surpass any restrictions that already exist under current statutes and regulations.

Comment: In addition to the many commenters supporting our proposed retrospective methodology, we received many other comments that opposed our proposal and expressed support for some type of prospective payment model. Some commenters expressed the view that our proposed model was complex, complicated by variation in payment policies across Medicare FFS payment models, and needed further refinement. Others stated that as compared to a prospective payment model, a retrospective model is less effective at holding providers accountable or in stimulating the kinds of behavior changes that are needed to achieve the goals of the program. For example, because providers are expected to change their behavior in anticipation of a reward that might occur several months later, the model diminishes the incentive for providers to change their behavior. Moreover, bonuses and penalties are not sufficiently correlated with performance. Further, a retrospective model could limit the availability of resources for providers to invest in the changes needed to support and sustain behavior change and high-quality care.

Some of the criticisms we received focused on the potential effects of a retrospective model on beneficiaries' costs. For example, some commenters expressed concerns on whether beneficiaries would or even could see cost-sharing reductions when a provider achieves savings under a retrospective model. Another comment suggested that as compared to a prospective model, payments under a retrospective model are more difficult to be incorporated into tools designed to help consumers shop for facilities and providers and reduced pricing predictability for the consumer.

In light of these concerns, many commenters proposed that CMS adopt or eventually transition to some kind of prospective payment model or hybrid model. Commenters suggested that doing so would improve accountability for costs and quality, strengthen risk/reward relationships, better support efforts to transition away from FFS, encourage providers to adhere to evidence-based clinical guidelines, reduce unnecessary or duplicative care, and help participants invest early in supportive resources, such as health information technology, care coordination tools, and infrastructure development to support accountability for quality and costs. A commenter offered the view that information technology solutions are now available that support prospective payment models with minimal burden and disruption to hospitals—concerns that have discouraged the adoption of prospective models.

Some examples of prospective models that were suggested would be for CMS to—

  • Establish an extended DRG that includes hospital, physician, and PAC services for some period of time (for example, 30, 60, 90 days);
  • Make a prospective payment to hospitals that are then distributed to their partners based on volume, acuity, quality, and efficiency;
  • Withhold some percentage of the total payment that would be intended for downstream partners. Hospitals would subsequently distribute these payments to partners based on their ability to meet quality and efficiency targets;
  • Move toward a prospectively negotiated case rate to foster collaboration among all clinicians involved in patient care and provide predictable pricing. For example, give facilities a financial incentive to assume the greater risk and uncertainty inherent in a prospective bundle by reducing or eliminating the two percent discount from the payment benchmark or narrowing the definition of “related care” in the 90-day post-discharge; and
  • Allow physicians to lead a team where the participating physician and their patient decide which other providers and suppliers would be involved in and what the treatment plan would be for the episode. The team would designate or create a jointly governed management organization that would be paid through new prospective episode codes. Other providers, including the hospital, could be paid by that same organization or through existing Medicare payment systems. Medicare would pay a single bundled Start Printed Page 73329payment amount to cover the costs of all of the services in that episode. The hospital and other providers and suppliers on the team could be paid either through the management organization or through traditional Medicare payment systems, but only by one of these sources. Amounts paid through traditional payment systems would be deducted from the amount paid to the management organization.

In addition to comments supporting a prospective payment model, we received comments explicitly expressing concern about adopting such a model. For example, a commenter expressed the view that non-hospital providers and suppliers, including physicians and PAC providers, would likely be concerned with a policy that would allow hospitals complete authority to allocate payments among participating providers and suppliers or to be empowered with functions and authorities typically associated with Medicare Administrative Contractors (MACs). Moreover, a prospective payment methodology would exacerbate anti-competitive concerns with respect to the proposed model in general.

Response: We appreciate the comments we received in opposition to our proposed retrospective model. While we believe that our proposed retrospective payment model would be effective in encouraging providers to improve care quality while better controlling the costs of the care, we also share commenters' optimism on the potential benefits and effectiveness of prospective models with respect to improving accountability for costs and quality, strengthening risk/reward relationships, better supporting efforts to transition away from FFS, and encouraging providers to adhere to evidence-based clinical guidelines while reducing unnecessary or duplicative care. We also are pleased that information technology solutions are being developed to support prospective payment models.

We agree with commenters that there are complexities and potential complications associated with a retrospective model and anticipate that further refinements will likely be needed with whatever kind of bundling model that is implemented. Therefore, we do not believe that the complexities and potential complications with our proposed model are significantly different than what occurs with other Medicare payment models, particularly any of the more novel ones. Likewise, we do not believe that such complexities or complications would be mitigated simply by adopting a prospective model. Moreover, both CMS and some of the commenters have noted that adoption of a prospective model could result in potentially significant complexities and logistical issues as well.

We also do not agree with the view suggesting that adoption of a retrospective model could limit the availability of resources for providers to invest in the changes needed to improve care quality and costs. Under our retrospective model, participant hospitals and other providers and suppliers will continue to bill and be paid under FFS Medicare as they would in the absence of the model that should result in a revenue stream comparable to what they would be absent the model, all else equal.

While we agree with the comment stating that beneficiaries will not see a reduction in their cost-sharing for joint replacement services under this model, we do not see this as being unique to the CJR model or a reason to not test it. To the contrary, if successful, our model will improve the quality of care and outcomes for these beneficiaries as well as better control costs of care. For example, if successful, we believe the model could help to limit or mitigate avoidable costs incurred by these beneficiaries such as costs associated with avoidable hospital readmissions. Last, we also do not see the potential challenges of integrating a retrospective payment methodology into sites designed to compare health care options as a reason to not test our proposed model or as being an insurmountable problem.

Based on the comments that we received, we believe there is support for both prospective and retrospective payment models. We also continue to believe that a retrospective payment model can accomplish the objective of testing episode payments with a broad group of hospitals, by including financial incentives that will streamline care delivery while producing less administrative burden for providers than would be possible with a prospective model. Accordingly, we will be implementing a retrospective payment model at this time as we had proposed. We appreciate the various examples of prospective models that commenters suggested for CMS' consideration, and will consider these examples along with other options to potentially be tested in the future.

Final Decision: After considering the public comments we received, we are finalizing our proposal to implement a retrospective payment model.

c. Two-Sided Risk Model

We proposed to establish a two-sided risk model for hospitals participating in the CJR model. We proposed to provide episode reconciliation payments to hospitals that meet or exceed quality performance thresholds and achieve cost efficiencies relative to CJR target prices established for them, as was defined later in sections III.C.4. and III.C.5. of the proposed rule. Similarly, we proposed to hold hospitals responsible for repaying Medicare when actual episode payments exceed their CJR target prices in each of performance years 2 through 5, subject to certain proposed limitations discussed in section III.C.8. of the proposed rule. Target prices would be established for each participant hospital for each performance year.

We proposed that hospitals will be eligible to receive reconciliation payments from Medicare based on their quality and actual episode spending performance under the CJR model in each of CJR performance years 1 through 5. Additionally, we proposed to phase in the responsibility for hospital repayment of episode actual spending if episode actual spending exceeds their target price starting in performance year 2 and continuing through performance year 5. Under this proposal in performance year 1, participant hospitals would not be required to pay Medicare back if episode actual spending is greater than the target price.

We considered an episode payment structure in which, for all 5 performance years of the model, participant hospitals would qualify for reconciliation payments if episode actual spending was less than the episode target price, but would not be required to make repayments to Medicare if episode actual spending was greater than the episode target price. However, we noted our belief that not holding hospitals responsible for repaying excess episode spending would reduce the incentives for hospitals to improve quality and efficiency. We also considered starting the CJR payment model with hospital responsibility for repaying excess episode spending in performance year 1 to more strongly align participant hospital incentives with care quality and efficiency. However, we stated our view that hospitals may need to make infrastructure, care coordination and delivery, and financial preparations for the CJR episode model, and that those changes can take several months or longer to implement. With this consideration in mind, we proposed to begin hospitals' responsibility for repayment of excess episode spending Start Printed Page 73330beginning in performance year 2 to afford hospitals time to prepare, while still beginning some incentives earlier (that is, reconciliation payments in year 1) to improve quality and efficiency of care for Medicare beneficiaries. We solicited comment on the proposed incentive structure for CJR.

In an effort to further ensure hospital readiness to assume responsibility for circumstances that could lead to a hospital repaying to Medicare actual episode payments that exceed the episode target price, we proposed to begin to phase in this responsibility for performance year 2, with full responsibility for excess episode spending (as proposed in the proposed rule) applied for performance year 3 through performance year 5. To carry out this “phase in” approach, we proposed during the first year of any hospital financial responsibility for repayment (performance year 2) to set an episode target price that partly mitigates the amount that hospitals would be required to repay (see section III.C.4.b. of the proposed rule), as well as more greatly limits (as compared to performance years 3 through 5) the maximum amount a hospital would be required to repay Medicare across all of its episodes (see section III.C.8. of the proposed rule).

Comment: Several commenters expressed support for our proposal to establish downside risk for participants as well as our proposal to gradually phase-in risk beginning in year 2. We received very few comments requesting the elimination of risk from the model. A commenter suggested that it was unfair to require hospitals to bear risk given that there were no limitations on beneficiary choices. Also, some commenters suggested that CMS consider excluding specific kinds of hospitals from the model, for example small hospitals or hospitals with low volume.

Most of the comments we received, however, requested that CMS ease the glide path to downside risk by either delaying the requirement for two to three years or by incorporating features to better limit risk, for example, by adjusting stop-loss caps. Some commenters requested that we modify the CJR model to be more like a shared savings model as is used in Shared Savings Program or the Pioneer ACO model. In their view, this option would be particularly attractive to smaller organizations with lower episode volumes that face a higher risk of random episode cost variation or those with limited financial resources.

Some commenters requested these changes because of concerns that hospitals have little or no experience bearing risk and thus need additional time to be ready to do so. Other commenters stated that our proposed timeframe for implementing the model and requiring hospitals to assume risk was simply too aggressive and offered too little time for hospitals to put in place the care procedures and infrastructure needed to be successful in the model and in a position to bear risk. In recommending that CMS delay downside risk, a commenter observed that payment features of other Medicare efforts such as BPCI and the Pioneer ACO model have been refined more than once since their implementation, which suggested that more can be learned about the appropriate framework for a risk model, particularly given that the CJR model is untested.

Response: We appreciate the comments we received in support of our proposal to phase-in downside risk to CJR participants beginning in Year 2 of the model. We are also encouraged that very few commenters opposed a requirement for participants to assume downside risk at some point in the model.

We disagree with the view that it is unfair to require hospitals to bear risk while beneficiaries retain the ability to choose among providers. As is the case with other new payment models such as the Shared Savings Program, the CJR model is intended to identify ways to improve care quality and better control costs in the Medicare FFS program. While Medicare beneficiaries may choose between Medicare FFS and Medicare Advantage, the majority of beneficiaries—roughly two-thirds in 2015—continue to choose the former. Accordingly, it is in the interest of the Medicare program and its beneficiaries for CMS to identify new models that both maintain beneficiary choice while improving care quality and costs. Also, while we appreciate suggestions to exclude certain kinds of hospitals, for example, small hospitals or hospitals with a low-volume of cases, we believe our methodology for selecting geographic units, as discussed in section III.A.4.of this final rule, as well as additional protections for certain kinds of these hospitals, as discussed in section III.C.8.c. of this final rule sufficiently address these concerns.

We also understand that commenters would like a more gradual transition to downside risk, and in response to the commenters' concerns, CMS has taken steps for hospitals to do so. As discussed in section III.C.8. of this final rule, we are reducing the potential risk to participants in Year 2 by lowering the stop-loss limit from 10 percent to 5 percent (and from 20 percent to 10 percent in Year 3). We believe these actions should assist participants both with respect to preparing for the assumption of risk as well as reducing the level of risk they must initially bear. We do not support the proposal to change the CJR model to a shared savings model as it is inconsistent with our intent of testing whether a bundled payments model will promote quality and financial accountability for episodes of care surrounding an LEJR or reattachment of a lower extremity procedure. Last, we recognize that our model, as would any model or program, will evolve and may require some adjustments over time. To the extent that this occurs with the CJR model, we would make adjustments that were deemed necessary, as we would do with any of these other models and programs; however, we do not believe the potential for model adjustments is a reason to delay the requirement for hospitals to bear risk in the absence of data suggesting that a problem actually exists.

Final Decision: After considering the public comments we received, we are finalizing our proposal to phase-in risk beginning in Year 2 of the model.

3. Adjustments to Payments Included in Episode

We proposed to calculate the actual episode payment amount by summing together Medicare payments for each non-cancelled CJR episode during the model's performance year for Parts A and B claims for services included in the episode definition, as discussed in section III.B. of this final rule. We proposed three adjustments to this general approach for—(1) Special payment provisions under existing Medicare payment systems; (2) payment for services that straddle the end of the episode; and (3) high payment episodes. We noted there would be further adjustments to account for overlaps with other Innovation Center models and CMS programs; we refer readers to section III.C.7. of the proposed rule.

We did not propose to adjust hospital-specific or regional components of target prices for any Medicare repayment or reconciliation payments made under the CJR model; CJR repayment and reconciliation payments would be not be included per the episode definition in section III.B. of this final rule. We stated in the proposed rule our belief that including reconciliation payments and Medicare repayments in target price calculations would perpetuate the initial set of target prices once CJR performance years are captured in the 3-historical-years of data used to set target Start Printed Page 73331prices, as described in section III.C.4. of this final rule, beginning with performance year 3 when performance year 1 would be part of the 3-historical-years. Including any prior performance years' reconciliations or repayments in target price calculations would approximately have the effect of Medicare paying hospitals the target price, regardless of whether the hospital went below, above, or met the target price in the prior performance years before accounting for the reconciliation payments or repayments. We stated in the proposed rule our intent for target prices to be based on historical patterns of service actually provided, so we did not propose to include reconciliation payments or repayments for prior performance years in target price calculations.

a. Treatment of Special Payment Provisions Under Existing Medicare Payment Systems

Many of the existing Medicare payment systems have special payment provisions that have been created by regulation or statute to improve quality and efficiency in service delivery. IPPS hospitals are subject to incentives under the HRRP, the Hospital Value-Based Purchasing (HVBP) Program, the Hospital-Acquired Condition (HAC) Reduction Program, and the Hospital Inpatient Quality Reporting Program (HIQR) and Outpatient Quality Reporting Program (OQR). IPPS hospitals and CAHs are subject to the Medicare EHR Incentive Program. Additionally, the majority of IPPS hospitals receive additional payments for Medicare Disproportionate Share Hospital (DSH) and Uncompensated Care, and IPPS teaching hospitals can receive additional payments for Indirect Medical Education (IME). IPPS hospitals that meet a certain requirements related to low volume Medicare discharges and distance from another hospital receive a low volume add-on payment. As previously stated in section III.B.2.b. of this final rule, acute care hospitals may receive new technology add-on payments to support specific new technologies or services that substantially improve the diagnosis or treatment of Medicare beneficiaries and would be inadequately paid otherwise under the MS-DRG system. Also, some IPPS hospitals qualify to be sole community hospitals (SCHs) or MDHs, and they may receive enhanced payments based on cost-based hospital-specific rates for services; whether a SCH or MDH receives enhanced payments may vary year to year, in accordance with § 419.43(g) and § 412.108(g), respectively.

Medicare payments to providers of PAC services, including IRFs, SNFs, IPFs, HHAs, LTCHs, and hospice facilities, are conditioned, in part, on whether the provider satisfactorily reports certain specified data to CMS: The Inpatient Rehabilitation Facility Quality Reporting Program (IRF QRP), the Skilled Nursing Facility Quality Reporting Program (SNF QRP), the Inpatient Psychiatric Facility Quality Reporting Program (IPF QRP), the Home Health Quality Reporting Program (HH QRP), the Long-Term Care Hospital Quality Reporting Program (LTCH QRP), and the Hospice Quality Reporting Program. Additionally, IRFs located in rural areas receive rural add-on payments, IRFs serving higher proportions of low-income beneficiaries receive increased payments according to their low-income percentage (LIP), and IRFs with teaching programs receive increased payments to reflect their teaching status. SNFs receive higher payments for treating beneficiaries with human immunodeficiency virus (HIV). HHAs located in rural areas also receive rural add-on payments.

ASCs have their own Quality Reporting Program (ASC QRP). Physicians also have a set of special payment provisions based on quality and reporting: The Medicare EHR Incentive Program for Eligible Professionals, the Physician Quality Reporting System (PQRS), and the Physician Value-based Modifier Program.

In the proposed rule we stated our intent with the CJR model is not to replace the various existing incentive programs or add-on payments, but instead to test further episode payment incentives towards improvements in quality and efficiency beyond Medicare's existing policies. Therefore, we proposed that the hospital performance and potential reconciliation payment or Medicare repayment be independent of, and not affect, these other special payment provisions.

We proposed to exclude the special payment provisions as discussed previously when calculating actual episode payments, setting episode target prices, comparing actual episode payments with target prices, and determining whether a reconciliation payment should be made to the hospital or funds should be repaid by the hospital.

Not excluding these special payment provisions would create incentives that are not aligned with the intent of the CJR model. Not excluding the quality and reporting-related special payment provisions could create situations where a high-quality or reporting compliant hospital or both receiving incentive payments, or those hospitals that discharge patients to PAC providers that receive incentives for being reporting compliant, may appear to be “high episode payment” under CJR. Conversely, lower quality or hospitals not complying with reporting programs or both that incur payment reduction penalties, or hospitals that discharge to PAC providers that are not reporting compliant, may appear to be “low episode payment” under CJR. Such outcomes would run counter to CJR's goal of improving quality. Also, not excluding add-on payments for serving more indigent patients, having low Medicare hospital volume, being located in a rural area, supporting greater levels of provider training, choosing to use new technologies, and having a greater proportion of CJR beneficiaries with HIV from CJR actual episode payment calculations may inappropriately result in hospitals having worse episode payment performance. Additionally, not excluding enhanced payments for MDHs and SCHs may result in higher or lower target prices just because these hospitals received their enhanced payments in one historical year but not the other, regardless of actual utilization. In the proposed rule we stated our belief that excluding special payment provisions would ensure a participant hospital's actual episode payment performance is not artificially improved or worsened because of payment reduction penalties or incentives or enhanced or add-on payments, the effects of which we are not intending to test with CJR.

In addition to the various incentive, enhanced and add on payments, sequestration came into effect for Medicare payments for discharges on or after April 1, 2013, per the Budget Control Act of 2011 and delayed by the American Taxpayer Relief Act of 2012. Sequestration applies a 2 percent reduction to Medicare payment for most Medicare FFS services.

In order to operationalize the exclusion of the various special payment provisions in calculating episode expenditures, we proposed to apply the CMS Price (Payment) Standardization Detailed Methodology described on the QualityNet Web site at http://www.qualitynet.org/​dcs/​ContentServer?​c=​Page&​pagename=​QnetPublic%2FPage%2FQnetTier4&​cid=​1228772057350. This pricing standardization approach is the same as used for the HVBP program's Medicare spending per beneficiary metric.

We sought comment on this proposed approach to treating special payment Start Printed Page 73332provisions in the various Medicare payment systems.

Comment: Several commenters supported the exclusion of the various special payment provisions in calculating episode expenditures. They agreed that doing so would help isolate the effect of utilization and quality of delivered care differences and remove any distortions due to Medicare payment policies outside the control of providers.

A few commenters expressed concern about how hospitals would be paid the special payment adjustments that are removed in calculating episode expenditures. A commenter inquired whether CMS would account for vendor rebates for hip and knee implants and medical devices, because rebates are not uncommon and can impact the cost of an LEJR procedure to a hospital.

Response: We appreciate commenters' support to exclude the various special payment provisions in calculating episode expenditures.

As discussed in section III.C.2.b. of this final rule, we are finalizing our proposal such that all providers and suppliers caring for Medicare beneficiaries in CJR episodes will continue to bill and be paid as usual under the applicable Medicare payment system, and determination of any reconciliation payments or repayments to Medicare will be made retrospectively after the end of each performance year. Therefore, special payment adjustments will continue to be paid as usual under the applicable Medicare payment systems, but their effects will be excluded when reconciliation payment and repayment to Medicare determinations are made retrospectively. This final rule will not affect how hospitals are currently paid special payment adjustments.

Payments for hip and knee implants and medical devices will also continue as usual under the applicable Medicare payment systems. For inpatient admissions paid under IPPS, in particular, implants and medical devices not categorized as new technology add-on payment would be included in the MS-DRG payment and would not be reimbursed separately. To mirror the IPPS approach, we will not separately account for vendor rebates in the LEJR episode.

We note that as previously stated, we plan to utilize the CMS Price (Payment) Standardization approach in order to remove the effects of special payment provisions from calculations of historical and performance period episode spending. We will follow the methodology, with modifications as necessary to be consistent with our episode definition in section III.B. of this final rule and to ensure timely reporting of reconciliation results, for the performance year reconciliations, which begin 2 months after the conclusion of a performance year. We will account for the information available at the time due to claims runout, payment system updates, and the calculations necessary to fully implement the standardization methodology. We will utilize the methodology, consistent with our episode definition, for the target price calculations and subsequent reconciliation calculations 14 months after the conclusion of the performance year, in which we incorporate full claims runout and further account for overlap with other models. This approach will provide feedback and reconciliation payments, as available, to hospitals in a timely manner and as accurately as feasible, while ensuring the standardization approach is utilized for the subsequent calculation, which represents the final calculation for a given performance period.

Comment: Many commenters requested that CJR reconciliation payments made to participant hospitals be included when updating the set of 3-historical-years used for calculating CJR episode target prices. They stated that the participant hospitals would be providing care coordination services that may not be directly reimbursed under applicable Medicare FFS payment systems. These services would then, instead, be funded by reconciliation payments. While historical Medicare FFS claim payments would account for hospitals' costs for providing services reimbursed under Medicare FFS, they would not account for hospitals' costs for care coordination services not reimbursed under Medicare FFS. Commenters contended that if we do not include reconciliation payments when calculating target prices using the updated set of historical years, we may underestimate hospital costs and target prices.

Response: We agree that participant hospitals may undertake activities that promote care coordination and improved quality of care but are not directly reimbursed under applicable Medicare FFS payment systems. We appreciate commenters' suggestions to include reconciliation payments when updating the set of historical years used to calculate target prices. We also believe this logic could be extended to include repayments to Medicare to mirror the inclusion of reconciliation payments. However, in the proposed rule we did not propose an alternative to include reconciliation payments and repayments when updating the set of historical years used to calculate target prices, and because the first time this policy would take effect would be for performance year 3 (2018), we may revisit this policy in future rulemaking and allow for public comment on the aforementioned alternative. At this time we are not modifying our proposal to exclude CJR reconciliation payments and repayments to Medicare when updating the set of historical years used to set target prices.

Comment: A few commenters inquired whether claims from non-participating physicians or payments to physicians who have opted out of Medicare would be included for purposes of setting target prices and calculating actual episode spending for reconciliation and repayment amount calculations. Commenters contended that if claims from non-participating providers or payments to physicians who have opted out of Medicare are not included, target prices and actual episode spending may be underestimated.

Response: With the exception of those physicians and practitioners who have complied with our opt-out procedures (see 42 CFR 405.400 through 405.455), when a physician or supplier furnishes a service that is covered by Medicare, the physician or supplier is subject to the mandatory claim submission provisions of section 1848(g)(4) of the Social Security Act (the Act). Therefore, if a physician or supplier charges or attempts to charge a beneficiary for a service that is covered by Medicare, then the physician or supplier must submit a claim to Medicare. As a result, claims from both participating and non-participating physicians would be included in our target price and actual episode spending calculations.

Opt-out physicians are prohibited from billing and receiving payment (either directly or indirectly) from Medicare except for emergency and urgent care services provided the physician has not previously entered into a private contract with the beneficiary. Therefore, we agree that payments for services furnished by physicians who have opted out of Medicare would not be included in target price and actual episode expenditure calculations. However, we estimate only a small portion of physicians furnishing services to beneficiaries captured in the CJR model will have opted out of Medicare, and we estimate that physician services comprise less than 15 percent of the average CJR episode expenditure, and therefore we believe the impact of not capturing expenditures from physicians Start Printed Page 73333who have opted out of Medicare will be small.

Additionally, there may be some participant hospitals with a disproportionately higher share of episodes for which services were furnished by physicians who have opted out of Medicare. Such participant hospitals would experience lower actual episode expenditures because payments for physicians who have opted out of Medicare would not be included. These hospitals' lower actual episode expenditures would be balanced by lower target prices because the payments for physicians who have opted out of Medicare would also be excluded in the historical episode expenditures, though this argument is primarily relevant in the early years of the CJR model before we move to 100 percent regional pricing as discussed in section III.C.4.b.(5) of this final rule. In the later years of this model, participant hospitals with disproportionately greater share of episodes for which services were furnished by Medicare opt-out physicians may unfairly benefit from regional target prices that are primarily based on the inclusion of expenditures for physician services. However, we believe this advantage to be small because physician expenditures comprise only a small portion of the average episode, and we expect very few physicians to opt out of Medicare.

Comment: A commenter inquired whether CMS would include IPPS capital payments in calculating target prices and actual episode expenditures, and if CMS' plan was to include them, they requested that such payments be excluded. The commenter stated that capital payments may vary by hospitals, and excluding capital payments would be consistent with the pricing standardization approach we proposed to reduce variations due to Medicare payment policies. The commenter also noted that excluding capital payments would be consistent with the approach taken in BPCI.

Response: In response to comments, we clarify that we will include IPPS capital payments in target price and actual episode expenditure calculations. IPPS capital payments are included in Medicare FFS payments, which we proposed to use to calculate target prices and actual episode expenditures. Consistent with our proposed treatment of special payment provisions, we do not intend to distort incentives based on IPPS capital payments that may vary across hospitals due to Medicare payment policies, as opposed to practice pattern and quality differences. By using the claims standardization approach previously described in this section, though, we will be able to remove the effect of variations due to Medicare payment policies (including wage index differences). We recognize that this approach of including IPPS capital payments would be different than the approach taken in BPCI. However, we note that other Medicare FFS payment systems, such as those for SNF and IRF, also are intended to cover providers' capital costs. Carving out the capital portion for IPPS payments would not be consistent with the inclusion of the capital portion for other Medicare FFS payment systems. Lastly, including IPPS capital payments affords participant hospitals an opportunity to achieve greater reconciliation payments if they are able to achieve efficiencies for the costs that the capital portion of IPPS payments would cover, which may or may not actually be capital costs.

Comment: Several commenters expressed concern about the regions that were selected for both the CJR model and the proposed HHVBP model.

Response: We refer readers to comments and responses to comments in section III.A.3 of this final rule for further discussion on the inclusion of regions selected for both the CJR model and the proposed HHVBP model, and we reference it here because the proposed HHVBP model would be another special payment provision that could affect Medicare payment amounts. We reemphasize that the intent of the CJR model is not to replace the various existing incentive programs or add on payments, and the claims standardization approach previously described in this section will remove the effect of any special payment provision, whether they currently exist or may be introduced in the future. Therefore, we do not believe any special payment provisions due to the proposed HHVBP model or other potential future special payment provisions to have an impact on the payments included in the CJR model target price and reconciliation calculations.

Comment: A commenter requested clarification on how the CJR model would interact with Medicare beneficiaries who have exhausted their benefits, and recommended that we modify Medicare beneficiaries' benefits so as to not allow their benefits to be exhausted while part of a CJR episode.

Response: We appreciate the commenter's suggestion. However, we did not propose any changes to Medicare beneficiaries' benefits, and we will not finalize any such changes in this final rule.

Final Decision: We are finalizing our proposal, without modification, to exclude special payment provisions from episode calculations. We clarify that we will include IPPS capital payments in target price and actual episode expenditure calculations. We also clarify that we will utilize the CMS Price Standardization approach previously referenced to remove the effect of any current and potential future special payment provisions. We may revisit in future rulemaking any modification to our policy to exclude reconciliation and recoupment payments when updating the historical data used to set target prices.

b. Treatment of Payment for Services That Extend Beyond the Episode

As we proposed a fixed 90-day post-discharge episode as discussed in section III.B. of the proposed rule, we stated our belief that there would be some instances where a service included in the episode begins during the episode but concludes after the end of the episode and for which Medicare makes a single payment under an existing payment system. An example would be a beneficiary in a CJR episode who is admitted to a SNF for 15 days, beginning on Day 86 post-discharge from the anchor CJR hospitalization. The first 5 days of the admission would fall within the episode, while the subsequent 10 days would fall outside of the episode.

We proposed that, to the extent that a Medicare payment for included episode services spans a period of care that extends beyond the episode, these payments would be prorated so that only the portion attributable to care during the episode is attributed to the episode payment when calculating actual Medicare payment for the episode. For non-IPPS inpatient hospital (for example, CAH) and inpatient PAC (for example, SNF, IRF, LTCH, IPF) services, we proposed to prorate payments based on the percentage of actual length of stay (in days) that falls within the episode window. Prorated payments would also be similarly allocated to the 30-day post-episode payment calculation in section III.C.8.d. of this final rule. In the prior example, one-third of the days in the 15-day length of stay would fall within the episode window, so under the proposed approach, one-third of the SNF payment would be included in the episode payment calculation, and the remaining two-thirds (because the entirety of the remaining payments fall within the 30 days after the episode ended) would be included in the post-episode payment calculation.

For HHA services that extend beyond the episode, we proposed that the Start Printed Page 73334payment proration be based on the percentage of days, starting with the first billable service date (“start of care date”) and through and including the last billable service date, that fall within the CJR episode. Prorated payments would also be similarly allocated to the 30-day post-episode payment calculation in section III.C.8.d. of the proposed rule. For example, if the patient started receiving services from an HHA on day 86 after discharge from the anchor CJR hospitalization and the last billable home health service date was 55 days from the start of home health care date, the HHA claim payment amount would be divided by 55 and then multiplied by the days (5) that fell within the CJR episode. The resulting, prorated HHA claim payment amount would be considered part of the CJR episode. Services for the prorated HHA service would also span the entirety of the 30 days after the CJR episode spends, so the result of the following calculation would be included in the 30-day post-episode payment calculation: HHA claim payment amount divided by 55 and then multiplied by 30 days (the number of days in the 30-day post-episode period that fall within the prorated HHA service dates).

There may also be instances where home health services begin prior to the CJR episode start date, but end during the CJR episode. In such instances, we also proposed to prorate HHA payments based on the percentage of days that fell within the episode. Because these services end during the CJR episode, prorated payments for these services would not be included in the 30-day post-episode payment calculation discussed in section III.C.8.d. of the proposed rule. For example, if the patient's start of care date for a home health 60-day claim was February 1, the anchor hospitalization was March 1 through March 4 (with the CJR episode continuing for 90 days after March 4), and the patient resumed home care on March 5 with the 60-day home health claim ending on April 1 (that is, April 1 was the last billable service date), we would divide the 60-day home health claim payment amount by 60 and then multiply that amount by the days from the CJR admission through April 1 (32 days) to prorate the HHA payment. This proposed prorating method for HHA claims is consistent with how partial episode payments (PEP) are paid for on home health claims.

For IPPS services that extend beyond the episode (for example, readmissions included in the episode definition), we proposed to separately prorate the IPPS claim amount from episode target price and actual episode payment calculations as proposed in section III.C.8. of the proposed rule, called the normal MS-DRG payment amount for purposes of this final rule. The normal MS-DRG payment amount would be pro-rated based on the geometric mean length of stay, comparable to the calculation under the IPPS PAC transfer policy at § 412.4(f) and as published on an annual basis in Table 5 of the IPPS/LTCH PPS Final Rules. Consistent with the IPPS PAC transfer policy, the first day for a subset of MS-DRGs (indicated in Table 5 of the IPPS/LTCH PPS Final Rules) would be doubly weighted to count as 2 days to account for likely higher hospital costs incurred at the beginning of an admission. If the actual length of stay that occurred during the episode is equal to or greater than the MS-DRG geometric mean, the normal MS-DRG payment would be fully allocated to the episode. If the actual length of stay that occurred during the episode is less than the geometric mean, the normal MS-DRG payment amount would be allocated to the episode based on the number of inpatient days that fall within the episode. If the full amount is not allocated to the episode, any remainder amount would be allocated to the 30 day post-episode payment calculation discussed in section III.C.8.d. of the proposed rule. The proposed approach for prorating the normal MS-DRG payment amount is consistent with the IPPS transfer per diem methodology.

The following is an example of prorating for IPPS services that extend beyond the episode. If beneficiary has a readmission for MS-DRG 493—lower extremity and humerus procedures except hip, foot, and femur, with complications—into an IPPS hospital on the 89th day after discharge from a CJR anchor hospitalization, and is subsequently discharged after a length of stay of 5 days, Medicare payment for this readmission would be prorated for inclusion in the episode. Based on Table 5 of the IPPS/LTCH PPS Final Rule for FY 2015, the geometric mean for MS-DRG 493 is 4 days, and this MS-DRG is indicated for double-weighting the first day for proration. This readmission has only 2 days that falls within the episode, which is less than the MS-DRG 493 geometric mean of 4 days. Therefore, the normal MS-DRG payment amount associated with this readmission would be divided by 4 (the geometric mean) and multiplied by 3 (the first day is counted as 2 days, and the second day contributes the third day), and the resulting amount is attributed to the episode. The remainder one-fourth would be captured in the post-episode spending calculation discussed in section III.C.8. of the proposed rule. If the readmission occurred on the 85th day after discharge from the CJR anchor hospitalization, and the length of stay was 7 days, the normal MS-DRG payment amount for the admission would be included in the episode without proration because length of stay for the readmission falling within the episode (6 days) is greater than or equal to the geometric mean (4 days) for the MS-DRG.

We considered an alternative option of including the full Medicare payment for all services that start during the episode, even if those services did not conclude until after the episode ended, in calculating episode target prices and actual payments. Previous research on bundled payments for episodes of PAC services noted that including the full payment for any claim initiated during the fixed episode period of time will capture continued service use. However, prorating only captures a portion of actual service use (and payments) within the bundle.[20] As discussed in section III.B. of this final rule, the CJR model proposed an episode length that extends 90 days post-discharge, and Table 5 in section III.B.3.c. of the proposed rule demonstrates that the average length of stay in PAC during a 90-day episode with a MS-DRG 470 anchor hospitalization is 47.3 days. Therefore, the length of the episode under CJR (90 days) should be sufficient to capture the vast majority of service use within the episode, even if payments for some services that extend beyond the episode duration are prorated and only partly attributed to the episode.

The following is a summary of comments received and our responses.

Comment: Several commenters supported the pro-rating of payments for services that extend beyond the episode. They agreed that pro-rating would help ensure target prices and actual episode payments reflect services that were furnished during the episode. A commenter requested clarification on how payments for IRFs would be pro-rated. Another commenter stated that the first day for pro-rated surgical MS-DRGs paid under IPPS should be weighted by more than the two-times weight proposed; the commenter believed that a multiplier of up to 4.5 would more accurately describe hospitals' costs for the first day of surgical inpatient admissions reimbursed under Medicare IPPS.Start Printed Page 73335

Response: We appreciate commenters' support for pro-rating payments for services that extend beyond the episode. As described in section III.C.3.b of this final, IRF payments will be pro-rated based on the percentage of actual length of stay (in days) that falls within the episode window. Prorated IRF payments would also be similarly allocated to the 30 day post episode payment calculation in section III.C.8.d. of this final rule.

We agree that costs for inpatient stays may not be equal for each day of an inpatient admission, and the distribution of costs may differ between surgical and non-surgical inpatient stays. We acknowledge there may be different methodologies to calculate how much more costs are incurred on the first day of a stay. However, we will maintain consistency with the IPPS per diem transfer policy that uses a two-times weight for the first day for a subset of MS-DRGs as described in § 412.4(f) and published on an annual basis in Table 5 of the IPPS/LTCH PPS Final Rules. We also note that many surgical readmissions are excluded from the episode definition described in section III.B. of this final rule, which should mitigate the impact of this prorating approach on surgical readmissions that extend beyond the episode.

Final Decision: After consideration of the public comments we received, we are finalizing the proposal to prorate payments for services that extend beyond the episode when calculating actual episode payments, setting episode target prices, and calculating reconciliation and repayment amounts.

c. Pricing Adjustment for High Payment Episodes

Given the broad proposed LEJR episode definition and 90-day post-discharge episode duration proposed for CJR, we want to ensure that hospitals have some protection from the variable repayment risk for especially high payment episodes, where the clinical scenarios for these cases each year may differ significantly and unpredictably. We did not believe the opportunity for a hospital's systematic care redesign of LEJR episodes has significant potential to impact the clinical course of these extremely disparate high payment cases.

The BPCI Model 2 uses a generally similar episode definition as proposed for CJR and the vast majority of BPCI episodes being tested for LEJR are 90 days in duration following discharge from the anchor hospitalization. Similarly in the proposed rule, we stated our belief that the distribution of 90-day LEJR episode payment amounts utilizing the BPCI Model 2 episode definition as displayed in Figure 2 provides information that is relevant to policy development regarding CJR episodes.

Start Printed Page 73336

As displayed, the mean episode payment amount is approximately $26,000. Five percent of all episodes are paid at two standard deviations above the mean payment or greater, an amount that is slightly more than 2 times the mean episode payment amount. While these high payment cases are relatively uncommon, we stated in the proposed rule our belief that incorporation of the full Medicare payment amount for such high payment episodes in setting the target price and correspondingly in Medicare's aggregate actual episode payment that is compared to the target price for the episode may lead in some cases to excessive hospital responsibility for these episode expenditures. This may be especially true when hospital responsibility for repayment of excess episode spending is introduced in performance year 2. The hospital may have limited ability to moderate spending for these high payment cases. Our proposal to exclude IPPS new technology add-on payments and separate payment for clotting factors for the anchor hospitalization from the episode definition limits excessive financial responsibility under this model of extremely high inpatient payment cases that could result from costly hospital care furnished during the anchor hospitalization. However, in the proposed rule we stated our belief that an additional pricing adjustment in setting episode target prices and calculating actual episode payments is necessary to mitigate the hospital responsibility for the actual episode payments for high episode payment cases resulting from very high Medicare spending within the episode during the period after discharge from the anchor hospitalization, including for PAC, related hospital readmissions, and other items and services related to the LEJR episode.

Thus, in order to limit the hospital's responsibility for the previously stated high episode payment cases, we proposed to utilize a pricing adjustment for high payment episodes that would incorporate a high payment ceiling at two standard deviations above the mean episode payment amount in calculating the target price and in comparing actual episode payments during the performance year to the target prices.

Specifically, when setting target prices, we would first identify for each anchor MS-DRG in each region (discussed further in section III.C.4. of this final rule) the episode payment amount that is two standard deviations above the mean payment in the historical dataset used (discussed further in section III.C.4. of the proposed rule). Any such identified episode would have its payment capped at the MS-DRG anchor and region-specific value that is two standard deviations above the mean, which would be the ceiling for purposes for calculating target prices. We note that the calculation of the historical episode high payment ceiling for each region and MS-DRG anchor would be performed after other steps, including removal of effects of special payment provisions and others described in section III.C.4.c. of this final rule.

When comparing actual episode payments during the performance year to the target prices, episode payments for episodes in the performance year would also be capped at two standard deviations above the mean. The high episode payment ceiling for episodes in a given performance year would be calculated based on MS-DRG anchor-specific episodes in each region. We discuss further how the high episode payment ceiling would be applied when comparing episode payments during the performance year to target prices in section III.C.6. of this final rule.

While this approach generally lowers the target price slightly, it provides a basis for reducing the hospital's responsibility for actual episode spending for high episode payment cases during the model performance years. When performing the reconciliation for a given performance year of the model, we would array the actual episode payment amounts for all episodes being tested within a single region, and identify the regional actual episode payment ceiling at two standard deviations above the regional mean actual episode payment amount. If the actual payment for a hospital's episode exceeds this regional ceiling, we would set the actual episode payment amount to equal the regional ceiling amount, rather than the actual amount paid by Medicare, when comparing a hospital's episode spending to the target price. Thus, a hospital would not be responsible for any actual episode payment that is greater than the regional ceiling amount for that performance year. We proposed to adopt this policy for all years of the model, regardless of the reconciliation payment opportunity or repayment responsibility in a given performance year, to achieve stability and consistency in the pricing methodology. We stated in the proposed rule our belief that this proposal provides reasonable protection for hospitals from undue financial responsibility for Medicare episode spending related to the variable and unpredictable course of care of some Medicare beneficiaries in CJR episodes, while still fully incentivizing increased efficiencies for approximately the 95 percent of episodes for which we estimate actual episode payments to fall below this ceiling.[21] We sought comment on our proposal to apply a pricing adjustment in setting target prices and reconciling actual episode payments for high payment episodes.

The following is a summary of the comments received and our responses.

Comment: Many commenters supported the proposal for a high episode payment ceiling at two standard deviations above the mean episode payment amount in calculating the target price and in comparing actual episode payments during the performance year to the target prices. They agreed that such a ceiling would help limit financial exposure to participant hospitals from outlier episodes. Some commenters requested the option of choosing specific risk tracks as provided under BPCI (for example, high episode payment ceiling at 75th, 95th, or 99th percentile).

Response: We appreciate commenters' support for a high episode payment ceiling. We acknowledge that BPCI offers different risk tracks with different outlier protection features from which participants can choose, and that we did not propose to provide CJR participant hospitals with choice of risk tracks or outlier protection policy. However, with the blending of regional and hospital-specific historical episode expenditure data that we are finalizing in section III.C.4.b.(5) of this final rule to calculate target prices, applying different risk tracks or outlier protection policies to different hospitals would distort target price calculations; this is not an issue in BPCI because target prices are calculated using only hospital-specific historical episode expenditure data. Additionally, we continue to believe that setting a high episode payment ceiling at two standard deviations above the mean episode payment amount, along with the phasing in of responsibility for hospital repayment in performance year 1 as discussed in section III.C.2 of this final rule, will be sufficient to limit financial exposure due to outlier episodes. We will finalize our proposal to use a common outlier policy for all participant hospitals.

Comment: Many commenters requested that CMS risk adjust episode spending based on patients' hip fracture status, among other clinical and demographic dimensions.

Response: We refer readers to comments and responses to comments Start Printed Page 73337in section III.C.4.b.(1) of this final rule for further discussion on risk stratification for hip fracture status, and we reference it here because changes to risk stratification would impact how a high payment episode ceiling would function. As discussed in the responses to comments in section III.C.4.b.(1) of this final rule, we will modify our policy in this final rule so as to set different target prices both for episodes anchored by MS-DRG 469 vs. MS-DRG 470 and for episodes with hip fractures vs. without hip fractures. Given this change, we will also modify the proposed approach to apply the high payment episode ceiling. Specifically, instead of calculating and applying high payment episode ceilings for each region and anchor MS-DRG combination, we will now calculate and apply high payment episode ceilings for each region, anchor MS-DRG, and hip fracture status combination.

Final Decision: After consideration of the public comments we received, we are finalizing the proposal to apply high episode payment ceilings when calculating actual episode payments, setting episode target prices, and calculating reconciliation and repayment amounts. However, we do note that the approach to calculate and apply the high episode payment ceilings will be adapted to account for the risk stratification based on hip fracture status discussed in section III.C.4.b. of this final rule.

4. Episode Price Setting Methodology

a. Overview

Whether a participant hospital receives reconciliation payments or is made responsible to repay Medicare for the CJR model will depend on the hospital's quality and actual payment performance relative to episode quality and target prices. Quality performance and its tie to payments is further discussed in section III.C.5. of this final rule, and the remainder of this section will discuss the proposed approach to establishing target prices.

We proposed to establish CJR target prices for each participant hospital. For episodes beginning in performance years 1, 3, 4, and 5, a participant hospital would have eight target prices, one for each of the following:

  • MS-DRG 469 anchored episodes that were initiated between January 1 and September 30 of the performance year, if the participant hospital successfully submits data on the voluntary patient-reported outcome measure proposed in section III.C.5. of the proposed rule.
  • MS-DRG 470 anchored episodes that were initiated between January 1 and September 30 of the performance year, if the participant hospital successfully submits data on the proposed voluntary patient-reported outcome measure.
  • MS-DRG 469 anchored episodes that were initiated between October 1 and December 31 of the performance year, if the participant hospital successfully submits data on the proposed voluntary patient-reported outcome measure.
  • MS-DRG 470 anchored episodes that were initiated between October 1 and December 31 of the performance year, if the participant hospital successfully submits data on the proposed voluntary patient-reported outcome measure.
  • MS-DRG 469 anchored episodes that were initiated between January 1 and September 30 of the performance year, if the participant hospital does not successfully submit data on the voluntary patient-reported outcome measure.
  • MS-DRG 470 anchored episodes that were initiated between January 1 and September 30 of the performance year, if the participant hospital does not successfully submit data on the proposed voluntary patient-reported outcome measure.
  • MS-DRG 469 anchored episodes that were initiated between October 1 and December 31 of the performance year, if the participant hospital does not successfully submit data on the proposed voluntary patient-reported outcome measure.
  • MS-DRG 470 anchored episodes that were initiated between October 1 and December 31 of the performance year, if the participant hospital does not successfully submit data on the proposed voluntary patient-reported outcome measure.

For episodes beginning in performance year 2, a participant hospital would have 16 target prices. These would include the same combinations as for the other 4 performance years, but one set for determining potential reconciliation payments, and the other for determining potential Medicare repayment amounts, as part of the phasing in of two-sided risk discussed later in this section. Further discussion on our proposals for different target prices for MS-DRG 469 versus MS-DRG 470 anchored episodes, for episodes initiated between January 1 and September 30 versus October 1 and December 31, and for participant hospitals that do and do not successfully submit data on the proposed patient-reported outcome measure can be found in sections III.C.4.b. and III.C.5. of the proposed rule.

We intend to calculate and communicate episode target prices to participant hospitals prior to the performance period in which they apply (that is, prior to January 1, 2017, for target prices covering episodes initiated between January 1 and September 30, 2017; prior to October 1, 2017 for target prices covering episodes initiated between October 1 and December 31, 2017). We stated in the proposed rule our belief that prospectively communicating prices to hospitals will help them make any infrastructure, care coordination and delivery, and financial refinements they may deem appropriate to prepare for the new episode target prices.

The proposed approach to setting target prices incorporated the following features:

  • Set different target prices for episodes anchored by MS-DRG 469 versus MS-DRG 470 to account for patient and clinical variations that impact hospitals' cost of providing care.
  • Use 3 years of historical Medicare payment data grouped into episodes of care according to the episode definition in section III.B. of the proposed rule, hereinafter termed historical CJR episodes. The specific set of 3-historical-years used would be updated every other performance year.
  • Apply Medicare payment system (for example, IPPS, OPPS, IRF PPS, SNF, MPFS, etc.) updates to the historical episode data to ensure we incentivize hospitals based on historical utilization and practice patterns, not Medicare payment system rate changes that are beyond hospitals' control. Because different Medicare payment system updates become effective at two different times of the year, we would calculate separate target prices for episodes initiated between January 1 and September 30 versus October 1 and December 31.
  • Blend together hospital-specific and regional historical CJR episode payments, transitioning from primarily provider-specific to completely regional pricing over the course of the 5 performance years, to incentivize both historically efficient and less efficient hospitals to furnish high quality, efficient care in all years of the model. Regions would be defined as each of the nine U.S. Census divisions.
  • Normalize for provider-specific wage adjustment variations in Medicare payment systems when combining provider-specific and regional historical CJR episodes. Wage adjustments would Start Printed Page 73338be reapplied when determining hospital-specific target prices.
  • Pool together CJR episodes anchored by MS DRGs 469 and 470 to use a greater historical CJR episode volume and set more stable prices.
  • Apply a discount factor to serve as Medicare's portion of reduced expenditures from the CJR episode, with any remaining portion of reduced Medicare spending below the target price potentially available as reconciliation payments to the participant hospital where the anchor hospitalization occurred.

Further discussion on each of the individual features can be found in section III.C.4.b. of this final rule. In section III.C.4.c. of this final rule, we also provide further details on the proposed sequential steps to calculate target prices and how each of the pricing features would fit together.

The following is a summary of the comments received and our responses.

Comment: Commenters responded on several of the proposed pricing features, including how quality performance would affect payment, and we refer readers to comments and responses to comments in sections III.C.4.b and III.C.5 for further discussion on changes to how quality would be tied to payment as described in the proposed rule. We reference these comments here because any changes to the proposed episode price setting methodology and link between quality performance and payment would impact the number of target prices for each participant hospital.

Response: As further discussed in section III.C.4.b.(1) of this final rule, we are modifying the proposed rule to risk stratify (and set different prices) based on not just different anchor MS-DRGs but also patients' hip fracture status. As discussed in section III.C.4.b.(9) of this final rule, we are modifying our policy in this final rule so as to use lower discount factors for purposes of determining the hospital's responsibility for excess episode spending not only in performance year 2, but also in performance year 3. Additionally, as discussed in section III.C.5 of this final rule, we are modifying the proposed rule so as to provide different levels of quality incentive payments that would modulate participant hospitals' effective target price discount factor based on their quality performance. Because of these changes, each participant hospital in performance years 1, 4, and 5 will have 8 potential target prices for each combination of anchor MS-DRG (469 vs. 470), hip fracture status (with hip fracture vs. no hip fracture), and episode initiation date (between April 1 and September 30 vs. between October 1 and December 31 for performance year 1, and between January 1 and September 30 vs. between October 1 and December 31 for performance years 2 through 5). Each participant hospital in performance years 2 and 3 will have 16 target prices for the same combinations in performance years 1, 4, and 5, but with one group of 8 potential target prices for purposes of calculating reconciliation payments and another group of 8 potential target prices for purposes of determining hospital's responsibility for excess episode spending.

b. Pricing Features

(1) Different Target Prices for Episodes Anchored by MS-DRG 469 Versus MS-DRG 470

For each participant hospital we proposed to establish different target prices for CJR episodes initiated by MS-DRG 469 versus MS-DRG 470. MS-DRGs under the IPPS account for some of the clinical and resource variations that exist and that impact hospitals' cost of providing care. Specifically, MS-DRG 469 is defined to identify, and provide hospitals a higher Medicare payment to reflect the higher hospital costs for, hip and knee procedures with major complications or comorbidities. Therefore, we proposed to risk stratify and calculate separate target prices for each participant hospital for CJR episodes with MS-DRG 469 versus MS-DRG 470 anchor hospitalizations.

We considered risk adjusting the episode target prices by making adjustments or setting different prices based on patient-specific clinical indicators (for example, comorbidities). However, we did not believe there is a sufficiently reliable approach that exists suitable for CJR episodes beyond MS-DRG-specific pricing, and there is no current standard on the best approach. At the time of developing the proposed rule Tennessee, Ohio, and Arkansas are launching multi-payer (including Medicaid and commercial payers, excluding Medicare) bundles and include hip and knee replacement as an episode.[22 23 24] These states' hip and knee episode definitions and payment models are consistent with, though not the same as, the proposed CJR episode described in the proposed rule. However, each of these three states uses different risk adjustment factors. This variation across states supported our stated belief in the proposed rule that there is currently no standard risk adjustment approach widely accepted throughout the nation that could be used under CJR, a model that would apply to hospitals across multiple states. Therefore, we did not propose to make risk adjustments based on patient-specific clinical indicators.

We also considered making risk adjustments based on the participant hospital's average Hierarchical Condition Category (HCC) score for patients with anchor CJR hospitalizations. The CMS-HCC risk adjustment model quantifies a beneficiary's risk by examining the beneficiary's demographics and historical claims data and predicting the beneficiary's total expenditures for Medicare Parts A and B in an upcoming year. However, the CMS-HCC risk adjustment model's intended use is to pay Medicare Advantage (MA) plans appropriately for their expected relative costs. For example, MA plans that disproportionately enroll the healthy are paid less than they would have been if they had enrolled beneficiaries with the average risk profile, while MA plans that care for the sickest patients are paid proportionately more than if they had enrolled beneficiaries with the average risk profile. The CMS-HCC risk adjustment model is prospective. It uses demographic information (that is, age, sex, Medicare/Medicaid dual eligibility, disability status) and a profile of major medical conditions in the base year to predict Medicare expenditures in the next year.[25] As previously noted, the CMS-HCC risk adjustment model is used to predict total Medicare expenditures in an upcoming year, and may not be appropriate for use in predicting expenditures over a shorter period of time, such as the CJR episode, and may not be appropriate in instances where its use is focused on LEJRs. Therefore, since we have not evaluated the validity of HCC scores for predicting Medicare expenditures for shorter episodes of care or for specifically LEJR beneficiaries, we did not propose to risk Start Printed Page 73339adjust the target prices using HCC scores for the CJR model.

We also considered risk stratifying or setting different prices for different procedures, such as different prices for hip versus knee replacements, but we did not believe there would be substantial variation in episode payments for these clinical scenarios to warrant different prices or adjustments. Moreover, Medicare IPPS payments, which account for approximately 50 percent [26] of CJR episode expenditures, do not differentiate between hip and knee procedures, mitigating procedure-specific variation for the anchor hospitalization. Furthermore, there are no widely accepted clinical guidelines to suggest that PAC intensity would vary significantly between knee and hip replacements. We sought comment on our proposal to price episodes based on the MS-DRG for the anchor hospitalization, without further risk adjustment.

The following is a summary of the comments received and our responses.

Comment: Many commenters stated that proper risk adjustment is necessary for the success of this model, and that anchor MS-DRG-specific pricing can help but is not sufficient on its own. Proper risk adjustment would account for differences in episode spend due to patient variations that are out of providers' control. They stated that MS-DRGs may capture variations within the inpatient setting, but do not reflect patient variations post-discharge. Inappropriate risk adjustment could lead to access issues for higher risk patients and increased volume of LEJR procedures for younger/healthier patients by participant hospitals looking to lower their average episode expenditures.

Most commenters who wrote on the issue suggested risk adjustment or complete exclusion for episodes with hip fractures, partial hip replacements, and emergent (versus non-emergent or elective) procedures. Some commenters provided analysis on hip fractures, in particular, and demonstrated episodes with hip fractures are significantly more expensive than those without hip fractures. Other clinical and demographic dimensions offered for risk adjustment or exclusion include the following: Procedure (total hip [THA] vs. total knee [TKA] vs. partial hip [PHA] vs. ankle vs. limb reattachment); socioeconomic status; patient functional status; age; and comorbidities. Requests from commenters for risk adjustment based on the previously stated dimensions were usually paired with requests to also exclude patients from the CJR model, and we encourage readers to read comments in section III.B.2.a. of this final rule for additional details on the clinical and demographic dimensions requested for risk adjustment or exclusion.

Some commenters who wrote on the issue of risk adjustment disputed CMS' statement in the proposed rule that there is no standard risk adjustment approach widely accepted throughout the nation. They pointed to examples of existing risk adjustment approaches that could be used for CJR episodes, such as Optum's Procedure Episode Grouper (PEG), Truven's Medical Episode Grouper (MEG), Health Care Incentives Improvement Institute's (HCI3) risk adjustment model, CMS's HCCs model, and CMS's risk-adjusted quality/efficiency metric for elective LEJR episodes: Hospital-Level, Risk-Standardized Payment Associated with a 90-Day Episode of Care for Elective Primary Total Hip Arthroplasty (THA) and/or Total Knee Arthroplasty (TKA).

Response: In response to comments, we undertook further analysis. Our analysis showed that episodes with hip fractures, identified by historical anchor hospitalization claims with an ICD-9-CM hip fracture code as the principal diagnosis, have approximately 70 percent greater historical average episode expenditures than episodes without hip fractures, even for episodes within the same anchor MS-DRG, confirming analyses shared by some commenters that also showed episodes with hip fractures to have significantly greater average expenditures.[27] PHA episodes and emergent episodes had similarly higher historical average expenditures than TKA and THA episodes and non-emergent episodes, respectively. There are clearly patient-specific conditions that lead to significant episode expenditure variations, even within the same MS-DRG.

On the basis of the comments and our further analysis, we agree with commenters that proper risk adjustment is necessary to appropriately incentivize participant hospitals to deliver high quality and efficient care. We acknowledge that a comprehensive risk adjustment methodology beyond just setting different prices by anchor MS-DRGs could more accurately risk adjust episodes for patient-specific clinical and demographic factors that would drive variations in CJR episode expenditures.

We disagree with commenters, though, that there is an already existing, widely accepted risk adjustment methodology for CJR episodes. The HCC model, as discussed earlier in this section, is not designed to predict costs within CJR episodes and may not accurately predict CJR episode expenditures. Commercial claims groupers such as Optum's PEG, Truven's MEG, and HCI3's risk adjustment model utilize different episode definitions from how we will define CJR episodes. Additionally, these commercial groupers have yet to be validated for a Medicare population; we believe there may be a different set of risk factors that predict episode expenditures for Medicare beneficiaries than those used to predict episode expenditures for younger and generally healthier individuals with commercial insurance. We also acknowledge that CMS has designed a risk-adjusted quality/efficiency metric for elective LEJR episodes: Hospital-Level, Risk-Standardized Payment Associated with a 90-Day Episode of Care for Elective Primary Total Hip Arthroplasty (THA) and/or Total Knee Arthroplasty (TKA). This metric, though, has been developed for a different episode definition; most notably, this risk-adjusted metric excludes emergent episodes while the CJR episode definition does not exclude emergent episodes, as discussed in section III.B. of this final rule.

We do believe that there are opportunities to learn from existing comprehensive risk adjustment models, and we may explore how a comprehensive risk adjustment model such as these may be adapted for the CJR model in the future.

In the meantime, though, we also believe we can improve upon the proposed approach of only setting different target prices by anchor MS-DRG. Specifically, we can account for the impact of hip fracture status (with hip fracture vs. without hip fracture), procedure choice (PHA vs. TKA or THA), and emergence status (emergent vs. non-emergent) on episode expenditures. According to our analysis, though, there was significant correlation between incidence of hip fractures, partial hip procedures, and emergent procedures—94 percent of partial hip replacement episodes and 93 percent of emergent episodes are for patients with hip fractures. Because of the correlation between these three factors, we believe we can account for all three by risk stratifying based on hip fracture status alone. We believe hip fracture status is a more appropriate dimension on which Start Printed Page 73340to risk stratify because it reflects patients' clinical status, as opposed to partial hip replacements and emergent procedures which are influenced by providers' care delivery decisions.

In light of the comments and our additional analysis, we will modify our proposed policy to risk stratify, or set different target prices, both for episodes anchored by MS-DRG 469 vs. MS-DRG 470 and for episodes with hip fractures vs. without hip fractures. By adding hip fracture status to our risk stratification approach, we believe we can capture a significant amount of patient-driven episode expenditure variation. Additionally, because of the high correlation between incidence of hip fractures, partial hip procedures, and emergent procedures, we do not believe we need to add any procedure-specific and emergent status factors for risk stratification. We still believe, as stated in the proposed rule that PAC intensity would not vary significantly between TKA and THA for beneficiaries without hip fractures.

We will identify episodes with hip fractures using ICD-9-CM or ICD-10-CM diagnosis codes, where the hip fracture diagnosis is the principal diagnosis on the anchor hospitalization claim for an LEJR procedure. Our goal is to identify those CJR episodes where the primary surgical treatment for the hip fracture is an LEJR procedure furnished during the anchor hospitalization. The historical episodes with hip fracture diagnosis codes on the anchor hospitalization claim will be used to set the hip fracture episode target prices under the CJR model, and episodes during the CJR model with hip fracture diagnosis codes on the anchor hospitalization claim will be reconciled at the hip fracture episode target prices.

In order to develop the initial list of ICD-9-CM hip fracture diagnosis codes used to identify those historical episodes with hip fracture for calculating hip fracture episode target prices, to implement changes to the list to account for the transition to the ICD-10-CM diagnosis code set that will be used to identify episodes during the model performance years that will receive fracture episode target prices, and to make other changes as necessary based on annual ICD-10-CM coding changes or to address issues raised by the public throughout the model performance years, we are implementing the following subregulatory process, which mirrors the subregulatory process we will use for the episode definition exclusions list described in section III.B.2 of this final rule. We will use this process on an annual, or more frequent, basis to update the ICD-CM hip fracture diagnosis code list and to address issues raised by the public.

As part of this process, we will first develop the potential ICD-CM hip fracture diagnosis codes based on our assessment according to the following standards:

  • The ICD-CM diagnosis code is sufficiently specific that it represents a bone fracture for which a physician could determine that a hip replacement procedure, either a PHA or a THA, could be the primary surgical treatment.
  • The ICD-CM diagnosis code is the primary reason (that is, principal diagnosis code) for the anchor hospitalization.

We will then post a list of potential hip fracture diagnosis codes (whether ICD-9-CM diagnosis codes, as necessary to develop initial target prices, or ICD-10-CM diagnosis codes to be utilized during the model performance years) to the CMS Web site at http://innovation.cms.gov/​initiatives/​cjr/​ to allow for public input on our planned application of these standards, and then we will adopt the ICD-CM hip fracture diagnosis code list with posting to the CMS Web site of the final ICD-CM hip fracture diagnosis code list after our consideration of the public input.

With public release of this final rule, we are initiating this subregulatory process to develop a final ICD-9-CM hip fracture diagnosis code list that will be used to identify historical anchor hospitalizations for beneficiaries with hip fracture for purposes of determining episode spending in the historical period and developing initial target prices for the model. The potential ICD-9-CM hip fracture diagnosis code list is posted on the CJR Web site at http://innovation.cms.gov/​initiatives/​cjr/​. Given our objective to quickly develop target prices and provide them to participant hospitals in the timeframe described in section III.C.4. of this final rule, we will allow for public input on this list for 14 days after the public release of this final rule. Public comments will be submitted via an email address posted on the CJR Web site along with the list of potential ICD-9-CM hip fracture diagnosis codes previously referenced. We will consider the public's input and then, after consideration, we will post the final ICD-9-CM hip fracture diagnosis code list to the CMS Web site. This list will be used to calculate the first set of target prices communicated to participant hospitals. Within 30 days of public release of the final rule, we will again initiate this subregulatory process to identify ICD-10-CM hip fracture diagnosis codes by posting the potential ICD-10-CM hip fracture diagnosis code list on the CMS Web site and seeking public input, so we can provide in a timely manner the final list of ICD-10-CM hip fracture diagnosis codes prior the beginning of the first model performance year.

Final Decision: After consideration of the public comments we received, we are modifying the proposed rule to risk stratify (and set different target prices) based on not just different anchor MS-DRGs but also patients' hip fracture status. We will identify episodes with hip fractures using ICD-9-CM or ICD-10-CM diagnosis codes in the principal position on the claim for the anchor hospitalization. We are instituting a subregulatory process in order to allow for public comment and to finalize the ICD-9-CM and ICD-10-CM diagnosis codes to be used in identifying hip fracture cases in the CJR model, which we are initiating as of the public release of this final rule. We refer readers to the list of ICD-9-CM diagnosis codes posted on the CJR model Web site at http://innovation.cms.gov/​initiatives/​cjr/​.

This policy is codified at § 510.300(a).

(2) Three Years of Historical Data

We proposed to use 3 years of historical CJR episodes for calculating CJR target prices. The set of 3-historical-years used would be updated every other year. Specifically—

  • Performance years 1 and 2 would use historical CJR episodes that started between January 1, 2012 and December 31, 2014;
  • Performance years 3 and 4 would use historical episodes that started between January 1, 2014 and December 31, 2016; and
  • Performance year 5 would use episodes that started between January 1, 2016 and December 31, 2018.

We considered using fewer than 3 years of historical CJR episode data, but we are concerned with having sufficient historical episode volume to reliably calculate target prices. We also considered not updating the historical episode data for the duration of the model. However, we stated in the proposed rule our belief that hospitals' target prices should be regularly updated on a predictable basis to use the most recent available claims data, consistent with the regular updates to Medicare's payment systems, to account for actual changes in utilization. We are not proposing to update the data annually, given the uncertainty in pricing this could introduce for participant hospitals. We also note that the effects of updating hospital-specific data on the target price could be limited Start Printed Page 73341as the regional contribution to the target price grows, moving to two-thirds in performance year 3 when the first historical episode data update would occur.

The following is a summary of the comments received and our responses.

Comment: Commenters generally supported using historical expenditures to set target prices. Several commenters expressed concern that updating the 3 years of historical CJR episode data every other year would effectively make participant hospitals compete against themselves without consideration of whether they are already efficient. Some of these commenters cited that BPCI does not update its historical data for the entirety of the BPCI model, and some other commenters noted that Medicare Shared Savings Program resets its historical benchmark every three years with each new participation agreement. There were also a few commenters that supported updating the 3 years of historical CJR episode data every 2 years because it was better than doing so every year.

Some commenters also stated that if we do update the historical data, we should include previous reconciliation payments and repayments to Medicare for the participant hospitals. We refer readers to comments and responses to comments in section III.C.3 of this final rule for further discussion on this comment.

Some commenters proposed alternative approaches to getting to target prices other than using and updating historical data. Some commenters suggested using a negotiations/bidding process approach to get to target prices; Medicare would negotiate with or request bids from providers for providing services covered under the CJR episode definition. Some other commenters suggested applying some sort of inflation factor, such as a CMS market basket update, for future years of the model instead of updating the 3 years of historical CJR episode data. These alternatives to using and updating the historical CJR episode data would help prevent a participant hospital from competing against its historical self, even if it is already efficient, in order to qualify for reconciliation payments.

Response: We appreciate commenters' support for using historical expenditures to set target prices. We acknowledge that BPCI does not update participants' historical data and Medicare Shared Savings Program does not reset participating entities' benchmark for 3 years (until the beginning of a new agreement period). However, these programs employ alternative mechanisms to account for recent national trends reflecting changes in industry wide practice patterns. BPCI, for example, retrospectively applies a national trend factor to trend forward historical episode expenditure data and capture changes in nationwide practice patterns between the time period used in the historical data and the performance period. BPCI participants are not penalized or rewarded for mirroring nationwide practice pattern trends. In BPCI, however, participants' target prices are determined retrospectively after the close of each performance period. We intend to calculate and communicate target prices prior to the start of each performance year, as discussed in section III.C.4.a of this final rule, so we cannot utilize the retrospective national trend factor approach as used in BPCI.

Instead, we proposed to capture changes in nationwide practice patterns by updating every other year the historical CJR episode data used to set target prices. We recognize that this approach of updating the historical episode data every other year effectively assumes a zero percent change in utilization between the latest year of historical episode data and the performance year. We believe this can be a valid estimate for a few years (for example, 2014 as the latest year of historical episode data for 2017 target prices; update historical episode data for 2018 target prices), but it is less likely to hold true for longer periods of time (for example, 2014 as the latest year of historical episode data for 2020 target prices; no update to historical episode data). Therefore, we believe updating the historical episode data is necessary. While updating the historical episode data more frequently (that is, every year, instead of every other year) would lessen our reliance on an assumption of zero percent utilization change, doing so may exacerbate commenters' concerns that already efficient hospitals would have to compete against themselves, as discussed further later in this section.

We appreciate commenters' concerns that it may be unsustainable for already efficient participant hospitals to continuously improve, and that participant hospitals may undertake activities that promote care coordination and improved quality of care but are not directly reimbursed under applicable Medicare FFS payment systems. If we were using 100 percent hospital-specific pricing, updating the historical data used to set target prices without including reconciliation payments would create a lower and harder to achieve target price for participant hospitals that previously increased efficiency. As discussed in section III.C.3 of this final rule, we may revisit in future rulemaking our decision to exclude reconciliation payments and repayment amounts when updating the set of historical years used to set target prices. Additionally, as we transition to regional pricing over the course of the model, participant hospitals will no longer compete against their historical selves but rather strive to outperform their regional peers. Under regional pricing, an already efficient hospital may be able to achieve actual episode expenditures below the regional target price without having to become even more efficient. By performance year 3, when the first update to historical episode data would occur, the majority of the target price would be based on the regional component, not the hospital-specific component, as described in section III.C.4.b.(5) of this final rule.

We appreciate commenters' suggestions on using alternative approaches to setting target prices. We may consider such approaches for future model tests.

Final Decision: After consideration of the public comments we received, we are finalizing the proposal, without modification, to use three years of historical expenditures, updated every other year, to set target prices.

(3) Trending of Historical Data to the Most Recent Year of the Three

We acknowledge that some payment variation may exist in the 3 years of historical CJR episodes due to updates to Medicare payment systems (for example, IPPS, OPPS, IRF PPS, SNF PPS, etc.) and national changes in utilization patterns. Episodes in the third of the 3-historical-years may have higher average payments than those from the earlier 2 years because of Medicare payment rate increases over the course of the 3-historical-years. We do not intend to have CJR incentives be affected by Medicare payment system rate changes that are beyond hospitals' control. In addition to the changes in Medicare payment systems, average episode payments may change year over year due to national trends reflecting changes in industry-wide practice patterns. For example, readmissions for all patients, including those in CJR episodes, may decrease nationally due to improved industry-wide surgical protocols that reduce the chance of infections. We do not intend to provide reconciliation payments to (or require repayments from) hospitals for achieving lower (or higher) Medicare expenditures solely because they followed national changes in practice Start Printed Page 73342patterns. Instead, we aim to incentivize hospitals based on their hospital-specific inpatient and PAC delivery practices for LEJR episodes.

To mitigate the effects of Medicare payment system updates and changes in national utilization practice patterns within the 3 years of historical CJR episodes, we proposed to apply a national trend factor to each of the years of historical episode payments. Specifically, we proposed to inflate the 2 oldest years of historical episode payments to the most recent year of the 3-historical-years described in section III.C.4.b.(2) of the proposed rule. We proposed to trend forward each of the 2 oldest years using the changes in the national average CJR episode payments. We also proposed to apply separate national trend factors for episodes anchored by MS-DRG 469 versus MS-DRG 470 to capture any MS-DRG-specific payment system updates or national utilization pattern changes. For example, when using CY 2012-2014 historical episode data to establish target prices for performance years 1 and 2, under our proposal we would calculate a national average MS-DRG 470 anchored episode payment for each of the 3-historical-years. The ratio of the national average MS-DRG 470 anchored episode payment for CY 2014 to that of CY 2012 would be used to trend 2012 MS-DRG 470 anchored episode payments to CY 2014. Similarly, the ratio of the national average MS-DRG 470 anchored episode payment for CY 2014 to that of CY 2013 would be used to trend 2013 episode payments to CY 2014. The previously stated process would be repeated for MS-DRG 469 anchored episodes. Trending CY 2012 and CY 2013 data to CY 2014 would capture updates in Medicare payment systems as well as national utilization pattern changes that may have occurred.

We considered adjusting for regional trends in utilization, as opposed to national trends. However, we stated in the proposed rule our belief that any Medicare payment system updates and significant changes in utilization practice patterns would not be region-specific but rather be reflected nationally.

We sought comment on our proposal to nationally trend historical data to the most recent year of the 3 being used to set the target prices.

The following is a summary of the comments received and our responses.

Comment: Some commenters supported the use of national trends for trending historical data to the most recent of the 3 being used to set the target prices. Some commenters suggested blending regional, instead of national, trends to be consistent with how target prices will be blended, as discussed in section III.C.4.b.(5) of this final rule. Some commenters inquired how trending historical data would capture changes in Medicare FFS fee schedules.

Response: We appreciate commenters' support for the use of national trends for trending historical data. This trending of historical data to the most recent of the 3 being used to set target prices would capture both Medicare FFS fee schedule and practice pattern changes. Medicare FFS fee schedule changes would be captured in the trend factor calculations; for example, if Medicare FFS fee schedules change so as to increase overall payments by 4 percent between the oldest and most recent year of historical episode data, the national trend factor applied to the oldest year of historical episode data would be 1.04 (assuming no change in utilization patterns). Medicare FFS fee schedule changes apply across the nation, and we believe that major changes to practice patterns would be nationwide and not constrained to any one region.

Comment: Many commenters requested for risk adjustment based on patients' hip fracture status, among other clinical and demographic dimensions.

Response: We refer readers to comments and responses to comments in section III.C.4.b.(1) of this final rule for further discussion on risk stratification, and we reference it here because changes to risk stratification would impact how we would trend historical data to the most recent year of the three being used. As discussed in the responses to comments in section III.C.4.b.(1) of this final rule, we will modify our proposal so as to set different target prices both for episodes anchored by MS-DRG 469 vs. MS-DRG 470 and for episodes with hip fractures vs. without hip fractures. Given this change, we must also modify the proposed approach to apply national trend factor. Specifically, instead of calculating different national trend factors just for anchor MS-DRGs 469 vs. 470, we will calculate different national trend factors for each combination of anchor MS-DRG (469 vs. 470) and hip fracture status (with hip fracture vs. without hip fracture) using the methodology we proposed.

Final Decision: After consideration of the public comments we received, we are finalizing the proposal to trend historical data to the most recent of the 3 being used to set target prices, though instead of calculating different national trend factors just for anchor MS-DRGs 469 vs. 470, we will calculate different national trend factors for each combination of anchor MS-DRG (469 vs. 470) and hip fracture status (with hip fracture vs. without hip fracture).

(4) Update Historical Episode Payments for Ongoing Payment System Updates

We proposed to prospectively update historical CJR episode payments to account for ongoing Medicare payment system (for example, IPPS, OPPS, IRF PPS, SNF, MPFS, etc.) updates to the historical episode data and ensure we incentivize hospitals based on historical utilization and practice patterns, not Medicare payment system rate changes that are beyond hospitals' control. Medicare payment systems do not update their rates at the same time during the year. For example, IPPS, the IRF PPS, and the SNF payment system apply annual updates to their rates effective October 1, while the hospital OPPS) and MPFS apply annual updates effective January 1. To ensure we appropriately account for the different Medicare payment system updates that go into effect on January 1 and October 1, we proposed to update historical episode payments for Medicare payment system updates and calculate target prices separately for episodes initiated between January 1 and September 30 versus October 1 and December 31 of each performance year. The target price in effect as of the day the episode is initiated would be the target price for the whole episode. Note that in performance year 5, the second set of target prices would be for episodes that start and end between and including October 1 and December 31 because the fifth performance period of the CJR model would end on December 31, 2020. Additionally, a target price for a given performance year may apply to episodes included in another performance year. For example, an episode initiated in November 2016, and ending in February 2017 would have a target price based on the second set of 2016 target prices (for episodes initiated between October 1 and December 31, 2016), and it would be captured in the CY 2017 performance year (performance year 2) because it ended between January 1 and December 31, 2017. We refer readers to section III.C.3.c. of the proposed rule for further discussion on the definition of performance years.

We proposed to update historical CJR episode payments by applying separate Medicare payment system update factors each January 1 and October 1 to each of the following six components of each hospital's historical CJR payments:

  • Inpatient acute.
  • Physician.Start Printed Page 73343
  • IRF.
  • SNF.
  • HHA.
  • Other services.

A different set of update factors would be calculated for January 1 through September 30 versus October 1 through December 31 episodes each performance year. The six update factors for each of the previously stated components would be hospital-specific and would be weighted by the percent of the Medicare payment for which each of the six components accounts in the hospital's historical episodes. The weighted update factors would be applied to historical hospital-specific average payments to incorporate ongoing Medicare payment system updates. A weighted update factor would be calculated by multiplying the component-specific update factor by the percent of the hospital's historical episode payments the component represents, and summing together the results. For example, let us assume 50 percent of a hospital's historical episode payments were for inpatient acute care services, 15 percent for physician services, 35 percent for SNF services, and 0.0 percent for the remaining services. Let us also assume for this example that the update factors for inpatient acute care services, physician services, and SNF services are 1.02, 1.03, and 1.01, respectively. The weighted update factor in this example would be the following: (0.5 * 1.02) + (0.15*1.03) + (0.35*1.01) = 1.018. The hospital in this example would have its historical average episode payments multiplied by 1.018 to incorporate ongoing payment system updates. The specific order of steps, and how this step fits in with others, is discussed further in section III.C.4.c. of this final rule.

Each of a hospital's six update factors would be based on how inputs have changed in the various Medicare payment systems for the specific hospital. Additional details on these update factors will be discussed later in this section.

Region-specific update factors for each of the previously stated components and weighted update factors would also be calculated in the same manner as the hospital-specific update factors. Instead of using historical episodes attributed to a specific hospital, region-specific update factors would be based on all historical episodes initiated at any CJR eligible hospital within the region. For purposes of this rule, CJR eligible hospitals are defined as hospitals that are paid under IPPS and not a participant in BPCI Model 1 or in the risk-bearing period of Models 2 or 4 for LEJR episodes, regardless of whether or not the MSAs in which the hospitals are located were selected for inclusion in the CJR model. CJR episodes initiated at a CJR eligible hospital will for purposes of this rule be referred to as CJR episodes attributed to that CJR eligible hospital.

We considered an alternative option of trending the historical episode payments forward to the upcoming performance year using ratios of national average episode payment amounts, similar to how we proposed to trend the 2 oldest historical years forward to the latest historical year for historical CJR episode payments in section III.C.4.b.(3) of the proposed rule. Using ratios of national average episode payment amounts would have the advantage of also capturing changes in national utilization patterns in addition to payment system updates between the historical years and the performance year. However, such an approach would need to be done retrospectively, after average episode payments can be calculated for the performance year, because it would rely on the payments actually incurred in the performance period, data that would be not be available before the performance period. While the proposed approach of using component-specific weighted update factors may be more complicated than the previously stated alternative to use ratios of national average episode payment amounts, we stated in the proposed rule our belief that the additional complication is outweighed by the value to hospitals of knowing target prices before the start of an episode for which the target price would apply. We sought comment on this proposed approach of updating historical episode payments for ongoing Medicare payment system changes.

We did not propose to separately and prospectively apply an adjustment to account for changes in national utilization patterns between the historical and performance years. If a prospective adjustment factor for national utilization pattern changes were applied, it may only be meaningful in performance years 2 and 4, when the historical data used to calculate target prices would not be updated, but another year of historical data would be available. In any of the other 3 performance years, the latest available historical year of data would already be incorporated into the target prices. Given that we proposed to refresh the historical data used to calculate target prices every 2 years, we did not believe an additional adjustment factor to account for national practice pattern changes is necessary to appropriately incentivize participant hospitals to improve quality of care and reduce episode payments.

The following is a summary of the comments received and our responses.

Comment: Several commenters noted that the Medicare payment system update factors were complicated to calculate. Some commenters supported the use of calculating Medicare payment system update factors at the hospital-specific and regional levels to reflect practice pattern variations, while some others proposed using national update factors to incentivize reduction in medically unnecessary and/or inappropriate practice pattern variations.

A couple of commenters also inquired whether the Medicare payment system update factors accounted for changes Medicare FFS payment system changes. A commenter requested we freeze MS-DRG weights for MS-DRGs 469 and 470 if the weights decrease in any given year as part of the annual Medicare FFS IPPS payment system updates.

Response: We acknowledge that the Medicare payment system update factor calculations are complex, but we believe the complexity is necessary to account for Medicare FFS payment system changes. We will use these payment system update factors to ensure that we incentivize hospitals based on utilization and practice patterns, not Medicare payment system rate changes. While changes to Medicare FFS rates for individual services would be applicable nationwide, the relative composition of each service in historical episodes will likely vary by hospital and region. Calculating payment system update factors at the hospital-specific and regional levels will more accurately capture the effects of payment system changes.

We also note that we are finalizing a modification to the equations used to calculate update factors for those payment systems that apply annual updates to their rates effective October 1 of each year. In lieu of calculating the update factors for inpatient acute, SNF, and IRF services using the values applicable at the end of latest historical year used to calculate target prices, we will use a blend of the values applicable during the latest historical year. Such a change will account for the payment systems that update payment rates on a fiscal year cycle, ensure we are calculating update factors based on the payment rates that apply to a given period to the extent feasible, and result in more accurate target price calculations. We reflect this change in the sections III.C.4.b.(4)(a), Start Printed Page 73344III.C.4.b.(4)(c), and III.C.4.b.(4)(d) of this final rule

We believe freezing MS-DRG weights would run counter to our objective to accurately account for the effects of Medicare FFS payment system changes. If we freeze MS-DRG weights and the weights decrease, we may inappropriately overestimate target prices.

Comment: Some commenters requested to have a single set of target prices for the entire calendar year, as opposed to two different sets of target prices that would account for intra-year Medicare FFS payment systems updates: one set for January 1 through September 30, and a second set for October 1 through December 31. These commenters stated that a single target price for the entire year may be easier to communicate to participant hospitals, and that the effect of mid-calendar year changes in Medicare FFS (for example, October 1 IPPS changes) could be estimated and reconciled against a single set of target prices for the entire calendar year.

Response: We appreciate commenters' desire for simplicity. However, we would not know the extent of October 1 Medicare FFS payment system updates prior to January of the same year. Additionally, the October update includes payment system updates for IPPS, which accounts for the plurality of historical CJR episode expenditures. Without knowing the magnitude of Medicare FFS payment system updates, we do not believe we could reliably calculate target prices. Any estimate would likely require corrections after the end of the performance year, rendering the initial target price unreliable and unrepresentative of the target price used for reconciliation.

Comment: Several commenters recommended that we modify the definition of `CJR eligible hospitals,' the term used to identify hospitals included in calculations for the regional component of target prices (discussed further in section III.C.4.b.(5) of this final rule), to not exclude hospitals that are participants in BPCI Model 1 or in the risk bearing period of Models 2 or 4 for LEJR episodes. They recommended that some regions may have a greater proportion of these BPCI participants, and excluding them from the calculations for the regional component of target prices would not accurately reflect the region's historical expenditures. Additionally, with fewer hospitals included, the region component of target prices would be more significantly impacted by the performance of just CJR participant hospitals.

Response: We agree with commenters' arguments to include hospitals that are participants in BPCI Model 1 or in the risk bearing period of Models 2 or 4 for LEJR episodes when calculating the regional component of CJR target prices. Including these BPCI hospitals would more accurately reflect the region's historical expenditures, independent of the level of BPCI participation in the region. Therefore, we are not finalizing our proposal to exclude these hospitals from the regional calculation. We will modify the definition of “CJR eligible hospitals” to include these BPCI hospitals so that their data is included in the regional component of target prices. We will treat these BPCI participants as though they were any other non-BPCI-participating hospital—we would not apply the BPCI discount factor to claims payments nor include BPCI reconciliation or repayments for these BPCI hospitals. We do not intend to reduce target prices for participant hospitals just because they are located in a region with greater BPCI participation; instead, we want to ensure that we are calculating a representative regional component for target prices. In order to reduce potential confusion, we will also rename “CJR eligible hospitals” to be “CJR regional hospitals”.

We also clarify that BPCI LEJR episodes will be included in the historical data used to calculate the hospital-specific component of target prices. There may be some CJR participant hospitals who were previously participants in BPCI Model 2; there may be some BPCI Model 2 episodes in the historical data initiated by PGPs for which the LEJR procedure took place at the CJR participant hospital; or there may be some BPCI Model 3 episodes in the historical data for which the LEJR procedure took place at the CJR participant hospital. Including the BPCI LEJR episodes from the historical data used to calculate the hospital-specific component of target prices would parallel the previously discussed approach to include BPCI LEJR episodes in the regional component of target prices. Again, as previously discussed for the regional component of target prices, we would not apply the BPCI discount factor to claim payments nor include BPCI reconciliation or repayments for the hospital-specific component of target prices.

Final Decision: After consideration of the public comments we received, we are modifying our proposal to update historical episode payments for ongoing payment system updates so as to include in the definition of “CJR eligible hospitals” that are participants in BPCI Model 1 or in the risk bearing period of Models 2 or 4 for LEJR episodes, and rename “CJR eligible hospitals” to be “CJR regional hospitals.” We are also finalizing a modification to how we calculate update factors to more accurately capture payment system rate changes throughout the calendar year for inpatient acute, IRF, and SNF services. The modification is reflected in III.C.4.b.(4)(a), III.C.4.b.(4)(c), and III.C.4.b.(4)(d) of this final rule.

(a) Inpatient Acute Services Update Factor

The proposed inpatient acute services update factor would apply to payments for services included in the episode paid under the IPPS. This would include payments for the CJR anchor hospitalization and related readmissions at hospitals paid under IPPS, but not payments for related readmissions at CAHs during the episode window. Payments for related readmissions at CAHs would be captured under the update factor for other services in section III.C.4.b.(4)(f) of the proposed rule.

The update factor applied to the inpatient acute services component of each participant hospital and region's historical average episode payments would be based on how inputs for the Medicare IPPS have changed between the latest year used in the historical 3 years of episodes and the upcoming performance period under CJR. We proposed to use changes in the following IPPS inputs to calculate the inpatient acute services update factor: IPPS base rate and average of MS-DRG weights, as defined in the IPPS/LTCH Final Rules for the relevant years. The average MS-DRG weight would be specific to each participant hospital and region to account for hospital and region-specific inpatient acute service utilization patterns. Hospital-specific and region-specific average MS-DRG weights would be calculated by averaging the MS-DRG weight for all the IPPS MS-DRGs included in the historical episodes attributed to each participant hospital and attributed to CJR eligible hospitals in the region, respectively; including MS-DRGs for anchor admissions as well as those for subsequent readmissions that fall within the episode definition. Expressed as a ratio, the inpatient acute services adjustment factor would equal the following:

  • The numerator is based on values applicable for the upcoming performance period (PP) for which a target price is being calculated.Start Printed Page 73345
  • The denominator is based on a blend of values applicable in the latest of the 3 historical years used in the target price (TP) calculations, weighted to account for the values applicable prior to October 1, and values applicable starting October 1 when IPPS updates for the new fiscal year are in effect. Note that this weighting incorporates a modification to our proposed methodology for calculating update factors, as previously discussed in section III.C.4.(b)(4) of this final rule.

Therefore, the inpatient acute services update factor formula is shown as—

(b) Physician Services Update Factor

The proposed physician services update factor would apply to payments for services included in the episode paid under the MPFS for physician services. We proposed to use changes in the following MPFS inputs to calculate the physician services update factor of each participant hospital and region's historical average episode payments: RVUs; work, practice expense, and malpractice (MP) liability geographic practice cost indices (GPCIs); and national conversion factor, as defined in the MPFS Final Rule for the relevant years. Hospital-specific and region-specific RVU-weighted GPCIs would be calculated to account for hospital and region-specific physician service utilization patterns. Hospital-specific and region-specific RVU-weighted GPCIs would be calculated by taking the proportion of RVUs for work, practice expense, and MP liability for physician services included in the historical episodes and attributed to each participant hospital and attributed to CJR eligible hospitals in the region, respectively, and multiplying each proportion by the relevant GPCI.

Expressed as a ratio, the physician services update factor would equal the following:

  • The numerator is based on GPCI values applicable for the upcoming performance period (PP) for which a target price is being calculated.
  • The denominator is based on GPCI values applicable at the end of the latest of the 3 historical years used in the target price (TP) calculations.

Therefore, the proposed physician services update factor formula is shown as—

(c) IRF Services Update Factor

The proposed IRF services update factor applies to payments for services included in the episode paid under the Medicare inpatient rehabilitation facility prospective payment system (IRF PPS). We proposed to use changes in the IRF Standard Payment Conversion Factor, an input for the IRF PPS and defined in the IRF PPS Final Rule for the relevant years, to update Medicare payments for IRF services provided in the episode. The IRF Standard Payment Conversion Factor is the same for all IRFs and IRF services, so there is no need to account for any hospital-specific or region-specific IRF utilization patterns; each participant hospital and region would use the same IRF services update factor.

Expressed as a ratio, the IRF PPS update factor would equal the following:

  • The numerator is based on values applicable for the upcoming performance period (PP) for which a target price is being calculated.
  • The denominator is based on a blend of values applicable in the latest of the 3 historical years used in the target price (TP) calculations, weighted to account for the values applicable prior to October 1, and values applicable starting October 1 when IRF PPS updates for the new fiscal year are in effect. Note that this weighting incorporates a modification to our proposed methodology for calculating update factors, as previously discussed in section III.C.4.(b)(4) of this final rule.

Therefore, the IRF services update factor formula is shown as—

(d) SNF Services Update Factor

The proposed SNF services update factor would apply to payments for services included in the episode and paid under the SNF PPS, including payments for SNF swing bed services. The update factor applied to the SNF services component of each participant hospital and region's historical average episode payments would be based on how average Resource Utilization Group (RUG-IV) Case-Mix Adjusted Federal Rates for the Medicare SNF PPS (defined in the SNF PPS Final Rule) have changed between the latest year used in the historical 3 years of episodes and the upcoming performance period under CJR. The average RUG-IV Case-Mix Adjusted Federal Rates would be specific to each participant hospital and region to account for hospital and region-specific SNF service utilization patterns. Hospital-specific and region-specific average RUG-IV Case-Mix Adjusted Federal Rates would be calculated by averaging the RUG-IV Case-Mix Adjusted Federal Rates for all SNF services included in the historical episodes attributed to each participant hospital and attributed to CJR eligible hospitals in the region, respectively. We note that the RUG-IV Case-Mix Adjusted Federal Rate may vary for the same RUG, depending on whether the SNF was categorized as urban or rural.

Expressed as a ratio, the SNF services update factor would equal the following:

  • The numerator is based on values applicable for the upcoming Start Printed Page 73346performance period (PP) for which a target price is being calculated.
  • The denominator is based on a blend of values applicable in the latest of the 3 historical years used in the target price (TP) calculations, weighted to account for values applicable prior to October 1, and values applicable starting October 1 when SNF PPS updates for the new fiscal year are in effect. Note that this weighting incorporates a modification to our proposed methodology for calculating update factors, as previously discussed in section III.C.4.(b)(4) of this final rule.

Therefore, the SNF services update factor formula is shown as—

(e) HHA Services Update Factor

The proposed HHA services update factor would apply to payments for services included in the episode and paid under the HH PPS, but exclude payments for Low Utilization Payment Adjustment (LUPA) claims (claims with four or fewer home health visits) because they are paid differently and would instead be captured in the update factor for other services in section III.C.4.b.(f) of the proposed rule. The update factor applied to the home health services component of each participant hospital and region's historical average episode payments would be based on how inputs for the Medicare HH PPS have changed between the latest year used in the historical 3 years of episodes and the upcoming performance period under CJR. We proposed to use changes in the HH PPS base rate and average of home health resource group (HHRG) case-mix weight, inputs for the HHA PPS and defined in the HHA PPS Final Rule for the relevant years, to calculate the home health services update factor. The average HHRG case-mix weights would be specific to each participant hospital and region to account for hospital and region-specific home health service utilization patterns. Hospital-specific and region-specific HHA services update factors would be calculated by averaging the HHRG case-mix weights for all home health payments (excluding LUPA claims) included in the historical episodes attributed to each participant hospital and attributed to CJR eligible hospitals in the region, respectively.

Expressed as a ratio, the HHA adjustment factor would equal the following:

  • The numerator is based on values applicable for the upcoming performance period (PP) for which a target price is being calculated.
  • The denominator is based on values applicable at the end of the latest of the 3 historical years used in the target price (TP) calculations.

Therefore, the proposed HHA services update factor formula is shown as—

(f) Other Services Update Factor

The other services update factor would apply to payments for services included in the episode and not paid under the IPPS, MPFS, IRF PPS, or HHA PPS (except for LUPA claims). This component would include episode payments for home health LUPA claims and CJR related readmissions at CAHs. For purposes of calculating the other services update factor, we proposed to use the Medicare Economic Index (MEI), a measure developed by CMS for measuring the inflation for goods and services used in the provision of physician services.[28] We would calculate the other services update factor as the percent change in the MEI between the latest year used in the TP calculation and its projected value for the upcoming performance period. Because MEI is not hospital or region-specific, each participant hospital and region would use the same other services update factor.

(5) Blend Hospital-Specific and Regional Historical Data

We proposed to calculate CJR episode target prices using a blend of hospital-specific and regional historical average CJR episode payments, including CJR episode payments for all CJR eligible hospitals in the same U.S. Census division as discussed further in section III.C.4.b.(6) of the proposed rule. Specifically, we proposed to blend two-thirds of the hospital-specific episode payments and one-third of the regional episode payment to set a participant hospital's target price for the first 2-performance years of the CJR model (CY 2016 and CY 2017). For performance year 3 of the model (CY 2018), we proposed to adjust the proportion of the hospital-specific and regional episode payments used to calculate the episode target price from two-thirds hospital-specific and one-third regional to one-third hospital-specific and two-thirds regional. Finally, we proposed to use only regional historical CJR episode payments for performance years 4 and 5 of the model (CY 2019 and CY 2020) to set a participant hospital's target price, rather than a blend between the hospital-specific and regional episode payments. The specific order of steps, and how this step fits in with others, is discussed further in section III.C.4.c. of the proposed rule. We welcome comment on the appropriate blend between hospital-specific and regional episode payments and the change in that blend over time.

We considered establishing episode target prices using only historical CJR hospital-specific episode payments for all 5 performance years of the model (that is, episode payments for episodes attributed to the participant hospital, as previously described in section III.C.2. of the proposed rule). Using hospital-specific historical episodes may be appropriate in other models such as BPCI Model 2 where participation is voluntary and setting a region-wide target price could lead to a pattern of selective participation in which inefficient providers decline to participate, undermining the model's ability to improve the efficiency and quality of care delivered by those providers, while already-efficient providers receive windfall gains even if they do not further improve efficiency. Because CJR model participants will be required to participate in the model, Start Printed Page 73347solely using hospital-specific historical episode data is not necessary to avoid this potential concern. Furthermore, using only hospital-specific historical CJR episode payments may provide little incentive for hospitals that already cost-efficiently deliver high quality care to maintain or further improve such care. These hospitals could receive a relatively low target price because of their historical performance but have fewer opportunities for achieving additional efficiency under CJR. They would not receive reconciliation payments for maintaining high quality and efficiency, while other hospitals that were less efficient would receive reconciliation payments for improving, even if the less historically efficient hospitals did not reach the same level of high quality and efficiency as the more historically efficient hospitals. Using only hospital-specific historical CJR episode payments may also not be sufficient to curb inefficient care or overprovision of services for hospitals with historically high CJR episode payments. In such instances, using hospital-specific historical episode payments for the CJR model could result in Medicare continuing to pay an excessive amount for episodes of care provided by inefficient hospitals, and inefficient hospitals would stand to benefit from making only small improvements. Thus, we did not propose to set target prices based solely on hospital-specific data for any performance years of the model.

We considered establishing the episode target price using only historical CJR regional episode payments for all 5 performance years of the model. Though regional target pricing would reward the most efficient hospitals for continuing to provide high quality and cost efficient care, we are concerned about providing achievable incentives under the model for hospitals with high historical CJR average episode payments. We stated in the proposed rule our belief that a lower regional price for such hospitals would leave them with little financial incentive in performance year 1, especially without any responsibility to repay payments in excess of the target price as described in section III.C.3. of the proposed rule. Thus, we did not propose to set target prices solely on regional data for the entire duration of the model.

Therefore, we proposed initially to blend historical hospital-specific and regional-historical episode payments and then transition to using regional-only historical episode payments in establishing target prices to afford early and continuing incentives for both historically efficient and less efficient hospitals to furnish high quality, efficient care in all years of the model. Our proposal more heavily weights a hospital's historical episode data in the first 2 years of the model (two-thirds hospital-specific, one-third regional), providing a reasonable incentive for both currently efficient and less efficient hospitals to deliver high quality and efficient care in the early stages of model implementation. Beginning in performance year 3, once hospitals have engaged in care redesign and adapted to the model parameters, we proposed to shift to a more heavily weighted regional contribution (one-third hospital-specific, two-thirds regional in performance year 3) and ultimately to a regional target price for performance years 4 and 5. We stated in the proposed rule our belief that by performance year 4, setting target prices based solely on regional historical data would be feasible because hospitals would have had 3 years under this model to more efficiently deliver high quality care, thereby reducing some of the variation across hospitals. We stated in the proposed rule our belief that transitioning to regional only pricing in the latter years of the model would provide important information about the reduction in unnecessary variation in LEJR episode utilization patterns within a region that can be achieved.

We stated in the proposed rule our belief that transitioning to regional-only pricing in the latter years of the model may provide valuable information regarding potential pricing strategies for successful episode payment models that we may consider for expansion in the future. As discussed previously, substantial regional and hospital-specific variation in Medicare LEJR episode spending currently exists for beneficiaries with similar demographic and health status, so we are proposing that the early CJR model years will more heavily weight historical hospital-specific experience in pricing episode for a participant hospital. Once the hospital has substantial experience with care redesign, we expect that unnecessary hospital-specific variation in episode spending will be minimized so that regional-only pricing would be appropriate as we have proposed. We noted that, like episode payment under the CJR model, Medicare's current payment systems make payments for bundles of items and services, although of various breadths and sizes depending on the specific payment system. For example, the IPPS pays a single payment, based on national prices with geography-specific labor cost adjustments, for all hospital services furnished during an inpatient hospitalization, such as nursing services, medications, medical equipment, operating room suites, etc. Under the IPPS, the national pricing approach incentivizes efficiencies and has, therefore, led to a substantial reduction in unnecessary hospital-specific variation in resource utilization for an inpatient hospitalization. On the other hand, the episode payment approach being tested under BPCI Model 2 relies solely on provider-specific pricing over the lifetime of the model, assuming the number of episode cases is sufficient to establish a reliable episode price, an approach that has potential limitations were expansion to be considered. Thus, we stated in the proposed rule our belief that our proposal for CJR will provide new, important information regarding pricing for even larger and broader bundles of services once unnecessary provider-specific variation has been minimized that would supplement our experience with patterns and pricing under existing payment systems and other episode payment models. We expect that testing of CJR will contribute further information about efficient Medicare pricing strategies that result in appropriate payment for providers' resources required to furnish high quality, efficient care to beneficiaries who receive LEJR procedures. This is essential information for any consideration of episode payment model expansion, including nationally, in the future, where operationally feasible and appropriate pricing strategies, including provider-specific, regional, and national pricing approaches would need to be considered.

We proposed an exception to the blended hospital-specific and regional pricing approach for hospitals with low historical CJR episode volume. We proposed to define hospitals with low CJR episode volume as those with fewer than 20 CJR episodes in total across the 3-historical-years used to calculate target prices. We stated in the proposed rule our belief that calculating the hospital-specific component of the blended target price for these historically low CJR episode volume hospitals may be subject to a high degree of statistical variation. Therefore, for each performance year, we proposed to use 100 percent regional target pricing for participant hospitals who have fewer than twenty historical CJR episodes in the 3-historical-years used to calculate target prices, as described in section III.C.4.b.(2) of the proposed rule. We note that the 3-historical-years used Start Printed Page 73348to calculate target prices would change over the course of the model, as described in section III.C.4.b.(2) of the proposed rule, and when that happens, the twenty episode threshold would be applied to the new set of historical years. If all IPPS hospitals nationally participated (for estimation purposes, only) in CJR, we estimate about 5 percent of hospitals would be affected by this proposed low historical CJR episode volume provision.[29] A minimum threshold of twenty episodes is almost equal to the minimum number of admissions required in the Medicare HRRP. HRRP payment adjustment factors are, in part, determined by procedure/condition-specific readmission rates for a hospital. HRRP requires at least 25 procedure/condition-specific admissions to calculate the procedure/condition-specific readmission rate and to be included in the hospital's overall HRRP payment adjustment factor. Though the proposed minimum threshold of twenty episodes is slightly less than the 25 admissions required for HRRP, we stated in the proposed rule our belief that because we would not be calculating infrequent events such as readmissions, we can achieve a stable price with slightly fewer episodes.

We also proposed an exception to the blended hospital-specific and regional pricing approach for participant hospitals that received new CCNs during the 24 months prior to the beginning of, or during, the performance year for which target prices are being calculated. These participant hospitals with new CCNs may have formed due to a merger between or split from previously existing hospitals, or may be new hospitals altogether. As a general principle, we aim to incorporate into the target prices all the historical episodes that would represent our best estimate of CJR historical payments for these participant hospitals with new CCNs. For participant hospitals with new CCNs that formed from a merger between or split from previously existing hospitals, we proposed to calculate hospital-specific historical payments using the episodes attributed to the previously existing hospitals. These hospital-specific historical payments would then be blended with the regional historical payments according to the approach previously described in this section. For participant hospitals with new CCNs that are new hospitals altogether, we proposed to use the approach previously described in this section for hospitals with fewer than 20 CJR episodes across the 3-historical-years used to calculate target prices. In other cases, due to an organizational change a hospital may experience a change to an already existing CCN during the 24 months prior to the beginning of, or during, the performance year for which target prices are being calculated. For example, one hospital with a CCN may merge with a second hospital assigned a different CCN, and both hospitals would then be identified under the single CCN of the second hospital. While there may be more than 20 CJR episodes under the second hospital's CCN in total across the 3-historical-years used to calculate target prices, in this scenario our use of only those cases under the second hospital's CCN in calculating hospital-specific historical payments would fail to meet our general principle of incorporating into target prices all the historical episodes that would represent our best estimate of CJR historical payments for these now merged hospitals. In this scenario, we proposed to calculate hospital-specific payments for the remaining single CCN (originally assigned to the second hospital only) using the historical episodes attributed to both previously existing hospitals. These hospital-specific historical payments would then be blended with the regional historical payments according to the approach previously described in this section in order to determine the episode price for the merged hospitals bearing a single CCN.

We sought comment on this proposed approach for blending hospital-specific and regional historical payments.

The following is a summary of the comments received and our responses.

Comment: Many commenters supported the proposal to blend hospital-specific and regional historical episode data to calculate target prices. They explained that this balanced the incentivizes for already efficient hospitals to continue great performance, and allowed hospitals with historically high episode expenditures sufficient time to create care pathways and implement practice pattern changes before getting to 100 percent regional pricing in years 4 and 5 of the CJR model. Some other commenters recommended for hospital-specific pricing only because any definition of region would not properly account for variations due to factors such as patient characteristics, socioeconomic factors, and access to care.

Some commenters recommended that instead of blending regional and hospital-specific historical average CJR episode payments, we use the higher of the two to reward hospitals that are already efficient.

Some commenters recommended that we delay the transition to regional pricing in order to afford more time for hospitals with high historic episode expenditures, some commenters supported our proposal to get to 100 percent regional pricing by year 4, and some others recommended that we accelerate the transition to regional pricing to appropriately reward already efficient participant hospitals.

Response: We appreciate commenters' support for blending hospital-specific and regional historical episode data to calculate target prices. We appreciate that the pace of transitioning to regional pricing may benefit some participant hospitals more than others. However, we believe that the proposed approach to get to 100 percent regional pricing by year 4 strikes an appropriate balance between providing participant hospitals time to adapt while providing important information about the reduction in unnecessary variation in LEJR episode utilization patterns within a region that can be achieved.

We believe that only using hospital-specific pricing would not reward already efficient participant hospitals for maintaining high performance; participant hospitals that are already delivering efficient and high quality care would find it challenging to improve upon their own historical performance in order to qualify for reconciliation payments. Similarly, we believe that using the higher of regional and hospital-specific prices would not sufficiently incentivize inefficient participant hospitals to become more efficient; participant hospitals that have historically high episode expenditures would have less of an incentive to become significantly more efficient over the course of the model if they can qualify for reconciliation payments by improving only slightly relative to their own historical performance, while still being less efficient than their regional peers.

We acknowledge the importance of properly accounting for variations in patient-specific clinical characteristics, socioeconomic conditions, and access to care to appropriately incentivize participant hospitals to deliver high quality and efficient care. We refer readers to response to comments in section III.C.4.b.(1) of this final rule for further discussion on risk stratification to account for such variations. We also acknowledge that incorporating a regional component of historical episode data into a participant hospital's target prices may increase the presence of the variations as Start Printed Page 73349commenters stated, thereby making appropriate risk adjustment and/or risk stratification that much more important. As discussed in the response to comments in section III.C.4.b.(1) of this final rule, we will risk stratify based on anchor MS-DRG and hip fracture status, and we may explore more comprehensive risk adjustment approaches.

Comment: Several commenters recommended modifying the definition of low volume as it is used to determine which participant hospitals receive 100 percent regional target prices because they do not have a sufficient number of CJR episodes in the 3-historical-years of data used to calculate target prices. Commenters suggested increasing the low volume threshold for hospital-specific and regional target pricing from 20 to, for example, 100 episodes, because 20 episodes was not sufficient to remove random variation.

Response: We agree with commenters that a greater number of participant hospital-specific episodes would better remove the effects of random variation. However, if we increase the low volume threshold for blending hospital-specific and regional target prices, more participant hospitals would receive 100 percent regional prices in the first three performance years of the model, and their target prices would not incorporate any data from hospital-specific historical experience. Let us take as an example a participant hospital that has 50 episodes in the 3-historical-years of data used to calculate target prices for performance year 1, and let us assume that the hospital-specific portion of its target price is higher than the regional component. This participant hospital would need to become more efficient so as to achieve actual episode expenditures below its target prices. By blending the hospital-specific and regional components of the target price, this hospital has a higher target price than it would have had it received a 100 percent regional price. With the higher target price, the participant hospital has a greater opportunity to improve its efficiency and qualify for reconciliation payments. The blending of regional and hospital-specific target prices affords historically less efficient hospitals an opportunity to be rewarded for improvement in the earlier performance years, while they prepare for transitioning to 100 percent regional pricing by performance year 4. We want to afford this transition opportunity to as many participant hospitals as possible, while minimizing the effect of random variations for hospitals with few historical episodes. In the proposed rule, we compared our proposed low volume threshold of 20 episodes to the threshold used for Medicare's HRRP program. We continue to believe that 20 episodes in the 3-historical-years of data used to calculate target prices is the appropriate “low volume” threshold for blending target prices that mitigates effects of random variation while still incorporating hospital-specific historical experience and affording participant hospitals an opportunity to transition to 100 percent regional pricing.

Final Decision: After consideration of the public comments we received, we are finalizing the proposal to blend hospital-specific and regional historical expenditures in setting target prices, though we note that the term “CJR eligible hospitals' is being renamed to “CJR regional hospitals” as discussed in response to comments in section III.C.4.b.(4) of this final rule.

(6) Define Regions as U.S. Census Divisions

In all 5 performance years we proposed to define “region” as one of the nine U.S. Census divisions [30] in Figure 3.

Start Printed Page 73350

We considered using states, HRRs, and the entire U.S. as alternative options to U.S. Census divisions in defining the region used in blending provider-specific and regional historical episode data for calculating target prices. However, HRR definitions are specifically based on referrals for cardiovascular surgical procedures and neurosurgery, and may not reflect referral patterns for orthopedic procedures. Using the entire U.S. would not account for substantial current regional variation in utilization, which is significant for episodes that often involve PAC use, such as LEJR procedures.[32] Finally, we considered using states as regions but were concerned that doing so would not allow for sufficient LEJR episode volume to set stable regional components of target prices, especially for participant hospitals in small states. We believe U.S. Census divisions provide the most appropriate balance between very large areas with highly disparate utilization patterns and very small areas that would be subject to price distortions due to low volume or hospital-specific utilization patterns.

We sought comment on our proposal to define a region as the U.S. Census division for purposes of the regional component of blended target prices under CJR.

The following is a summary of the comments received and our responses.

Comment: Some commenters supported the use of U.S. Census divisions as regions. Some commenters, though, stated U.S. Census divisions are too large with significant practice and PAC access variations, resulting in different average historical expenditures across hospitals in the same U.S. Census division. Some commenters suggested an alternative of using MSAs as regions; MSAs would align with the provider selection process, and the smaller unit for regions would better capture regional practice pattern differences. Other commenters, including MedPAC, stated that we should define the entire nation as the region (that is, national pricing) because we should be striving towards eliminating regional variations in practice patterns.

Response: We appreciate commenters' support for the use of U.S. Census divisions as regions. Especially given that commenters proposed both larger regions (that is, national pricing) and smaller regions (that is, MSAs), we still believe U.S. Census divisions provide the most appropriate balance between very large areas with highly disparate utilization patterns and very small areas that would be subject to price distortions due to low volume or hospital specific utilization patterns.

Comment: Several commenters noted that some of the selected MSAs for participation in CJR span two different U.S. Census divisions. These commenters stated that the true cost for hospitals in the same MSA would likely not be different, and significant differences in pricing would create unfair market advantages due to a hospital's address within an MSA. They suggested blending the regional target price component of the two U.S. Census divisions that are being spanned for each of these MSAs, reflecting the distribution of the population within the MSA/census regions.

Response: We agree with commenters that the true cost for hospitals in the same MSA may not be different, and significant differences in pricing may create unfair market advantages due to a hospital's address within an MSA. We will modify our proposal and apply the same regional target price component to target pricing for all participant hospitals within an MSA, even if the MSA spans two U.S. Census divisions. There are three selected MSAs for participation in CJR that span two U.S. Census divisions: St. Louis, Cincinnati, and Cape Girardeau.

We considered the approach suggested by commenters—blending the two regional target price components based on the population distribution. However, using 2010 U.S. Census data, we determined that at least 75 percent of the population in the previously mentioned MSAs resides in just one of the U.S. Census divisions that the MSA spans. For simplicity, we will completely group MSAs that span U.S. Census divisions together with the U.S. Census divisions in which the Census estimates the majority of people reside, as shown in Table 9:

Table 9—Region Grouping for Selected MSAS That Span U.S. Census Divisions

MSAOriginal U.S. Census divisions spanned by MSA (state included in MSA)U.S. Census division used for CJR region
St. Louis, MO-ILWest North Central (MO), East North Central (IL)West North Central.
Cincinnati, OH-KY-INEast North Central (OH, IN), East South Central (KY)East North Central.
Cape Girardeau, MO-ILWest North Central (MO), East North Central (IL)West North Central.

Final Decision: After consideration of the public comments we received, we are modifying our proposal to define regions as U.S. Census divisions so as to ascribe the same regional component of target prices for participant hospitals in MSAs that span U.S. Census divisions. Specifically, as described in Table 9, selected MSAs that span U.S. Census divisions will be attributed to one U.S. Census division for purposes of calculating the regional component of CJR target prices.

(7) Normalize for Provider-Specific Wage Adjustment Variations

We note that some variation in historical CJR episode payments across hospitals in a region may be due to wage adjustment differences in Medicare's payments. In setting Medicare payment rates, Medicare typically adjusts facilities' costs attributable to wages and wage-related costs (as estimated by the Secretary from time to time) by a factor (established by the Secretary) reflecting the relative wage level in the geographic area of the facility or practitioner (or the beneficiary residence, in the case of home health and hospice services) compared to a national average wage level. Such adjustments are essential for setting accurate payments, as wage levels vary significantly across geographic areas of the country. However, having the wage level for one hospital influence the regional-component of hospital-specific and regional blended target prices for another hospital with a different wage level would introduce unintended pricing distortions not based on utilization pattern differences.Start Printed Page 73351

In order to preserve how wage levels affect provider payment amounts, while minimizing the distortions introduced when calculating the regional-component of blended target prices, we proposed to normalize for wage index differences in historical episode payments when calculating and blending the regional and hospital-specific components of blended target prices. Calculating blended target prices from historical CJR episodes would help ensure we incentivize hospitals based on historical utilization and practice patterns, not Medicare payment system rate changes that are beyond hospitals' control.

We proposed to normalize for provider-specific wage index variations using the IPPS wage index applicable to the anchor hospitalization (that is, the IPPS wage index used in the calculation of the IPPS payment for the anchor hospitalization). The anchor hospitalization accounts for approximately 50 percent of the total episode expenditures, and the IPPS wage index is applied to IPPS payments in a similar manner as wage indices for other Medicare payment systems are applied to their respective payments.[33] Therefore, we proposed that the IPPS wage index applicable to the anchor hospitalization for each historical episode be used to normalize for wage index variations in historical episode payments across hospitals when calculating blended target prices. We proposed to specifically perform this normalization using the wage normalization factor (0.7 * IPPS wage index + 0.3) to adjust the labor-related portion of payments affected by wage indices. The 0.7 approximates the labor share in IPPS, IRF PPS, SNF, and HHA Medicare payments. We would normalize for provider-specific wage index variations by dividing a hospital's historical episode payments by the wage normalization factor.

We proposed to reintroduce the hospital-specific wage variations by multiplying episode payments by the wage normalization factor when calculating the target prices for each participant hospital, as described in section III.C.4.c. of the proposed rule. When reintroducing the hospital-specific wage variations, the IPPS wage index would be the one that applies to the hospital during the period for which target prices are being calculated (for example, FY 2016 wage indices for the target price calculations for episodes that begin between January 1 and September 30, 2016). The specific order of steps, and how this step fits in with others, is discussed further in section III.C.4.c. of the proposed rule. We sought comment on our proposal to normalize for wage index differences using participant hospitals' wage indices in order to calculate blended target prices.

The following is a summary of the comments received and our responses.

Comment: Commenters emphasized the need to account for wage index differences. Not accounting for these differences accurately may unfairly disadvantage some hospitals. Some commenters expressed concern about using 0.7 as the labor share for the labor related portions of Medicare FFS payments; the weight index weight varies by Medicare FFS payment system, and in IPPS in particular, the weight can be either 0.688 or 0.620, depending on the IPPS hospital's wage index. Some other commenters noted that using only the IPPS wage index for the anchor hospital would not accurately normalize expenditures for PAC providers who have their own wage indices. Some of these commenters recommended we blend hospital and PAC providers' wage indices. Some commenters requested clarification on how we would account for wage index differences between baseline and performance periods.

Response: We acknowledge the need to accurately account for wage index differences so that we incentivize based on practice patterns and not Medicare FFS fee schedule differences. We recognize that the proposed approach of using the anchor hospital's wage index and 0.7 as the labor share for the labor related portions of Medicare FFS payments would only approximately normalize and reapply wage indices.

In response to commenters, we will modify our proposal and normalize for wage indices at the claim level for both historical episode expenditures and actual episode expenditures in each performance year by using the wage index normalization algorithm included in the CMS Price (Payment) Standardization Detailed Methodology discussed in section III.C.3 of this final rule, the same methodology we finalized to exclude the various special payment provisions in calculating episode expenditures. By normalizing claims for wage indices in the historical episode expenditure data at the claim level, we will accurately account for wage indices and labor shares for various providers and suppliers under the different Medicare FFS payment systems. This will be a more accurate way than what we proposed to achieve the same goal of accounting for wage index differences so that we incentivize based on practice patterns and not Medicare FFS wage adjustment differences. We will also normalize claims for wage indices in performance year data, as we discuss further in response to comments in section III.C.6.a. of this final rule.

We believe it is still important to reintroduce wage index variations near the end of the target price calculation methodology. Participant hospitals may use their reconciliation payments to invest in care coordination or care delivery infrastructure, and we expect that the costs for such investments would vary by geography due to differences in local wages. For example, we expect that hiring a care coordinator would cost a participant hospital more in the New York metro region than in a rural part of New Mexico. If we do not reintroduce wage index variations into target price calculations, we would calculate reconciliation and repayment amounts that would not capture labor cost variation throughout the country, and participant hospitals in higher labor cost regions may see relatively less financial incentive to invest in improved care quality and efficiency. We intend to incentivize all hospitals to reduce episode spending under the CJR model, regardless of local labor cost variations.

We will use the proposed approach to reintroduce wage index variations—apply the participant hospital's wage index to episode spending, using 0.7 as the labor share. While commenters are correct that the IPPS labor share can be 0.688 or 0.620, depending on the participant hospital's wage index, the labor share for PAC providers also varies across Medicare FFS payment systems: ~0.695 for SNF PPS and IRF PPS, and ~0.785 for HH PPS. Given this range for the labor share across Medicare FFS payment systems, we believe that using 0.7 is an appropriate estimate of the labor share for reintroducing wage index variations. Additionally, as commenters pointed out, PAC providers have their own wage indices. Because wage index variations are reintroduced near the end of the target price calculation methodology and after other features, such as blending, pooling, and update factors are applied, we do not believe there is a simple approach to reintroduce wage index variations at the claim level. We acknowledge that using the participant hospital's wage index and 0.7 as the labor share would only be an approximation of the wage index variations, but this approximation would not change whether a participant hospital qualifies for reconciliation Start Printed Page 73352payments or is obligated to repay Medicare because we would apply wage index normalization at the claim level for both target price calculations (as previously discussed) as well as calculations of actual episode spending (as discussed in response to comments in section III.C.6.a. of this final rule), and the wage index variation would be reintroduced in the same manner to both target price calculations (as previously discussed) and actual episode spending calculations (as discussed in response to comments in section III.C.6.a. of this final rule). We believe that this approach to reintroducing wage index variations is sufficient to modulate the reconciliation and repayment amounts to reflect local labor cost variations.

Final Decision: After consideration of the public comments we received, we are modifying our proposal so as to normalize for wage indices at the claim level by using the wage index normalization algorithm included in the CMS Price (Payment) Standardization Detailed Methodology discussed in section III.C.3., the same claim-level standardization methodology we finalized in section III.C.3.a. to exclude the various special payment provisions in calculating episode expenditures. We are finalizing the proposal to reintroduce wage index differences into calculations of historical and actual episode spending based on the participant hospital's wage index and 0.7 as the labor cost share.

(8) Combination of CJR Episodes Anchored by MS-DRGs 469 and 470

We proposed to pool together CJR episodes anchored by MS-DRGs 469 and 470 for target price calculations to use a greater historical CJR episode volume and set more stable target prices. We note that we would still calculate separate target prices for episodes anchored by MS-DRGs 469 versus 470, described later in this section.

To pool together MS-DRG 469 and 470 anchored episodes, we proposed to use an anchor factor and hospital weights. The anchor factor would equal the ratio of national average historical MS-DRG 469 anchored episode payments to national average historical MS-DRG 470 anchored episode payments. The national average would be based on episodes attributed to any CJR eligible hospital. The resulting anchor factor would be the same for all participant hospitals. For each participant hospital, a hospital weight would be calculated using the following formula, where episode counts are participant hospital-specific and based on the episodes in the 3-historical-years used in target price calculations:

A hospital-specific pooled historical average episode payment would be calculated by multiplying the hospital's hospital weight by its combined historical average episode payment (sum of MS-DRG 469 and 470 anchored historical episode payments divided by the number of MS-DRG 469 and 470 historical episodes).

The calculation of the hospital weights and the hospital-specific pooled historical average episode payments would be comparable to how case mix indices are used to generate case mix-adjusted Medicare payments. The hospital weight essentially would count each MS-DRG 469 triggered episode as more than one episode (assuming MS-DRG 469 anchored episodes have higher average payments than MS-DRG 470 anchored episodes) so that the pooled historical average episode payment, and subsequently the target price, is not skewed by the hospital's relative breakdown of MS-DRG 469 versus 470 anchored historical episodes.

The hospital-specific pooled historical average payments would be modified by blending and discount factors, as described in section III.C.4.c. of the proposed rule. Afterwards, the hospital-specific pooled calculations would be “unpooled” by setting the MS-DRG 470 anchored episode target price to the resulting calculations, and by multiplying the resulting calculations by the anchor factor to produce the MS-DRG 469 anchored target prices.

We would calculate region-specific weights and region-specific pooled historical average payments following the same steps proposed for hospital-specific weights and hospital-specific pooled average payments. Instead of grouping episodes by the attributed hospital as is proposed for hospital-specific calculations, region-specific calculations would group together episodes that were attributed to any CJR eligible hospital located within the region. The hospital-specific and region-specific pooled historical average payments would be blended together as discussed in section III.C.4.b.(3) of the proposed rule. The specific order of steps, and how this step fits in with others, is discussed further in section III.C.4.c. of the proposed rule.

We considered an alternative option of independently setting target prices for MS-DRG 470 and 469 anchored episodes without pooling them. However, hospital volume for MS-DRG 469 was substantially less than for MS-DRG 470. In 2013 across all IPPS hospitals, there were more than 10 times as many MS-DRG 470 anchored episodes as compared to MS-DRG 469 anchored episodes.[34] In the same analysis, the median number of episodes for a hospital with at least 1 episode for the MS-DRG anchored episode was more than 80 for MS-DRG 470 anchored episodes, though fewer than 10 for MS-DRG 469 anchored episodes. Calculating target prices for MS-DRG 469 anchored episodes separately for each participant hospital may result in too few historical episodes to calculate reliable target prices. We also considered pooling together MS-DRG 469 and 470 anchored episodes without any anchor factor or hospital weights. However, internal analyses suggest that average episode payments for these two MS-DRG anchored episodes significantly differed; CJR episodes initiated by MS-DRG 469 had payments almost twice as large as those initiated by MS-DRG 470.[35] This difference is reasonable given that Medicare IPPS payments differ for MS-DRG 469 and 470 admissions, and inpatient payments comprise approximately 50 percent of CJR episode payments. Thus, pooling together MS-DRG 469 and 470 anchored episodes without any anchor factor or hospital weights would introduce distortions due only to case-mix differences.

The following is a summary of the comments received and our responses.Start Printed Page 73353

Comment: Many commenters requested for risk adjustment based on patients' hip fracture status, among other clinical and demographic dimensions.

Response: We refer readers to comments and responses to comments in section III.C.4.b.(1) of this final rule for further discussion on risk stratification, and we reference it here because changes to risk stratification would impact how we would combine CJR episodes anchored by MS-DRGs 469 and 470. As discussed in the responses to comments in section III.C.4.b.(1) of this final rule, we will modify our proposal so as to risk stratify and set different target prices both for episodes anchored by MS-DRG 469 vs. MS-DRG 470 and for episodes with hip fractures vs. without hip fractures. To fully incorporate this change, we will also modify the proposed approach to calculate anchor factors and hospital and regional weights so as to apply them to four groups of target prices, instead of two groups; otherwise, the approach will be the same as proposed. Specifically, we will have three anchor factors, instead of one:

Additionally, hospital and regional weights will be calculated using the following formula:

Final Decision: After consideration of the public comments we received, we are finalizing our proposal, with modification to calculate anchor factors and hospital and regional weights while incorporating the previously discussed changes to risk adjust not only on anchor MS-DRG but also hip fracture status. Additionally, note that the term “CJR eligible hospitals” is being renamed to “CJR regional hospitals” as discussed in response to comments in section III.C.4.b.(4) of this final rule.

(9) Discount Factor

When setting an episode target price for a participant hospital, we proposed to apply a discount to a hospital's hospital-specific and regional blended historical payments for a performance period to establish the episode target price that would apply to the participant hospital's CJR episodes during that performance period and for which the hospital would be fully, or partly, accountable for episode spending in relationship to the target price, as discussed in section III.C.3. of the proposed rule. We expect participant hospitals to have significant opportunity to improve the quality and efficiency of care furnished during episodes in comparison with historical practice, because this model would facilitate the alignment of financial incentives among providers caring for beneficiaries throughout the episode. This discount would serve as Medicare's portion of reduced expenditures from the CJR episode, with any episode expenditure below the target price potentially available as reconciliation payments to the participant hospital where the anchor hospitalization occurred. We proposed to apply a 2 percent discount for performance years 1 through 5 when setting the target price. We stated our belief in the proposed rule that applying a 2 percent discount in setting the episode target price allows Medicare to partake in some of the savings from the CJR model, while leaving considerable opportunity for participant hospitals to achieve further episode savings below the target price that they would be paid as reconciliation payments, assuming they meet the quality requirements as discussed in section III.C.5 of the proposed rule.

The proposed 2 percent discount is similar to the range of the discounts used for episodes in the ACE demonstration.[36] In the Medicare ACE, Start Printed Page 73354a demonstration program that included orthopedic procedures such as those included in CJR, participant hospitals negotiated with Medicare discounts of 2.5 to 4.4 percent of all Part A orthopedic services and 0.0 to 4.4 percent of all Part B orthopedic services during the inpatient stay (excluding PAC). Hospitals received the discounted payment and reported that they were still able to achieve savings.[37] We stated our belief in the proposed rule that there is similar, if not potentially more, opportunity for savings in the CJR payment model because it includes acute inpatient, as well as PAC, an area of episode spending that accounts for approximately 25 percent of CJR episode payments and exhibits more than 2 times the episode payment variation [38] than that of acute inpatient hospitalization.[39] We stated in the proposed rule our belief that with the proposed 2 percent discount, participant hospitals have an opportunity to create savings for themselves as well as Medicare, while also maintaining or improving quality of care for beneficiaries.

The proposed 2 percent discount also matches the discount used in the BPCI Model 2 90-day episodes, and is less than the discount used in BPCI Model 2 30-day and 60-day episodes (3 percent). Hundreds of current BPCI participants have elected to take on responsibility for repayment in BPCI Model 2 with a 2 to 3 percent discount. Because many BPCI participants volunteered to participate in a bundled payment model with a discount, we stated in the proposed rule our belief that a discount percent that is within, and especially a discount of 2 percent that is at the lower end of, the BPCI discount range would allow CJR participant hospitals to create savings for both themselves and Medicare.

As stated previously in section III.C.3. of the proposed rule, we proposed to phase in the financial responsibility of hospitals for repayment of actual episode spending that exceeds the target price starting in performance year 2. In order to help hospitals transition to taking on this responsibility, we proposed to apply a reduced discount of one percent in performance year 2 for purposes of determining the hospital's responsibility for excess episode spending, but maintain the 2 percent discount for purposes of determining the hospital's opportunity to receive reconciliation payment for actual episode spending below the target price. For example, under this proposal in performance year 2, a hospital that achieves CJR actual episode payments below a target price based on a 2 percent discount would retain savings below the target price, assuming the quality thresholds for reconciliation payment eligibility are met (discussed in section III.C.5. of the proposed rule) and the proposed performance year stop-gain limit (discussed in section III.C.8. of the proposed rule) does not apply. Medicare would hold responsible for repayment hospitals whose CJR actual episode payments exceed a target price based on a one percent discount, assuming the proposed performance year 2 stop-loss limit (discussed in section III.C.8. of the proposed rule) does not apply. Hospitals that achieve CJR actual episode payments between a 2 percent-discounted target price and 1 percent-discounted target price would neither receive reconciliation payments nor be held responsible for repaying Medicare. The decision on which percent-discounted target price applies will be made by evaluating actual episode payments in aggregate after the completion of performance year 2, and the same percent-discounted target price would apply to all episodes that are initiated in performance year 2. We proposed to apply this reduced one percent discount for purposes of hospital repayment responsibility only in performance year 2 and apply the 2 percent discount for excess episode spending repayment responsibility for performance years 3 through 5. Under this proposal, the discount for determination of reconciliation payment for episode actual spending below the target price would not deviate from 2 percent through performance years 1 through 5.

In section III.C.5. of the proposed rule, we proposed voluntary submission of data for a patient-reported outcome measure. We proposed to incent participant hospitals to submit data on this measure by reducing the discount percentage by 0.3 percentage points for successfully submitting data, as defined in section III.D. of the proposed rule. By successfully submitting data on this metric for episodes ending in performance years 1, 2, 3, 4, and or 5, we would adjust the discount percentage in the corresponding year(s) as follows:

  • For episodes beginning in performance year 2, set the discount percentage in a range from 2 percent to 1.7 percent for purposes of determining the hospital's opportunity to receive reconciliation payment for actual episode spending below the target price, and set the discount percentage in a range from 1 percent to 0.7 percent for purposes of determining the amount the hospital would be responsible for repaying Medicare for actual episode spending above the target price.
  • For episodes beginning in performance years 3 through 5, set the discount percentage in a range from 2 percent to 1.7 percent for purposes of reconciliation payment and Medicare repayment calculations.

The determination of whether the hospital successfully submitted data on the patient-reported outcome measure cannot be made until after the performance year ends and data is reported. Therefore, participant hospitals would be provided target prices for both scenarios whether the successfully submit data or not and such determination will happen at the time of payment reconciliation (discussed further in section III.C.6. of the proposed rule).

We sought comment on our proposed discount percentage of 2 percent for CJR episodes, our proposal to reduce the discount to 1 percent on a limited basis in performance year 2, and our proposal to reduce the discount by 0.3 percentage points for successfully reporting patient-reported outcomes data in the corresponding year.

The following is a summary of the comments received and our responses.

Comment: Many commenters expressed concern about participant hospitals taking on financial risk in the CJR model. We refer readers to comments in section III.C.2, of this final rule for more discussion of such comments, and we reference them here because these comments may impact how the proposed discount factor is used to phase in risk for participant hospitals.

Response: As discussed in the responses to comments in section III.C.2. of this final rule, we appreciate commenters' concerns about participant hospitals' ability to manage risk. In the proposed rule, we proposed to use several design elements to phase in risk to better help transition participant hospitals. One of these design elements to phase risk in was the use of a reduced discount factor by 1 percentage point for Start Printed Page 73355purposes of calculating repayment amounts in performance year 2, as discussed earlier in this section. In response to commenters' concerns, we will extend the use of a reduced discount factor for purposes of calculating repayment amounts to apply not only in performance year 2, but also in performance year 3.

Comment: Many commenters offered a variety of suggestions to CMS's proposal and alternatives considered to link quality and payment in the CJR model, including varying the discount percentage incorporated in the target price at reconciliation based on the participant hospital's quality performance. We refer readers to comments in section III.C.5 of this final rule for greater discussion of comments on linking quality and payment in the CJR mode.

Response: As discussed in the responses to comments later in this final rule in section III.C.5. of this final rule, we are modifying the proposed rule so as to use a composite score methodology to link quality and payment in the CJR model. With this composite score methodology, each hospital will receive a discount factor of 3 percent, though the discount factor would be 2 percent for purposes of calculating repayments to Medicare in performance years 2 and 3, reflecting the proposed discount factor reduction by 1 percentage point and the extension to performance year 3 of this reduction, to phase in downside risk, as discussed in the previous response.

Each participant hospital may qualify for a quality incentive payment. The quality incentive payment would not be a separate payment stream, but rather it would alter a hospital's effective discount factor used to calculate its target prices. Depending on a participant hospital's quality performance, in performance years 1, 4, and 5, the quality incentive payments could result in effective discount factors ranging from 3 percent to 1.5 percent. In performance years 2 and 3, the quality incentive payments could result in effective discount factors for purposes of calculating reconciliation payments ranging from 3 percent to 1.5 percent, and for purposes of calculating repayment amounts from 2 percent to 0.5 percent. We note that the lower effective discount factors for calculating repayment amounts in performance years 2 and 3 reflect the reduction by 1 percentage point in discount factor to phase in downside risk.

If hospitals' quality performance during the CJR model mirrors historical quality performance, we expect the majority of the participant hospitals to qualify for an effective discount factor of 2 percent each performance year for purposes of reconciliation payment calculations, the same discount factor proposed for all participant hospitals in the proposed rule. By using a range of discount factors, we will offer more participant hospitals an opportunity to qualify for reconciliation payments, and we will be able to better reward the highest quality participant hospitals.

We refer readers to responses to comments in section III.C.5 of this final rule for more details on quality incentive payments, effective discount factors, the link between quality and payment, and how participant hospitals may perform based on historical quality performance.

Comment: Some commenters recommended that we not apply a discount factor to any hospital because it would effectively function as a rate cut for MS-DRGs 469 and 470. Some of these commenters suggested we could achieve savings using a shared savings methodology (for example, participant hospitals would receive 50 percent of actual episode performance below undiscounted target prices, and would repay 50 percent of actual episode performance above undiscounted target prices).

Response: We disagree with commenters that a discount factor is the equivalent of a rate cut. We are providing participant hospitals the opportunity to qualify for reconciliation payments for delivering high quality and efficient care for LEJR episodes, and reconciliation payments may likely exceed the value of the discount factor.

The discount factor will serve as Medicare's portion of reduced expenditures from the CJR episode. We acknowledge that there are other potential mechanisms, including shared savings methodologies, to provide savings to Medicare while also incentivizing participant hospitals. However, we also believe that a discount model, as proposed, can also incentivize participant hospitals to deliver high quality and efficient care while also providing savings to Medicare. We appreciate commenters' suggestions and we may consider alternative methodologies, such as shared savings, in the future.

Comment: Several commenters requested that we not apply a discount factor to hospitals that are already efficient because they would not be able to achieve further efficiencies. It would be challenging for these efficient hospitals to qualify for reconciliation payments if benchmarked against a target price that incorporates a discount factor.

Response: Commenters' concerns could be valid if we were basing target prices only on hospital-specific episode expenditure data. However, because we are blending hospital-specific and regional components in the target price calculation, and transitioning to completely regional target prices by performance year 4, target prices for more efficient hospitals likely would be higher than what they would be under a hospital-specific only pricing approach. We believe that with the blending and transition to regional pricing, historically efficient and high quality participant hospitals have a significant opportunity to qualify for reconciliation payments. Additionally, as discussed in the response to comments in section III.C.5. of this final rule, we are modifying our proposal to provide lower effective discount factors used to calculate target prices for participant hospitals with better quality performance. Therefore, high quality participant hospitals will have a lower hurdle to overcome to qualify for reconciliation payments. We will continue to incorporate a discount percentage into the target price for every participant hospital, and we will use a reduced discount factor for participant hospitals with high quality performance, as stated previously in this section's responses to comments and in section III.C.5. of this final rule.

Comment: Commenters requested upfront investments to fund care delivery (for example, care coordination), infrastructure, and quality reporting changes that participant hospitals may need to make, similar to how some ACOs use upfront investments in other models and programs (for example, an initiative similar to the ACO Investment Model for Medicare Shared Savings Program participants). Commenters suggested we fund these upfront investments in a number of ways, including the following: a supplemental lump sum payment at the start of the model; increase, instead of discount, historical episode expenditures by 2 percent; or transition in an increasing discount factor, getting to 2 percent by the end of the model.

Response: We thank the commenter for the suggestion and for recognizing the importance of potential care delivery, infrastructure, and quality reporting changes participant hospitals may need to make for an episode-based payment model such as CJR. However, we do not believe that an additional upfront payment mechanism such as a per-beneficiary-per-month payment or an additional payment per episode will be necessary for hospitals to Start Printed Page 73356successfully participate in this model. In BPCI, a similar episode-based payment model, participants have been able to improve episode expenditure performance without such additional upfront payment mechanisms.

Additionally, we believe there may be low investment opportunities for participant hospitals to achieve high quality and efficiency and qualify for reconciliation payments in this model. For example, participant hospitals may refer to high quality and efficient PAC providers when appropriate, and updates to discharge and referral patterns may be informed using already publicly available quality data and historical episode expenditure data provided by CMS and discussed in section III.E. of this final rule. PAC expenditures account for a significant proportion of historical CJR episode expenditures (approximately 30 percent [40] ), and changes to discharge and referral patterns could have significant impact on participant hospitals' actual episode expenditure performance. We note that this rationale may not hold true for other models (for example, patient-centered medical homes, ACOs) where providers are responsible for beneficiaries' cost of care over a longer period of time.

We also reiterate that as discussed in section III.C.5.b. of this final rule, the quality measures selected for this model are already in use for mandatory CMS quality reporting programs, such as the IQR program. Hospitals will not experience an additional reporting burden under this model for such measures. In addition, while we are including testing of a voluntary patient-reported outcomes measures, as discussed in section III.C.5.b.2. of this final rule, reporting of this measure will be voluntary. We do not believe there is any required additional burden on participant hospitals to report quality data.

Given the success of participants in a similar model, the possibility to achieve reconciliation payments with relatively low investment approaches, and the lack of required additional quality reporting burden, we will not make additional upfront payments through mechanisms such as per-beneficiary-per-month payments or additional payments per episode.

Final Decision: After consideration of the public comments we received, we are modifying our proposal to use a composite score methodology to link quality and payment in the CJR model. With this composite score methodology, a participant hospital may qualify for a reconciliation payment and for different effective discount factors depending on its quality performance. We refer readers to section III.C.5. of this final rule more details on how quality and payment will be linked.

c. Approach To Combine Pricing Features

In section III.C.4.(b) of the proposed rule we discuss the various features we proposed to incorporate into our approach to set target prices. We refer readers to that section for more information on rationale and alternatives considered for each feature. In this section we discuss how the different pricing features, as well as the episode definition (section III.B. of the proposed rule) and adjustments to payments included in the episodes (section III.C.3. of the proposed rule), would fit together and be sequenced to calculate CJR episode target prices for participant hospitals. The following steps would be used to calculate MS-DRG 469 and 470 anchored episode target prices for both January 1 through September 30 and October 1 through December 31 each performance year. The output of each step would be used as the input for the subsequent step, unless otherwise noted.

  • (1) Calculate historical CJR episode payments for episodes that were initiated during the 3-historical-years (section III.C.4.b.(2) of the proposed rule) for all CJR eligible hospitals for all Medicare Part A and B services included in the episode. We note that specific Per Beneficiary Per Month (PBPM) payments may be excluded from historical episode payment calculations as discussed in section III.C.7.d. of the proposed rule.
  • (2) Remove effects of special payment provisions (section III.C.3.a. of the proposed rule).
  • (3) Prorate Medicare payments for included episode services that span a period of care that extends beyond the episode (section III.C.3.b of the proposed rule.).
  • (4) Normalize for hospital-specific wage adjustment variation by dividing the episodes outputted in step (3) by the hospital's corresponding wage normalization factor described in section III.C.4.b.(7) of the proposed rule.
  • (5) Trend forward 2 oldest historical years of data to the most recent year of historical data. As discussed in section III.C.4.b.(3) of the proposed rule, separate national trend factors would be applied to episodes anchored by MS-DRG 469 versus MS-DRG 470.
  • (6) Cap high episode payment episodes with a region and MS-DRG anchor-specific high payment ceiling as discussed in section III.C.3.c. of the proposed rule, using the episode output from the previous step. We have posted region-specific historical average episode payments on the CJR Web site at http://innovation.cms.gov/​initiatives/​CJR/​. Note that these historical average episode payments were based on our proposed policies and do not represent actual target prices or the regional portion of actual target prices under the model.
  • (7) Calculate anchor factor and participant hospital-specific weights (section III.C.4.b.(8) of the proposed rule) using the episode output from the previous step to pool together MS-DRG 469 and 470 anchored episodes, resulting in participant hospital-specific pooled historical average episode payments. Similarly, calculate region-specific weights to calculate region-specific pooled historical average episode payments.
  • (8) Calculate participant hospital-specific and region-specific weighted update factors as described in section III.C.4.b.(4) of the proposed rule. Multiply each participant hospital-specific and region-specific pooled historical average episode payment by its corresponding participant hospital-specific and region-specific weighted update factors to calculate participant hospital-specific and region-specific updated, pooled, historical average episode payments.
  • (9) Blend together each participant hospital-specific updated, pooled, historical average episode payment with the corresponding region-specific updated, pooled, historical average episode payment according to the proportions described in section III.C.4.b.(5) of the proposed rule. Participant hospitals that do not have the minimum episode volume across the historical 3 years would use 0.0 percent and 100 percent as the proportions for hospital and region, respectively.
  • (10) Reintroduce hospital-specific wage variations by multiplying the participant hospital-specific blended, updated, and pooled historical average episode payments by the corresponding hospital-specific wage normalization factor, using the hospital's IPPS wage index that applies to the hospital during the period for which target prices are being calculated (section III.C.4.b.(7) of the proposed rule).
  • (11) Multiply the appropriate discount factor, as discussed in section III.C.4.b.(9) of the proposed rule to each participant hospital's wage-adjusted, blended, updated, and pooled historical Start Printed Page 73357average episode payment. For performance years 1, 3, 4, and 5, two discount factors would be used, one if the hospital successfully submits data on the patient-reported outcomes measure proposed in section III.C.5. of the proposed rule, and one if the hospital does not successfully submit the data. For performance year 2, 4 discount factors would be used to account for the 4 combinations of the following: (a) Whether or not the hospital successfully submits data on the patient-reported outcomes measure; and (b) for the different discount factors proposed for purposes of calculating reconciliation payments vs. calculating repayment amounts. The result of this calculation would be the participant hospital-specific target prices for MS-DRG 470 anchored episodes.
  • (12) Multiply participant hospitals' target prices for MS-DRG 470 anchored episodes by the anchor factor (section III.C.4.b.(8) of the proposed rule) to calculate hospitals' target prices for MS-DRG 469 anchored episodes.

The previously stated twelve steps would be used to calculate target prices for episodes that begin between January 1 and September 30, as well as for episodes that begin between October 1 and December 31, for each performance year. The target price calculations for the two different time periods each performance year would differ by the IPPS wage index used in step (11) and the update factors used in step (8). By following these twelve steps, we would calculate target prices for each participant hospital for each performance year. We refer readers to section III.C.4.b. of the proposed rule for further details on each of the specific steps.

We sought comment on the proposed approach to sequence and fit together the different pricing features, the episode definition (section III.B. of the proposed rule), and adjustments to payments included in the episodes (section III.C.3. of the proposed rule) to calculate CJR episode target prices for participant hospitals.

The following is a summary of the comments received and our responses.

Comment: Many commenters requested for risk adjustment based on patients' hip fracture status, among other clinical and demographic dimensions. Commenters also recommended that we modify the definition of “CJR eligible hospitals”, the term used to identify hospitals included in calculations for the regional component of target prices, to not exclude hospitals that are participating in BPCI.

Response: We refer readers to comments and responses to comments in sections III.C.4.b.(1) and III.C.4.b.(4) of the final rule for further discussion on risk stratification and CJR eligible hospitals, respectively. We reference them here because changes to risk stratification and CJR eligible hospitals would impact how we would combine CJR pricing features. Given the changes to the proposed rule described in sections III.C.3, III.C.4.b, and III.C.5, we are modifying the different pricing features would fit together and be sequenced to calculate CJR episode target prices for participant hospitals. The following steps would be used to calculate different target prices in each performance year for each combination of anchor MS-DRG (469 vs. 470), hip fracture status (with hip fracture vs. without hip fracture), and period during which target prices are applicable within a performance year (episodes initiated January 1 through September 30 vs. October 1 through December 31 each performance year). The output of each step would be used as the input for the subsequent step, unless otherwise noted.

  • (1) Calculate historical CJR episode payments for episodes that were initiated during the 3-historical-years (section III.C.4.b.(2) of this final rule) for all CJR eligible hospitals for all Medicare Part A and B services included in the episode. We note that specific PBPM payments may be excluded from historical episode payment calculations as discussed in section III.C.7.d. of this final rule.
  • (2) Remove effects of special payment provisions (section III.C.3.a. of this final rule) and normalize for wage index differences (section III.C.4.b.(7) of this final rule) by standardizing Medicare FFS payments at the claim-level.
  • (3) Prorate Medicare payments for included episode services that span a period of care that extends beyond the episode (section III.C.3.b of this final rule.).
  • (4) Trend forward 2 oldest historical years of data to the most recent year of historical data. As discussed in section III.C.4.b.(3) of this final rule, separate national trend factors would be applied for each combination of anchor MS-DRG (469 vs. 470) and hip fracture status (with hip fracture vs. no hip fracture).
  • (5) Cap high episode payment episodes with a region and MS DRG anchor specific high payment ceiling as discussed in section III.C.3.c. of this final rule, using the episode output from the previous step. We have posted region specific historical average episode payments on the CJR final rule Web site at http://innovation.cms.gov/​initiatives/​CJR/​.
  • (6) Calculate anchor factor and participant hospital specific weights (section III.C.4.b.(8) of this final rule) using the episode output from the previous step to pool together MS DRG 469 and 470 anchored episodes with and without hip fracture, resulting in participant hospital specific pooled historical average episode payments. Similarly, calculate region specific weights to calculate region specific pooled historical average episode payments.
  • (7) Calculate participant hospital specific and region specific weighted update factors as described in section III.C.4.b.(4) of this final rule. Multiply each participant hospital specific and region specific pooled historical average episode payment by its corresponding participant hospital specific and region specific weighted update factors to calculate participant hospital specific and region specific updated, pooled, historical average episode payments.
  • (8) Blend together each participant hospital specific updated, pooled, historical average episode payment with the corresponding region specific updated, pooled, historical average episode payment according to the proportions described in section III.C.4.b.(5) of this final rule. Participant hospitals that do not have the minimum episode volume across the historical 3 years would use 0.0 percent and 100 percent as the proportions for hospital and region, respectively. For purposes of this final rule, we will define the output of this step as the pre-discount target price for MS DRG 470 anchored episodes without hip fracture.
  • (9) Multiply the output of step (8) by the appropriate anchor factors (step (6) of this target price calculation process and detailed in section III.C.4.b.(8) of this final rule) for MS DRG 469 anchored episodes with hip fracture, MS DRG 469 anchored episodes without hip fracture, and MS DRG 470 anchored episodes with hip fracture. For purposes of this final rule, we will define the outputs of this step as the pre-discount target prices for MS DRG 469 anchored episodes with hip fracture, MS DRG 469 anchored episodes without hip fracture, and MS DRG 470 anchored episodes with hip fracture.
  • (10) Multiply the pre-discount target prices for MS DRGs 469 and 470 episodes with and without hip fracture by the appropriate effective discount factor that incorporates any quality incentive payment, as briefly described in section III.C.4.b.(9) of this final rule Start Printed Page 73358and more specifically detailed in the response to comments in section III.C.5. of this final rule and Tables 19, 20, and 21. The results of these calculations will be participant hospitals' target prices for MS DRG 469 anchored episodes with hip fracture, MS DRG 469 anchored episodes without hip fracture, MS DRG 470 anchored episodes with hip fracture, and MS DRG 470 anchored episodes without hip fracture.

The previously stated 10 steps will be used to calculate target prices for episodes that begin between January 1 and September 30 (between April 1 and September 30 for performance year 1), as well as for episodes that begin between October 1 and December 31, for each performance year. The target price calculations for the two different time periods each performance year will differ by the update factors used in the seventh step. By following these ten steps, we will calculate target prices for each participant hospital for each performance year. We refer readers to section III.C.4.b. of this final rule for further details on each of the specific steps.

Final Decision: After consideration of the public comments we received, we are modifying our proposal to incorporate changes described in sections III.C.3, III.C.4.b, and III.C.5 of this final rule when fitting together and sequencing episode target price features used to calculate CJR episode target prices for participant hospitals.

These final policies are set forth at § 510.300 and § 510.305.

5. Use of Quality Performance in the Payment Methodology

a. Background

Over the past several years Medicare payment policy has moved away from FFS payments unlinked to quality and towards payments that are linked to quality of care. Through the Affordable Care Act, we have implemented specific IPPS programs like the HVBP program (subsection (o) of section 1886 of the Act), the Hospital Acquired Condition Reduction Program (HACRP) (subsection (q) of section 1886) and the HRRP (subsection (p) of section 1886), where quality of care is linked with payment. We have also implemented the Shared Savings Program, an ACO program that links shared savings payment to quality performance. Since the implementation of the HRRP in October 2012, readmission rates for various medical conditions like THA and TKA (THA/TKA) have improved. Trend analyses show a decrease in readmission rates and specifically with THA/TKA risk-standardized readmissions rates (RSRR) from 5.4 percent (July 2010-June 2011) to 4.8 percent (July 2012-June 2013).[41] Additionally, hospital THA/TKA RSCR decreased from 3.4 percent (April 2010 through March 2011) to 3.1 percent (April 2012 through March 2013). Despite the downward trend of THA/TKA RSRRs and RSCRs, the wide dispersion in these readmission rates suggests there is still room for hospitals to improve their performance on these measures as illustrated by a THA/TKA RSRR distribution of 2.8 to 9.4 percent (July 2010-June 2013) and a THA/TKA RSCR distribution of 1.5 to 6.4 percent (April 2010-March 2013). In the proposed rule, we stated our belief that the CJR model would provide another mechanism for hospitals to improve quality of care, while also achieving cost efficiency. Incentivizing high-value care through episode-based payments for LEJR procedures is a primary objective of CJR. Therefore, incorporating quality performance into the episode payment structure is an essential component of the CJR model. We also stated our belief that the financial opportunity discussed in section III.C.2. of the proposed rule would provide the appropriate incentives necessary to reward a participant hospital's achievement of episode savings when the savings are greater than the discounted target price. For the reasons stated previously, we discussed our belief that it would be important for the CJR model to link the financial reward opportunity with achievement in quality of care for Medicare beneficiaries undergoing LEJR.

As discussed in section III.C. of this final rule, which outlines the payment structure proposed for the CJR model, each participant hospital would have target prices calculated for MS-DRG 469 and 470 anchored episodes; each anchored episode would include an anchor hospitalization for an LEJR procedure and a 90-day period after the date of discharge from the anchor hospitalization. These episode target prices represent expected spending for all related Part A and Part B spending for such episodes, with a discount applied. Hospitals who achieve actual episode spending below a target price for a given performance period would be eligible for a reconciliation payment from CMS, subject to the proposed stop-gain limit policy as discussed in section III.C.8. of this final rule.

In the proposed rule, we proposed quality performance standards that must also be met in order for a hospital to be eligible to receive a reconciliation payment under CJR. Specifically, we described our proposal to include a performance measure result threshold on select outcomes-based quality measures as a requirement for participants to receive a reconciliation payment if actual episode spending is less than the target price under CJR in a performance year, in addition to a payment adjustment for successful reporting of a voluntary measure in development. Beginning in performance year one and continuing throughout the duration of the model, we proposed to make reconciliation payments only to those CJR hospital participants that met or exceeded a minimum measure result threshold. We also discussed an alternative approach to determining CJR reconciliation payment eligibility and adjusting payment based on a quality score developed from performance on three outcomes-based quality measures and success in reporting the voluntary measurement in development.

b. Implementation of Quality Measures in the Payment Methodology

In section III.D. of the proposed rule, we proposed three measures to assess quality of care of the hospitals participating in the CJR model. We also proposed voluntary data submission for a patient-reported outcome measure in development. In section III.C.5. of the proposed rule, we proposed using three measures to determine eligibility for a reconciliation payment, as well as proposed rewarding hospitals that voluntarily submit data for the patient-reported outcome measure. We also discussed an alternative approach to determining reconciliation payment eligibility and adjusting payment based on a composite quality score calculated from the three required outcome measures and success on reporting voluntary data on the patient-reported outcome measure.

(1) General Selection of Quality Measures

The CJR model is designed to provide financial incentives to improve coordination of care for beneficiaries that we expect to lead to avoidance of post-surgical complications and hospital readmissions, as well as to improve patient experience through care redesign and coordination. Furthermore, we acknowledge that achievement of savings while ensuring high-quality care for Medicare FFS beneficiaries in LEJR episodes would require close collaboration among hospitals, Start Printed Page 73359physicians, PAC providers, and other providers and suppliers. In order to encourage care collaboration among multiple providers of patients undergoing THA and TKA, we proposed three measures, as described in detail in section III.D.2. of this final rule, to determine hospital quality of care and to determine eligibility for a reconciliation payment under the CJR model. The measures we proposed are as follows:

  • Hospital-level 30-day, all-cause RSRR following elective primary THA and/or TKA (National Quality Forum (NQF)#1551), an administrative claims-based measure.
  • Hospital-level RSCR following elective primary THA and/or TKA (NQF #1550), an administrative claims-based measure.
  • HCAHPS Survey measure (NQF #0166).

Beginning in performance year 1 and continuing throughout the duration of the model, we proposed to make reconciliation payments only to those CJR participant hospitals that met or exceeded a minimum performance threshold on the measures previously listed. We proposed that hospitals must meet or exceed the measure reporting thresholds and other requirements described in section III.C.5 and III.D. of this final rule on all three measures in order to be eligible for a reconciliation payment.

These three outcome measures were chosen due to their: (1) Alignment with the goals of the CJR model; (2) hospitals' familiarity with the measures due to their use in other CMS hospital quality programs, including programs that tie payment to performance such as HVBP and HRRP; and (3) assessment of CMS priorities to improve the rate of LEJR complications and readmissions, while improving patient experience. In the proposed rule, we stated our belief that the three quality measures we proposed for reconciliation payment eligibility reflected these goals and accurately measured hospitals' level of achievement on such goals.

(2) Adjustment to the Payment Methodology for Voluntary Submission of Data for Patient-Reported Outcome (PRO) Measure

During our consideration of quality metrics for the CJR model, we examined the feasibility of linking voluntary data submission of patient-reported outcomes, beyond the current three required measures discussed in section III.D.2. of this final rule for use in the model, with the possibility of incentivizing participant hospitals under the episode payment model to participate in this voluntary submission of data. We specifically examined potential patient-reported outcome measures since this type of outcome measure aligns with the CJR model goal of improving LEJR episode quality of care, including a heightened emphasis on patient-centered care where patients provide meaningful input to their care. Furthermore, the availability of patient reported outcome data would provide additional information on a participant hospital's quality performance, especially with respect to a patient's functional status, beyond the current three required measures discussed in section III.D.2. of this final rule for use in the model. We noted that we have a measure in development, the Hospital-Level Performance Measure(s) of Patient-Reported Outcomes Following Elective Primary THA or TKA measure or both (hence forth referred to as “THA/TKA patient-reported outcome-based measure”), that would support the National Quality Strategy domain of patient and family engagement, and could capture meaningful information that would not otherwise be available on patient outcomes that are related to the quality of LEJR episodes under CJR. In the proposed rule, we stated our belief that incorporating this measure into CJR by adjusting the payment methodology for successful voluntary data submission on the THA/TKA patient-reported outcome-based measure (henceforth referred to as “THA/TKA voluntary data”) would provide participant hospitals with valuable information on functional outcomes that would assist them in assessing an important patient-centered outcome, engaging other providers and suppliers in care redesign for LEJR episodes, as well as provide them with the potential for greater financial benefit from improved LEJR episode efficiencies. We did not believe it would be appropriate at this time to hold any participant hospitals financially accountable for their actual THA/TKA voluntary data, as we proposed to require for the three measures described in section III.C.5.b.(5) of this final rule.

Instead, we proposed to adjust the episode payment methodology for participant hospitals that successfully submit THA/TKA voluntary data by reducing the discount percentage used to set the target price from 2.0 percent to 1.7 percent of expected episode spending based on historical CJR episode data, hereinafter referred to as the voluntary reporting payment adjustment. The proposed payment policies with respect to reconciliation payment eligibility and the discount percentage based on hospital voluntary data submission are summarized in Table 10 for performance years 3 through 5 where we proposed that hospitals have full repayment responsibility. The proposed specific percentages that would apply for purposes of the repayment amount and reconciliation payment are outlined for performance years 1 and 2 in the discussion that follows.

Table 10—Proposed Reconciliation Payment Eligibility and Discount Percentage Included in the Target Price for Each Participant Hospital Based on Quality Performance in Performance Years 3 Through 5

Discount percentage included in target price/reconciliation payment eligibilityMeets thresholds for all 3 required quality measuresDoes not meet thresholds for one or more of 3 required quality measures
Successfully submits THA/TKA voluntary data1.7%/eligible1.7%/ineligible.
Does not successfully submit THA/TKA voluntary data2.0%/eligible2.0%/ineligible.

We refer readers to section III.D.3.a. of this final rule for further discussion of the THA/TKA patient-reported outcome-based measure and our proposed definition of successful reporting. In addition, we refer readers to section III.C.4.b.(9) of this final rule for discussion of the proposed discount of 2.0 percent (without the voluntary reporting payment adjustment) to establish the target price. In the proposed rule, we stated our belief that a voluntary reporting payment adjustment of 0.3 percent of expected episode spending would, on average, cover the participant hospitals' additional administrative costs of voluntarily reporting patient risk variables and patient-reported reported function for outcome calculation. We estimated the value of this discount Start Printed Page 73360reduction, on average, to be about $75 per LEJR episode at a participant hospital, which we believed would be sufficient to pay hospitals for the resources required to survey beneficiaries pre- and post-operatively about functional status and report this information required for measure development to CMS. We also believed that voluntary reporting on this patient-reported outcome measure would be integral to implementation of the CJR model, as it would allow us to further develop and evaluate the measure for potential use in this model in the future as a measure of quality that is important and not captured in any other available measures.

We proposed that the voluntary reporting payment adjustment would be available for all years of the model, unless we find the measure to be unfeasible or have adequately developed the measure such that continued voluntary data collection is no longer needed for measure development during the course of the model. In those situations, we would notify participant hospitals that the voluntary reporting payment adjustment was no longer available as we would cease collecting the data.

We proposed that when we provide the episode target price to each participant hospital at 2 times during the performance year, we would provide different target prices reflecting the 2.0 percent and 1.7 percent discounts. At the time of reconciliation for the performance year, we would determine which participant hospitals successfully reported the THA/TKA voluntary data for that performance year. The effects of this voluntary reporting payment adjustment would vary for each year of the model, depending on the proposed reconciliation payment and repayment policies for that performance year. For hospitals that achieved successful reporting of the THA/TKA voluntary data in performance year 3, 4, or 5, we would use the target price reflecting the 1.7 percent discount (compared with the 2.0 percent discount for nonreporting or unsuccessfully reporting hospitals) to calculate the hospital's reconciliation payment or repayment amount. Based on this comparison, consistent with the proposal described in section III.C.6. of this final rule, we would make a reconciliation payment if actual episode spending was less than the target price (and the thresholds for reconciliation payment eligibility are met for the three required quality measures) or make participant hospitals responsible for repaying Medicare if actual episode spending exceeded the target price. For performance year 2, when we proposed that repayment responsibility would be phased-in, for participant hospitals with successful THA/TKA voluntary data reporting, we would use a target price reflecting the 1.7 percent discount (compared with the 2.0 percent discount for nonreporting or unsuccessfully reporting hospitals) to determine if actual episode spending was below the target price, whereupon the participant hospital would receive a reconciliation payment if the quality thresholds on the three required measures were met. In order to help hospitals transition to taking on repayment responsibility, we proposed to apply a reduced discount of 0.7 percent for successful THA/TKA voluntary data reporting hospitals (compared with 1.0 percent for nonreporting or unsuccessfully reporting hospitals) in performance year 2 for purposes of determining the hospital's repayment responsibility for excess episode spending. For performance year 1, when we proposed that there would be no repayment responsibility, for participant hospitals with successful THA/TKA voluntary data reporting, we would use a target price reflecting the 1.7 percent discount (compared with the 2.0 percent discount for nonreporting or unsuccessfully reporting hospitals) to determine if actual episode spending was below the target price, whereupon the participant hospital would receive a reconciliation payment if the quality thresholds on the three required measures were met. In the proposed rule, we stated our belief that this proposed voluntary reporting payment adjustment would provide the potential for increased financial benefit for participant hospitals due to a higher target price (that reflects a lower discount percentage) that successfully report the measure. Participant hospitals that successfully reported the voluntary data would be subject to a lower repayment amount (except for performance year 1 when hospitals have no repayment responsibility) or a higher reconciliation payment (assuming the thresholds are met on the three required measures for reconciliation payment eligibility), than hospitals that did not successfully report the voluntary data.

In general, we proposed that participant hospitals that met the performance thresholds for the three required quality measures and reduced actual episode spending below the target price, as well as successfully reported the THA/TKA voluntary data, would be eligible to retain an additional 0.3 percent of the reduced episode expenditures relative to participant hospitals that successfully reported the three required quality measures but did not report voluntary data, funds which would offset additional administrative costs that the participant hospitals would incur in reporting on the measure. Additionally, for performance years 2-5 where we proposed that participant hospitals would have payment responsibility, participant hospitals with increased actual episode spending above the target price would not be required to repay 0.3 percent of the increased episode expenditures (relative to participant hospitals that do not report voluntary data), funds that would offset additional administrative costs that the participant hospitals would incur in reporting on the measure. These costs would include the hospital staff time required for training on the measure, as well as then gathering and reporting on multiple patient risk variables from LEJR episode beneficiaries' medical records and locating beneficiaries and administering via phone survey questions on functional status, which would also then be reported to CMS. Thus, we expected that the proposal would encourage reporting by a number of participant hospitals, and it would have the potential to benefit those hospitals that successfully reported on the measure. Therefore, this proposal could financially benefit reporting hospitals that would also collect valuable information on patient functional outcomes that could inform their LEJR care redesign. While this measure remains in development from our perspective to ensure translation of data across care settings and the respective hospital communities during the 90-day post-discharge episode of care, participant hospitals would gain anecdotal, locally relevant information regarding the patient-reported outcomes of their own patients that could inform participant hospitals' continuous quality improvement efforts.

We considered two alternative options to adjust the CJR payment methodology by modifying the required quality measure thresholds for reconciliation payment eligibility for those participant hospitals that successfully submit the THA/TKA voluntary data. First, we considered adjusting the threshold that hospitals must meet on the three required quality measures for reconciliation payment eligibility if reduced episode spending was achieved from the unadjusted 30th percentile threshold to the adjusted 20th percentile threshold for performance years 1, 2, and 3, and from the unadjusted 40th percentile to the adjusted 30th percentile for performance years 4 and 5. Second, we Start Printed Page 73361considered only requiring hospitals to meet the 30th percentile threshold on two of three outcome measures for performance years 1, 2, and 3, and the 40th percentile threshold on two of three outcome measures for performance years 4 and 5. These options would provide the opportunity for some participant hospitals, specifically those that missed the unadjusted percentile for one or more of the three required quality measures by a specified margin, to receive reconciliation payments if actual episode spending was less than the target price. However, these options could benefit only a subset of participant hospitals that successfully reported the THA/TKA voluntary data. For the majority of participant hospitals that we expect would meet the unadjusted thresholds for all three required measures, these options would not provide any incentive to voluntarily report the data because the hospitals would not benefit from voluntarily reporting the additional measure. We decided not to propose either of these options to adjust the CJR payment methodology for participant hospitals that voluntarily report data on the new measure because the limited benefit could result in few hospitals choosing to report on the measure, thereby limiting our progress in developing the measure. We noted that these two considered options and our proposal were not mutually exclusive.

We sought comment on the proposed voluntary reporting payment adjustment of reducing the discount percentage from 2.0 percent to 1.7 percent for CJR participant hospitals that voluntarily and successfully report on the THA/TKA voluntary data. Given our interest in robust hospital participation in reporting on the THA/TKA voluntary data under CJR, we were specifically interested in information on the additional resources and their associated costs that hospitals would incur to report THA/TKA voluntary data, as well as the relationship of these costs to the potential financial benefit participant hospitals could receive from the proposed reduced discount of 1.7 percent. Based on such information, we would consider whether a change from the proposed discount factor reduction due to successful voluntary data submission would be appropriate. We also sought comment on whether the alternative payment methodology adjustments considered, or combination of adjustments, would more appropriately incentivize CJR participant hospitals to submit THA/TKA voluntary data. In the proposed rule, we stated our belief that development of the THA/TKA patient-reported outcome measure would benefit from reporting by a broad array of participant hospitals, including those that currently deliver high quality, efficient LEJR episode care and those that have substantial room for improvement on quality and or cost-efficiency.

We summarize the public comments we received on the proposed voluntary reporting payment adjustment and provide our responses in section III.C.5.b.(5)(c)(iii) of this final rule. We did not receive public comments on the alternative payment methodology adjustments that we discussed in the proposed rule. Furthermore, in light of our interest in encouraging CJR participant hospital THA/TKA voluntary data reporting, we also considered alternative approaches to collect this information or provide hospitals with funds to help cover their associated administrative costs other than adjustments to the CJR model payment methodology. One alternative would be for hospitals to collect and report on patient pre-operative information collected 0 to 90 days before surgery, while CMS would engage a contractor to collect and report the post-operative information collected 9 to 12 months after surgery. This approach would reduce some of the administrative burden of collection and reporting on hospitals, although participant hospitals would need to provide CMS with certain beneficiary information, including contact information that would be needed for a CMS contractor to contact the beneficiary at a later date. We sought comment on this alternative, including whether hospitals would incur significant additional administrative costs to report on the data prior to surgery and how CMS could best provide funds to offset some of those costs, through an adjustment to the CJR payment methodology or other means. We also sought comment on the information participant hospitals would need to provide to CMS so that a CMS contractor could collect and report the post-operative data, and the most efficient ways for hospitals to provide this information to us. Finally, we considered an approach that would provide hospitals with separate payment outside of an adjustment to the CJR payment methodology to specifically assist in covering their administrative costs of reporting THA/TKA voluntary data, in order to achieve robust hospital participation in reporting. We sought comment on the hospital administrative costs that would be incurred for reporting, as well as on approaches we could take to ensure that hospitals achieved successful reporting under such an approach if separate payment was made. Finally, we expressed our interest in comments regarding the comparative strength of these various alternatives in encouraging hospitals to participate in reporting THA/TKA voluntary data.

We did not receive any public comments on the alternatives we discussed other than adjustments to the payment methodology to collect THA/TKA voluntary data and provide hospitals with funds to cover the required resources. We summarize these comments we received in section III.C.5.b.(5)(c)(iii) of this final rule and provide our responses.

(3) Measure Risk-Adjustment and Calculations

All three proposed outcome measures are risk-adjusted, and we refer readers to section III.D.2. of this final rule for a full discussion of these measures and risk-adjustment methodologies. We believed that risk-adjustment for patient case-mix is important when assessing hospital performance based on patient outcomes and experience and understanding how a given hospital's performance compares to the performance of other hospitals with similar case-mix.

(4) Applicable Time Period

We proposed to use a 3-year rolling performance or applicable period for the Hospital-level 30-day, all-cause RSRR following elective primary THA and/or TKA (NQF #1551) and the Hospital-level RSCR following elective primary THA and/or TKA (NQF #1550) measures. We also specifically proposed to align with the HIQR Program's 3-year rolling performance period for the RSRR and RSCR measures since we believed that a 3-year performance period yields the most consistently reliable and valid measure results (FY 2015 IPPS/LTCH final rule, 79 FR 50208 through 50209). For the HCAHPS Survey measure, we proposed to follow the same performance period as in the HIQR Program (FY 2015 IPPS/LTCH final rule, 79 FR 50259). HCAHPS scores are created from 4 consecutive quarters of survey data; publicly reported HCAHPS results are also based on 4 quarters of data. For the voluntary data collection for the proposed THA/TKA patient-reported outcome-based performance measure, the optimal reporting time period had not been determined at the time of issuance of the CJR model proposed rule. Therefore, we proposed defining the applicable time period as Start Printed Page 7336212 month intervals that may begin between July 1, 2016 and December 31, 2016, and continue in subsequent performance years for a total of four or fewer performance periods. Participant hospitals will submit required data to CMS in a mechanism similar to the data submission process for the HIQR Program within sixty days of the end of each 12 month period. As described in section III.C.5.b.(2) of the proposed rule, the proposed voluntary reporting payment adjustment of reducing the discount percentage from 2.0 percent to 1.7 percent for CJR participant hospitals that successfully report on the THA/TKA voluntary data would begin in year 2 and also apply to subsequent years of the model. We are not finalizing the proposed voluntary reporting payment adjustment, as discussed further in section III.C.5.b.(5)(c)(iii) of this final rule. We note that we summarize the public comments we received on the proposed applicable time period and provide our responses in section III.D.3.d. of this final rule, and we summarize the public comments we received on the reporting time period for the THA/TKA patient-reported outcome and limited risk variable data and provide our responses in section III.D.3.a.(9) of this final rule.

(5) Criteria for Applicable Hospitals and Performance Scoring

(a) Identification of Participant Hospitals for the CJR Model

As discussed in section III.A.2. of this final rule, all CJR participant hospitals will be IPPS hospitals.

(b) Methodology To Determine Performance on the Quality Measures

To determine performance on the quality measures, we proposed to calculate measure results for all three measures as outlined in the Quality Measures section III.D.2. of this final rule. Performance on the three measures for the CJR model participant hospitals would be compared to the national distribution of measure results for each of these measures obtained through the HIQR Program. The HIQR Program is an IPPS program in which public reporting is a focus of the program for the nation's acute care hospitals, and we proposed using the absolute value of the CJR model participant hospital's result to determine if that participant hospital was eligible for a reconciliation payment. In essence, we intended to take the HIQR Program measure results (also posted publicly) for the proposed measures, identify the proposed threshold, and apply the thresholds as outlined in section III.C.5.b.(5)(c)(iii) of this final rule. In the proposed rule, we stated our belief that it would be reasonable to use the HIQR Program distribution of measure results to identify a measure result threshold because—(1) The hospitals in the HIQR Program represent most acute care hospitals in the nation; (2) the CJR model participant hospitals are a subset of the hospitals in the HIQR Program; and (3) the expectation that the CJR model participant hospitals meet a measure result threshold based on a national distribution of measure results would encourage the CJR model participant hospitals to strive to attain measure results consistent with or better than hospitals across the nation. For a detailed description of how we proposed to determine the measure result thresholds for consideration of a reconciliation payment adjustment, see section III.C.5.b.(3) and III.C.5.b.(5)(c) of this final rule. We would not want to encourage CJR model participant hospitals to strive for measure results or quality of care performance that may be lower than the national measure results. Given that the CJR participant hospitals are a subset of the HIQR Program participant hospitals, they are familiar with these three measures and may have put into place processes that will help to improve quality of care in the LEJR patient population. Finally, once the measure results were calculated, we proposed to use these results to determine eligibility for reconciliation payment, which is discussed in detail in the next section.

We summarize the public comments we received on the proposed calculation of the measure results and application of performance thresholds and provide our responses in sections III.D.2 and III.C.5.b.(5)(c)(iii) of this final rule, respectively.

To be considered to have successfully reported the voluntary data collection and submission for the THA/TKA voluntary data, we proposed that successfully reporting would mean participant hospitals must meet all of the following:

  • Submit the data elements listed in section III.D.3.a.(2) of this final rule.
  • Data elements listed in section III.D.3.a.(3) of this final rule must be submitted on at least 80 percent of their eligible elective primary THA/TKA patients (patients eligible for pre-operative THA/TKA voluntary data submission are those described in section III.D.3.a.(3) of this final rule); patients eligible for post-operative THA/TKA voluntary data submission are those described in section III.D.3.a.(3) of this final rule and also having a THA/TKA procedure date during the anchor hospitalization at least 366 days prior to the end of the data collection period. Therefore, participant hospitals would not be expected to collect and submit post-operative THA/TKA voluntary data on patients who are fewer than 366 days from the date of surgery.
  • THA/TKA voluntary data submission must occur within 60 days of the end of the most recent performance period.

Hospitals that meet these three standards and successfully submit THA/TKA voluntary data would be eligible for the proposed voluntary reporting payment adjustment of reducing the discount percentage from 2.0 percent to 1.7 percent for CJR participant hospitals that voluntarily and successfully report on the THA/TKA voluntary data. We note that we are not finalizing this voluntary reporting payment adjustment proposal as discussed in section III.C.5.b.(5)(c)(iii) of this final rule. However, we continue to believe that encouraging collection and submission of the THA/TKA voluntary data through the CJR model would increase availability of patient-reported outcomes to both participant hospitals that collect and submit data on their own patients in the model (and their patients as well); further development of an outcomes measure that provides meaningful information on patient-reported outcomes for THA/TKA procedures that are commonly furnished to Medicare beneficiaries; provide another quality measure that may be incorporated into the CJR model policy linking quality to payment in future performance years, pending successful development of the measure; and inform the quality strategy of future payment models. Collecting data on at least 80 percent of hospital's eligible THA/TKA patients would provide sufficiently representative data to allow for development and testing of the THA/TKA patient-reported outcome-based performance measure.

We invited public comment on the proposal to calculate measure results for all three measures as outlined in the Quality Measures section III.D.2. of this final rule. We also sought public comment on our proposal for hospitals to meet three requirements, previously outlined, in order to be considered as successfully submitting THA/TKA voluntary data.

We summarize the public comments on the proposals to calculate measure results and determine measure result thresholds and provide our responses in sections III.D.2. and III.C.5.b.(5)(c)(iii) of this final rule, respectively. We summarize the public comments on the proposals for successful THA/TKA Start Printed Page 73363voluntary data submission and provide our responses in section III.D.3.a. of this final rule.

(c) Methodology To Link Quality and Payment

(i) Background

In proposing a methodology for linking payment for LEJR episodes to quality under this model, we considered several alternatives. Specifically, we considered making reconciliation payments to hospitals tied to achievement and improvement in quality performance or, alternatively, establishing minimum quality performance thresholds for selected quality measures from the beginning of the model or a later year, which would reward achievement but not necessarily improvement. While we proposed as discussed section III.C.5.b.(5)(c) of this final rule to establish minimum thresholds for participant hospital performance on three selected quality measures for reconciliation payment eligibility each performance year from the beginning of the model, we also discussed in detail an alternative we considered, which would make quality incentive payments related to hospital achievement and improvement on the basis of a composite quality score developed for each performance year. The composite quality score would affect reconciliation payment eligibility and change the effective discount included in the target price experienced by a participant hospital at reconciliation.

Similar to the proposal described in section III.C.5.b.(5)(c) of this final rule, the alternatives considered would require a determination of participant hospital performance on all three proposed required quality measures, described in section III.D.2. of this final rule, based on the national distribution of hospital measure result performance, but instead of identifying the participant hospital's performance percentile for comparison with a threshold requirement, we would do so for purposes of assigning points toward a hospital composite quality score. Both the hospital-level 30-day, all cause Risk-Standardized Readmission Rate (RSRR) following elective primary THA and/or TKA (NQF #1551) measure and the hospital-level Risk-Standardized Complication Rate (RSCR) following elective primary THA and/or TKA (NQF #1550) measure directly yield rates for which a participant hospital performance percentile could be determined and compared to the national distribution in a straightforward manner. As discussed in section III.D.2.c. of this final rule, we proposed to use the HCAHPS Linear Mean Roll Up (HLMR) score calculated using the HCAHPS Survey measure (NQF #0166). Once the HLMR scores are calculated, the participant hospital performance percentile could also be determined and compared to the national distribution in a straightforward manner. In addition, the alternatives considered would account for the successful submission of voluntary THA/TKA data on the patient-reported outcome measure, as discussed in section III.C.5.b.(2) of this final rule, in the calculation of the composite quality score.

(ii) Alternatives Considered To Link Quality and Payment

We considered assigning each participant hospital a composite quality score, developed as the sum of the individual quality measure scores described later in this section, which were set to reflect the intended weights for each of the quality measures and the successful submission of THA/TKA voluntary data in the composite quality score. The participant hospital's composite quality score would affect reconciliation payment eligibility and could also provide the opportunity for quality incentive payments under the CJR model. Each quality measure would be assigned a weight in the composite quality score and possible scores for the measures would be set to reflect those weights. A composite quality score for each performance year would be calculated for each participant hospital based on its own performance that would affect reconciliation payment eligibility and the hospital's opportunity to receive quality incentive payments under the model. The composite quality score would also change the effective discount included in the target price experienced by the hospital at reconciliation for that performance year. We would weigh participant hospital performance on each of the three measures and successful submission of voluntary THA/TKA data according to the measure weights displayed in Table 11.

Table 11—Quality Measure Weights Under the Composite Quality Score Alternative Considered in the Proposed Rule

Quality measureWeight in composite quality score (%)
Hospital-level 30-day, all-cause RSRR following elective primary THA and/or TKA (NQF #1551)20
Hospital-level RSCR following elective primary THA and/or TKA (NQF #1550)40
HCAHPS Survey (NQF #0166)30
Voluntary THA/TKA data submission on patient-reported outcome measure10

We would assign the lowest weight of 10 percent to the successful submission of THA/TKA data on the patient-reported outcome measure because these data represent a hospital's meaningful participation in advancing the quality measurement of LEJR patient-reported outcomes but not actual outcome performance for LEJR episodes under the CJR model. In the proposed rule, we stated our belief the three required measures that represent LEJR outcomes deserve higher weights in the composite quality score. We would assign a modest weight of 20 percent to the readmissions measure because, while we believed that readmissions are an important quality measure for LEJR episodes, the episode payment methodology under the model already provides a strong financial incentive to reduce readmissions that otherwise would contribute significantly to greater actual episode payments. Furthermore, hospitals generally have already made significant strides over the past several years in reducing readmissions due to the inclusion of this measure in other CMS hospital programs that make payment adjustments based on performance on this measure. We believed that a higher weight than 20 percent would overvalue the contribution of readmissions performance as an indicator of LEJR episode quality in calculating the composite quality score. Furthermore, other CMS hospital programs may also make a payment adjustment based on hospital performance on the readmissions measure, so we would not want this measure to also strongly influence reconciliation payment eligibility and the opportunity for quality incentive payments under the Start Printed Page 73364CJR model. We would assign a higher weight of 30 percent to the HCAHPS Survey measure because we believed that incorporating this quality measure, which reflects performance regarding patients' perspectives on care, including communication, care transitions, and discharge information, is a highly meaningful outcome measure of LEJR episode quality under the CJR model. However, we did not believe it would be appropriate assign the HCAHPS Survey measure the highest weight of the four measures, as the measure is not specific to LEJR episode care, but rather to all clinical conditions treated by participant hospitals. Finally, we would assign the highest weight, 40 percent, to the complications measure. We believed this measure should be weighted the most because it is specific to meaningful outcomes for primary THA and TKA that are the major procedures included in LEJR episodes under the CJR model. The measure includes important complications of LEJR episodes, such as myocardial infarction, pneumonia, surgical site bleeding, pulmonary embolism, death, mechanical joint complications, and joint infections occurring within various periods of time during the LEJR episode. LEJR episodes under the CJR model are broadly defined so that reducing complications should be a major focus of care redesign that improves quality and efficiency under this model, yet because complications may not be as costly as readmissions, the payment incentives under the model would not as strongly target reducing complications as reducing readmissions. We sought comment on this weighting of the individual quality scores in developing a composite quality score for each participant hospital.

Under such an approach, we would first score individually each participant hospital on the Hospital-level 30-day, all-cause RSRR using the elective primary THA and/or TKA (NQF #1551) measure; Hospital-level RSCR following using the elective primary THA and/or TKA measure (NQF #1550); and HCAHPS Survey measure (NQF #0166) based on the participant hospital's performance percentile as compared to the national distribution of hospitals' measure performance, assigning scores according to the point values displayed in Table 12. These individual measure scores were set to reflect the measure weights included in Table 11 so they could ultimately be summed without adjustment in calculating the composite quality score.

Table 12—Individual Scoring Under the Composite Quality Score Alternative Considered for Three Required Quality Measures in the Proposed Rule

Performance percentileComplications measure quality score (points)HCAHPS survey quality score (points)Readmissions measure quality score (points)
≥90th8.006.004.00
≥80th and <90th7.405.553.70
≥70th and <80th6.805.103.40
≥60th and <70th6.204.653.10
≥50th and <60th5.604.202.80
≥40th and <50th5.003.752.50
≥30th and <40th4.403.302.20
<30th0.000.000.00

Given the current national distribution of hospital performance on these measures, in the proposed rule we stated our belief that small point increments related to higher measure performance deciles would be the most appropriate way to assign more points to reflect meaningfully higher quality performance on the measures. The absolute differences for each decile among the three measures reflect the intended weight of the measure in the composite quality score. We would assign any low volume participant hospital without a reportable value for the measure to the 50th performance percentile of the measure, so as not to disadvantage a participant hospital based on its low volume alone because that hospital may in actuality provide high quality care. These three measures are well-established measures in use under CMS hospital programs, so we did not believe that scores below the 30th percentile reflect quality performance such that they should be assigned any individual quality measure score points for LEJR episodes under CJR. However, we also considered reducing scores incrementally across the bottom three deciles in order to provide greater incentives for quality improvement for hospitals that may not believe they can attain the 30th performance percentile on one or more of the three measures and to avoid creating a “cliff” at the 30th performance percentile. We sought comment on this scoring approach to the three required quality measures.

Additionally, we would assign a measure quality score of one point for participant hospitals that successfully submit THA/TKA voluntary data and 0 points for participant hospitals that do not successfully submit these data. Because we would not use the actual THA/TKA voluntary data on the patient-reported outcome measure in assessing LEJR episode quality performance under the model, we believed this straightforward binary approach to scoring the submission of THA/TKA voluntary data for the patient-reported outcome measure development would be appropriate.

We note that the Shared Savings Program utilizes a similar scoring and weighting methodology, which is described in detail in the CY2011 Shared Savings Program Final Rule (see § 425.502). The HVBP and HACRP programs also utilize a similar scoring methodology, which applies weights to various measures and assigns an overall score to a hospital (79 FR 50049 and 50102).

We would sum the score on the three quality measures and the score on successful submission of THA/TKA voluntary data to calculate a composite quality score for each participant hospital. Then we would incorporate this score in the model payment methodology by first, requiring a minimum composite quality score for reconciliation payment eligibility if the participant hospital's actual episode spending is less than the target price and second, by making quality incentive payments that change the effective discount percentage included in the target price experienced by the hospital in the reconciliation process. The payment policies we would apply are displayed in Tables 13, 14, and 15 for the performance years of the model. Start Printed Page 73365Under the CJR model as proposed, there would be no participant hospital repayment responsibility in performance year 1 and this responsibility would begin to be phased-in in performance year 2, with full implementation in performance year 3.

Table 13—Performance Year 1: Relationship of Composite Quality Score To Reconciliation Payment Eligibility and the Effective Discount Percentage Experienced at Reconciliation Under the Composite Quality Score Alternative Considered in the Proposed Rule

Composite quality scoreEligible for reconciliation paymentEligible for quality incentive paymentEffective discount percentage for reconciliation payment (%)Effective discount percentage for repayment amount
≤5.00NoNo3.0Not applicable.
>5.00 and ≤9.25YesNo3.0Not applicable.
>9.25 and ≤15.20YesYes2.0Not applicable.
>15.20YesYes1.5Not applicable.

Table 14—Performance Year 2: Relationship of Composite Quality Score to Reconciliation Payment Eligibility and the Effective Discount Percentage Experienced at Reconciliation Under the Composite Quality Score Alternative Considered in the Proposed Rule

Composite quality scoreEligible for reconciliation paymentEligible for quality incentive paymentEffective discount percentage for reconciliation payment (%)Effective discount percentage for repayment amount (%)
≤5.00NoNo3.02.0
>5.00 and ≤ 9.25YesNo3.02.0
>9.25 and ≤ 15.20YesYes2.01.0
>15.20YesYes1.50.5

Table 15—Performance Years 3-5: Relationship of Composite Quality Score to Reconciliation Payment Eligibility and the Effective Discount Percentage Experienced at Reconciliation Under the Composite Quality Score Alternative Considered in the Proposed Rule

Composite quality scoreEligible for reconciliation paymentEligible for quality incentive paymentEffective discount percentage for reconciliation payment (%)Effective discount percentage for repayment amount (%)
≤5.00NoNo3.03.0
>5.00 and ≤ 9.25YesNo3.03.0
>9.25 and ≤ 15.20YesYes2.02.0
>15.20YesYes1.51.5

Under this approach, the CJR model discount included in the target price without consideration of the composite quality score would be 3.0 percent, not the 2.0 percent described under our payment proposal in section III.C.4.b.(9) of this final rule. In the proposed rule, we stated our belief that a discount percentage of 3.0 percent without explicit consideration of episode quality is reasonable as it is within the range of discount percentages included in the ACE demonstration and it is the Model 2 BPCI discount factor for 30 and 60 day episodes, where a number of BPCI participants are testing LEJR episodes subject to the 3.0 percent discount factor. Hospitals that provide high quality episode care would have the opportunity to receive quality incentive payments that would reduce the effective discount percentage as displayed in Tables 13, 14, and 15. Depending on the participant hospital's actual composite quality score, quality incentive payments could be valued at 1.0 percent to 1.5 percent of the hospital's benchmark episode price (that is, of the expected episode spending prior to application of the discount factor to calculate a target price).

Under this methodology, we would require hospitals to achieve a minimum composite quality score of greater than 5.00 to be eligible for a reconciliation payment if actual episode spending was less than the target price. Participant hospitals with below acceptable quality performance reflected in a composite quality score less than or equal to 5.00 would not be eligible for a reconciliation payment if actual episode spending was less than the target price. A level of quality performance that is below acceptable would not affect participant hospitals' repayment responsibility if actual episode spending exceeds the target price. We believed that excessive reductions in utilization that lead to low actual episode spending and that could result from the financial incentives of an episode payment model would be limited by a requirement that this minimum level of LEJR episode quality be achieved for reconciliation payments to be made. This policy would encourage hospitals to focus on appropriate reductions or changes in utilization to achieve high quality care in a more efficient manner. Therefore, these hospitals would be ineligible to Start Printed Page 73366receive a reconciliation payment if actual episode spending was less than the target price.

For hospitals with composite quality scores of less than or equal to 5.00, we also considered a potential alternative approach. Under this approach, we would still permit this group of hospitals to receive reconciliation payments but would impose a quality penalty that would increase their effective discount percentage to 4.0 percent for purposes of calculating the reconciliation payment or recoupment amount in performance years 3 through 5, 4.0 percent for calculating the reconciliation payment and 3.0 percent for calculating the repayment amount in performance year 2, and 4.0 percent for calculating the reconciliation payment in performance year 1 where participant hospitals have no repayment responsibility. A potential advantage of this approach is that it would provide stronger incentives for quality improvement for participant hospitals with low performance on quality, even if they did not expect to be able to reduce actual episode spending below the target price. In addition, this approach would provide financial incentives to improve the efficiency of care even for hospitals that did not expect to meet the minimum quality score for reconciliation payment eligibility, while still providing strong incentives to provide high-quality care. The disadvantage of this approach is that it could provide reconciliation payments even to hospitals that did not achieve acceptable quality performance.

Participant hospitals with an acceptable composite quality score of >5.00 and ≤9.25 would be eligible for a reconciliation payment if actual episode spending was less than the target price because their quality performance was at the acceptable level established for the CJR model. They would not be eligible for a quality incentive payment at reconciliation because their episode quality performance, while acceptable, was not good or excellent. Therefore, these hospitals would be eligible to receive a reconciliation payment if actual episode spending was less than the target price.

Participant hospitals with a good composite quality score of >9.25 and ≤15.20 would be eligible for a quality incentive payment at reconciliation if actual episode spending was less than the target price because their quality performance exceeded the acceptable level required for reconciliation payment eligibility under the CJR model. In addition, they would be eligible for a quality incentive payment at reconciliation for good quality performance that equals 1.0 percent of the participant hospital's benchmark price, thereby changing the effective discount percentage included in the target price experienced by the hospital at reconciliation. Thus, participant hospitals achieving this level of quality for LEJR episodes under CJR would either have less repayment responsibility (that is, the quality incentive payment would offset a portion of their repayment responsibility) or receive a higher payment (that is, the quality incentive payment would add to the reconciliation payment) at reconciliation than they would have otherwise based on a comparison of actual episode spending to the target price that reflects a 3.0 percent discount. Therefore, these hospitals would be eligible to receive a reconciliation payment if actual episode spending was less than the target price and would also receive a quality incentive payment.

Finally, hospitals with an excellent composite score quality score of >15.20 would be eligible to receive a reconciliation payment if actual episode spending was less than the target price because their quality performance exceeded the acceptable level required for reconciliation payment eligibility under the CJR model. In addition, they would be eligible for a higher quality incentive payment at reconciliation for excellent quality performance that equals 1.5 percent of the participant hospital's benchmark price, thereby changing the effective discount percentage included in the target price experienced by the hospital at reconciliation. Thus, participant hospitals achieving this level of quality for LEJR episodes under CJR would either have less repayment responsibility (that is, the quality incentive payment would offset a portion of their repayment responsibility) or receive a higher payment (that is, the quality incentive payment would add to the reconciliation payment) at reconciliation than they would have otherwise based on a comparison of actual episode spending to the target price that reflects a 3.0 percent discount. Therefore, these hospitals would be eligible to receive a reconciliation payment if actual episode spending was less than the target price and would also receive a quality incentive payment.

Under this methodology, the proposed stop-loss and stop-gain limits discussed in section III.C.8. of this final rule would not change. We believed this approach to quality incentive payments based on the composite quality score could have the effect of increasing the alignment of the financial and quality performance incentives under the CJR model to the potential benefit of participant hospitals and their collaborators as well as CMS, although it would substantially increase the complexity of the methodology to link quality and payment. We sought comment on this alternative approach to basing reconciliation payment eligibility and quality incentive payments on the participant hospital's composite quality score under the CJR model, as well as the composite quality scoring ranges applicable to the respective payment policies.

While we described in detail this alternative considered to link quality to payment under CJR, we did not propose this methodology for several reasons. First, the Shared Savings Program and HVBP program utilize many more measures than we proposed for the CJR model. For example, the Shared Savings Program initially incorporated thirty-three measures across four quality domains (79 FR 67916 and 67917). The range of measures in the Shared Savings Program and the HVBP program lends itself to a scoring approach, which can account for many measures and allows providers to achieve a high score despite performing well on some measures but achieving lower performance on others. There is a detailed description of the Shared Savings Program scoring methodology on the CMS Web site at: https://www.cms.gov/​Medicare/​Medicare-Fee-for-Service-Payment/​sharedsavingsprogram/​Quality_​Measures_​Standards.html. We believed that given the more limited set of measures chosen for the CJR model, a scoring approach such as the alternative described in this section could diminish the importance of each measure. Use of a scoring approach would not allow hospital performance on two different outcomes to be easily reviewed and understood with respect to the impact of individual measure performance on Medicare's actual payment for the episode under the model. Second, we believed the measures proposed for this model represent goals of clinical care that should be achievable by all hospitals participating in the model that heighten their focus on these measures, especially the readmissions and complications measures, for LEJR episodes based on the financial incentives in the model. Finally, we believed that a methodology that assesses performance based on absolute values of a specific set of measures that Start Printed Page 73367are already in use, as we proposed for the CJR model, would be the most appropriate methodology to provide achievable and predictable quality targets for participant hospitals on measures that monitor the most meaningful quality of care outcomes in a model where some acute care hospitals that might not choose to participate in a voluntary model are also included. Our proposed method as discussed in the next section reflected our expectation that hospitals achieve a certain level of performance on measures to ensure that hospitals provide high-quality care under the model.

Finally, we also considered an approach whereby participant hospitals would not be penalized with regard to their eligibility for reconciliation payments in CJR for failure to meet the specified thresholds for the quality measures in performance year 1 of the model; in other words, we would delay the proposal described in the next section to performance year 2 rather than beginning in performance year 1. We considered calculating participant hospital performance on the required measures for the model, and, if actual episode spending was less than the target price, the participant hospital would receive a full reconciliation payment of savings achieved beyond the target price, regardless of performance on the quality measures. However, we did not believe this would be appropriate for the CJR model, given that two of the measures are administrative claims-based and thus impose no additional reporting burden on hospitals; rather, these two measures are established measures in existing CMS quality programs, and a central goal of the model is improving care for Medicare beneficiaries in LEJR episodes. We noted that the HCAHPS Survey measure (NQF #0166) is also an established measure in the HIQR Program and would not impose additional reporting burden on hospitals.

We summarize the public comments we received on these alternatives considered to link quality and payment and provide our responses in section III.C.5.b.(5)(c)(iii) of this final rule. We note that we will be adopting the composite score methodology for the CJR model, as discussed in our responses to comments in section III.C.5.b.(5)(c)(iii) of this final rule.

(iii) Threshold Methodology and Final Policy To Link Quality and Payment

For the reasons outlined in the previous section, we did not propose to use similar methodologies to other CMS programs that would tie CJR episode reconciliation payment eligibility and reconciliation payment and Medicare repayment amounts to a composite quality score on specified quality measures, but as discussed later in this section, we instead proposed to simply assess performance or achievement on a quality measure by setting a measure result threshold for each measure beginning in performance year 1 of the model.

We proposed that the CJR measure result threshold would be based on the measure results from the HIQR Program, a nationally-established program, and would use its national distribution of measure results. These are the same measure results posted on Hospital Compare or in the Hospital Compare downloadable database (https://data.medicare.gov/​data/​hospital-compare) for the HIQR Program. We refer readers to the earlier discussion of the HIQR Program, which utilizes measures to assess most acute care hospitals in the nation. Determining the CJR model target thresholds are discussed in the next section.

As previously described, we proposed for the CJR model the following three required measures to assess LEJR episode quality of care:

  • Hospital-level 30-day, all-cause RSRR following elective primary THA and/or TKA (NQF #1551).
  • Hospital-level RSCR following elective primary THA and/or TKA (NQF #1550).
  • HCAHPS Survey (NQF #0166).

We also proposed to make a voluntary reporting payment adjustment for CJR participant hospitals who successfully and voluntarily submit data for the THA/TKA patient-reported outcome-based performance measure (henceforth referred to as “THA/TKA voluntary data”) as described in sections III.C.5.b.(2) and III.D.3.a. of this final rule. We proposed that participant CJR hospitals must meet or surpass a specified threshold for each required measure beginning for performance year 1 of the model in order to be eligible for a reconciliation payment if actual episode payments are less than the target price. The calculation of the HCAHPS Survey measure is described in section III.D.2.c. of this final rule. We proposed to use the individual measure results calculated as specified in section III.D. of this final rule for the three required measures to determine hospital eligibility for reconciliation payment for each performance year of the CJR model. Also, as discussed in section III.C.4. of this final rule, which outlines the proposed pricing structure for the CJR model, target prices for MS-DRG 470-anchored episodes and for MS-DRG 469-anchored episodes would be calculated for hospitals participating in the model for an episode of care extending 90 days after discharge from the anchor hospitalization. Participant hospitals that achieve actual episode payment below the specified target price for a given performance period would be eligible for a reconciliation payment, provided that the participant hospital also met episode quality thresholds on the three required measures for the performance period.

We proposed to use the following quality criterion to determine if a participant hospital qualifies for a reconciliation payment based on the episode quality thresholds on the three required measures:

The hospital's measure result is at or above the 30th percentile of the national hospital measure results calculated for all HIQR-Program participant hospitals for each of the three required measures for each performance period (for a detailed description of how we determined the performance period and reconciliation payment eligibility, see section III.C.5. of this final rule).

Using HIQR Program's 3 year rolling period as outlined in section III.D.2.d. (Applicable Time Period) of this final rule, if a participant hospital performed at or above the 30th percentile of all HIQR Program hospitals for each of the three required measures and if actual episode payment was less than the target price for the specified performance year, we would make a reconciliation payment to the hospital. Failure to achieve the threshold on one or more measures would result in the participant hospital not receiving a reconciliation payment regardless of whether the actual episode payment was less than the target price for that performance period. We proposed that for hospitals with insufficient volume to determine performance on an individual measure, these hospitals would be considered to be performing at the threshold level and their results would be publicly posted with all other participant hospitals' measure results (for a detailed summary of public reporting, see section III.D.5. of this final rule). We did not believe it would be appropriate to potentially penalize high quality, efficient hospitals due to their low volume, given that meeting the required quality measure thresholds would be required for reconciliation payment eligibility.

We also proposed for performance years 4 and 5 to increase the measure result threshold to the 40th percentile. We believed that increasing the measure result threshold to the 40th percentile Start Printed Page 73368would encourage participants to strive for continued quality improvement throughout the 5 performance years of the model. We sought comment on our proposal to make a reconciliation payment to a participant hospital that achieves actual episode spending below the target price for a performance year and performs at or above the 30th percentile of HIQR program participant hospitals for all three required quality measures in performance years 1 through 3 or the 40th percentile in performance years 4 and 5, as well as our proposal to consider low volume hospitals to be performing at the threshold level.

We proposed to require hospitals to meet the threshold for all three measures for the following reasons. The measures proposed for this model are fully developed, NQF-endorsed, and implemented measures in CMS IPPS programs. These measures are also publicly reported on the Hospital Compare Web site. Hospitals are familiar with the complications and readmissions quality measures and with the HCAHPS Survey, as they are currently included in the HIQR Program, HVBP program, and HRRP (79 FR 50031, 50062, 50208, 50209 and 50259), and we believed that there would be minimal additional administrative burden for hospitals. All three measures are widely utilized nationally; thus, a nationally-based threshold would be an appropriate benchmark. In addition, the goal of the CJR model is LEJR episode care redesign that includes effective care coordination and management of care transitions. Strategies to prevent and efficiently manage post-procedure complications and hospital readmissions following an LEJR procedure are consistent with the goals of the model; a hospital cannot succeed in this model without engaging in care redesign efforts that would address aspects of care included in these measures. Failure to perform successfully on these key quality measures (defined by meeting the minimum thresholds) would indicate that hospitals are not achieving quality consistent with the goals of the model to specifically incentivize greater improvement on these measures than hospitals not participating in the CJR model, and should not be eligible to receive a reconciliation payment from Medicare even if reduced episode spending is achieved. Finally, the approach we proposed is consistent with CMS' goal of moving hospitals and other providers to value-based payment that ties payment to quality. In the 5 performance years of this model, performance on quality measures would only be applied to determining eligibility for a reconciliation payment; quality measures would not be used to determine participant hospitals' financial responsibility, except for the proposed voluntary reporting payment adjustment described in described in section III.C.5.b.(3) of this final rule. In essence, participant hospitals' responsibility to repay Medicare the difference between their target price and their actual episode payment, should actual episode payments exceed the target price, would not be impacted by performance on quality measures.

Finally, we proposed to increase the measure result thresholds for the final 2 performance years of the model, to ensure that CJR participant hospitals continue to maintain a high level of quality performance or improve performance on these measures as they gain experience with implementation of this payment model. More specifically, we proposed that in order for a participant hospital to receive a reconciliation payment for actual episode spending that is less than the target price for performance years 4 and 5, the participant hospital's measure result must be at or above the 40th percentile of the national hospital measure results calculated for all HIQR-Program participant hospitals for each of the three required measures for each performance period. As previously noted, we proposed to use the most recently available HCAHPS 4-quarter roll-up to calculate the HLMR. In the proposed rule, we stated our belief that holding the participant hospitals to a set measure result threshold for the first 3 years, and increasing this threshold for performance years 4 and 5, would emphasize the need to maintain and improve quality of care while cost efficiencies are pursued. We sought comment on our proposed approach to incorporating quality performance into eligibility for reconciliation payments under the CJR model for participant hospitals.

Table 16 displays the proposed thresholds that participant hospitals must meet on the various measures over the 5 model performance years.

Table 16—Proposed Thresholds for Required Quality Measures To Determine Participant Hospital Reconciliation Payment Eligibility Over 5 Years

MeasurePY1 ThresholdPY2 ThresholdPY3 ThresholdPY4 ThresholdPY5 Threshold
Hospital-level 30-day, all-cause RSRR following elective primary THA and/or TKA (NQF #1551)30th percentile30th percentile30th percentile40th percentile40th percentile.
Hospital-level RSCR following elective primary THA and/or TKA (NQF #1550)30th percentile30th percentile30th percentile40th percentile40th percentile.
HCAHPS Survey (NQF #0166)30th percentile30th percentile30th percentile40th percentile40th percentile.

We sought comment on our proposed methodology to utilize quality measure performance in the payment methodology for CJR, as well as the proposed thresholds for participant hospital reconciliation payment eligibility over the performance years of the model.

As discussed in section III.C.5.b.(2) of this final rule, we stated our belief that hospitals that choose to submit THA/TKA voluntary data should have the potential to benefit financially through an adjustment to the payment methodology of the model. We proposed a voluntary reporting payment adjustment for hospitals that successfully submit the THA/TKA voluntary data by reducing the discount percentage incorporated into the target price from 2.0 percent to 1.7 percent. This voluntary reporting payment adjustment would start in performance year 1 and would be available through performance year 5 of the model for each year that the hospital successfully reports THA/TKA voluntary data. As proposed, reporting THA/TKA voluntary data would not affect eligibility for a reconciliation payment if actual episode payments are less than the target price. Participant hospitals would still need to meet the 30th or 40th percentile threshold, as applicable to the given performance year, on all Start Printed Page 73369three required quality measures (Table 16).

We considered, but did not propose, two other alternatives to adjust the payment methodology for participant hospitals that successfully report the THA/TKA voluntary data as described in section III.C.5.b.(2) of this final rule. These alternatives would change the threshold percentile for the three required quality measures or, alternatively, reduce the number of required measures in which the threshold must be met provided that successful THA/TKA voluntary data were reported for a performance year. First, we considered reducing the threshold for reconciliation payment eligibility that participant hospitals must meet on the three required quality measures from the 30th percentile threshold to the 20th percentile threshold for performance years 1, 2, and 3, and from the 40th percentile to the 30th percentile for performance year. Second, we considered only requiring hospitals to meet the 30th percentile threshold on two of three outcome measures for performance years 1, 2, and 3, and the 40th percentile threshold on two of three outcome measures in performance years 4 and 5. Under both of these alternatives, the eligibility for reconciliation payments could change based on the THA/TKA voluntary data. We sought comment on these alternative payment methodology adjustments that could impact reconciliation payment eligibility, unlike the proposed voluntary reporting payment adjustment. We note that the other alternative approaches to encouraging THA/TKA voluntary data reporting for CJR beneficiaries as discussed in section III.C.5.b.(2) of this final rule that would not require adjustments to the CJR payment methodology would also not affect reconciliation payment eligibility.

The following is a summary of the comments received and our responses on the proposals and alternatives discussed in section III.C.5. of the proposed rule, including the proposed threshold methodology for reconciliation payment eligibility, as well as the alternatives considered that would change the proposed threshold requirements for participant hospitals that successfully report voluntary THA/TKA data. As cross-referenced several times earlier in this section, these comments and our responses also discuss a number of other proposals, alternatives considered, and other topics related to linking quality and payment under the CJR model for which we sought public comment.

Comment: Some commenters questioned the rationale for linking quality to episode payment for participant hospitals under the CJR model, arguing that the model should not be focused on individual hospital performance but on the overall performance of hospitals within the model, with respect to both the cost and quality of LEJR episode performance. The commenters observed that BPCI, a bundled payment model that includes LEJR as the most commonly selected episode and shares many features with the proposed CJR model, does not tie payment to quality, although BPCI has quality reporting requirements. They claimed that CMS, hospitals, and other providers lack experience with pay-for-performance in a bundled payment context and, therefore, that the level of performance that should be expected from providers under bundled payment is not yet understood. A commenter urged CMS to focus on the big picture in the CJR model, specifically changes in critical aspects of performance versus the national average for all hospitals along the continuum, potential changes in the types or nature of services to beneficiaries undergoing LEJR procedures, and aggregate changes in patient outcomes. Commenters asserted that tying a hospital's payment to performance on quality measures was not the only or the best way to make maintaining or improving LEJR episode quality performance central to the CJR model. Several commenters stated that implementing pay-for-performance in an episode payment model was premature, and recommended that CMS, at most, adopt a pay-for-reporting methodology while quality data are being collected and analyzed to determine the appropriate level of quality performance that should be specifically rewarded.

Several commenters urged CMS to delay implementing the proposed quality performance thresholds for reconciliation payment eligibility until performance year 2, or later, where the performance period for measure data would correspond more fully or completely to performance years under the model. They recommended that the first year or two of the CJR model should be pay-for-reporting and, because the proposed THA/TKA Complications measure (NQF #1550) and the THA/TKA Readmissions measure (NQF #1551) are claims-based measures and the HCAHPS Survey measure (NQF #0166) is currently administered by hospitals, all participant hospitals would be expected to meet the CJR model quality performance requirements, which would only require public reporting in performance year 1 and possibly performance year 2. Several commenters in favor of pay-for-reporting in the first performance year asserted that such an approach would be consistent with other CMS value-based initiatives. A commenter also claimed that a year of pay-for-reporting would allow participant hospitals the time to establish internal systems for analyzing quarterly claims data and provide them with maximal opportunity to achieve savings that could be invested in these systems.

Response: We note that we currently have broad experience with pay-for-performance in Medicare programs, including the HRRP, HVBP Program, HAC Reduction Program, and the Shared Savings Program. These pay-for-performance programs have improved the quality of care for Medicare beneficiaries. For example, since the implementation of HRRP in 2012, readmission and complications rates for various medical conditions such as elective THA/TKA have been significantly reduced, thereby resulting in improvements in the quality of care for Medicare beneficiaries undergoing LEJR procedures. Furthermore, pay-for-performance is a feature of a number of Innovation Center models currently in testing. We refer readers to section III.D.5. of this final rule for further discussion of public reporting of pay-for-performance data during performance year 1 of the model.

While the current BPCI models do not specifically link payment to quality, the Request for Applications describing the BPCI model design features was released over 4 years ago, in August 2011. We now have two years of experience with BPCI Model 2 Awardees, the model that most closely resembles the CJR model, in the risk-bearing period, and the year 1 BPCI annual evaluation and monitoring report from February 2015 is publicly available on the CMS Web site at: https://innovation.cms.gov/​Files/​reports/​BPCI-EvalRpt1.pdf. We have developed and adopted a variety of new quality measures in programs and models since 2011, as well as gained experience with pay-for-reporting and pay-for-performance in a variety of models and programs involving a wide range of health care providers and clinical conditions. Given our extensive experience over the past several years with pay-for-performance approaches, the availability of existing measures that reflect the quality of care for elective THA/TKA episodes, and the breadth of the CJR model, which reaches Start Printed Page 73370substantially all IPPS hospitals in the selected MSAs, including those hospitals who otherwise would not participate in a voluntary payment model, we believe that a pay-for performance approach is necessary and appropriate beginning in the model's first performance year. IPPS hospitals have substantial experience over multiple years with CMS programs that include pay-for-performance and we believe, given the proposed quality measures for the CJR model, that CJR pay-for-performance in an episode payment model is a natural extension to bundled payment of pay-for-performance measures used in current CMS programs. While we acknowledge that pay-for-performance is not the only way for a model to heighten a focus on maintaining or improving the quality of LEJR episode care, we believe that the CJR model, like other Innovation Center models, should target both improved quality and reduced costs. Based on our experience in other programs and models, we believe that pay-for-performance under the CJR model shows great promise in moving participant hospitals toward greater efficiency and higher quality of LEJR episodes. In view of successful implementation of pay-for-performance in other CMS hospital programs using similar quality measures that has resulted in significant improvements in the quality of care, we believe IPPS hospitals have sufficient experience to be ready for pay-for-performance under the CJR model. We expect that other features of the model design, including our plans for data sharing, will help participant hospitals committed to care redesign toward these goals achieve success on both quality and cost performance for episodes.

We note that the quality measures finalized for the model as discussed in section III.D.2. of this final rule rely upon data that hospitals are already submitting and which are already analyzed by CMS for other programs, so we see no reason to adopt a period of pay-for-reporting for the first performance year of the model or longer. In the proposed rule, we considered a similar policy that would not penalize hospitals with regard to their eligibility for reconciliation payments for failure to meet the proposed quality measure thresholds in performance year 1. However, we continue to believe that adopting pay-for-reporting and not pay-for-performance in performance year 1 or longer would be inappropriate given that two of the proposed quality measures are administrative claims-based measures and impose no additional reporting burden on hospitals, the proposed measures are all established measures in existing CMS quality programs, and a central goal of the CJR model is improving care for Medicare beneficiaries in LEJR episodes. In this regard, the CJR model is different from some other CMS value-based initiatives where the data for some measures were newly submitted by providers or newly analyzed by CMS early in the initiative. Furthermore, we do not believe that participant hospitals need a year of pay-for-reporting to develop systems for analyzing episode claims under the model, as we expect hospitals to already be focused on improving their performance on these measures. The two measures finalized for the CJR model are aligned with the goals of the CJR model, are familiar to hospitals based on their use in other CMS hospital programs, and are aligned with CMS priorities to reduce LEJR complications while improving the patient experience. Because the measures reflect these goals and accurately measure hospitals' level of achievement and improvement on quality outcomes that are important to beneficiaries undergoing LEJR procedures, we are finalizing our proposal to implement a pay-for-performance approach in the CJR model in the first performance year by using quality performance in the episode payment methodology.

Comment: Some commenters supported the proposed strategy to link quality to payment through performance thresholds for quality measures that would result in reconciliation payment eligibility if the thresholds were met. Several commenters further reasoned that there should be no need to increase thresholds for reconciliation payment eligibility over the performance years of the model as CMS had proposed because the possibility of reconciliation payment provides an adequate quality improvement incentive. A commenter in favor of the proposed threshold approach recommended that CMS make the proposed THA/TKA voluntary patient-reported outcome (PRO) data submission mandatory and significantly increase incentives around their collection.

A number of commenters estimated that under CMS' proposal, more than half of the participant hospitals would be ineligible for reconciliation payments based on their current quality measure performance, even if episode savings were achieved during a performance year. The commenters stated that CMS should not use performance percentiles that would always exclude a predetermined number of participant hospitals from reconciliation payments, and hold hospitals to multiple quality performance standards for the same measure performance under different CMS models and programs. They contended that performance percentiles, as measures of relative performance, do not reflect best practices and, therefore, recommended that CMS require a level of absolute measure performance rather than relative performance when incorporating quality performance into the payment methodology under the CJR model. The commenters did not describe the absolute levels of performance that they would recommend on the quality measures for the CJR model. Several commenters claimed that the use of thresholds for reconciliation eligibility disadvantages small hospitals because only one or two patient instances could change the participant hospital's performance percentile and, therefore, affect the hospital's eligibility for reconciliation payments. Other commenters pointed out that the Shared Savings Program uses quality thresholds, but the methodology accounts for improvement and the program is voluntary, while hospital participation would be required in the CJR model and improvement was not considered in the pay-for-performance methodology CMS proposed.

Other commenters asserted that CMS' proposal linking quality measure performance to eligibility for reconciliation payments failed to reflect the quality of care delivered in the context of the model due to flaws in the proposed approach to determining participant hospital performance in relation to the thresholds. The commenters contended that the proposed methodology to determine performance on quality measures and link performance to reconciliation payment eligibility uses arbitrary distinctions in performance among hospitals that are not borne out by the data or even by CMS's own method of assigning ratings of performance on the Hospital Compare Web site. They stated that use of measure result point estimates to determine performance percentiles under CMS' proposal for performance thresholds for reconciliation payment eligibility may not be appropriate because: (1) The THA/TKA Complications measure (NQF #1550) and the THA/TKA Readmissions measure (NQF #1551) are a ratio comparing observed to expected outcome, where expected is based on the national performance, so an individual hospital's performance Start Printed Page 73371should be assessed within confidence intervals as the measure was originally specified, tested, and endorsed by the NQF; and (2) there may not be a clinically and statistically significant difference in the performance of hospitals immediately above and below the 30th percentile. The commenters observed that while the HRRP uses measure result point estimates (the same measure results proposed in section III.C.5.b.(5)(b) of the proposed rule, which proposed to use the absolute values of the CJR model participant hospital measure results) in calculating the excess readmission ratio in accordance with the statutory provision that defines this ratio, they stated that CMS has the flexibility under the statutory authority for the CJR model to use confidence intervals in determining outcome measure results for use in the payment methodology.

A number of commenters recommended that CMS adopt a threshold methodology that would utilize the confidence intervals used on the Hospital Compare Web site that distinguishes performance based on the three categories of comparison to the national rate on the THA/TKA Complications measure (NQF #1550) and the THA/TKA Readmissions measure (NQF #1551) to determine if a participant hospital is eligible for reconciliation payment. On Hospital Compare, hospitals are grouped into “no different than national rate,” “better than national rate,” or “worse than national rate” for each measure. The commenters recommending this methodology recommended against use of the HCAHPS Survey measure (NQF #0166).

Therefore, the commenters maintained that CMS should modify its proposal and set the quality performance thresholds for reconciliation payment eligibility at “worse than national rate,” rather than at the 30th percentile or above compared to the national rate. Specifically, the commenters suggested if performance on both the THA/TKA Complications measure (NQF #1550) and the THA/TKA Readmissions measure (NQF #1551) is statistically “worse than national rate,” then a participant hospital should not be eligible for reconciliation payment. Those hospitals that are deemed “no different than national rate” or “better than national rate” on both measures should automatically be deemed eligible for any potential reconciliation payment. Some commenters further urged CMS to also allow participant hospitals performing “worse than national rate” on one or both quality measures to receive reconciliation payments if CJR model episode savings were achieved as long as the hospital submits a corrective action plan to CMS describing their future strategies to improve quality of care, including contributing a portion of the reconciliation payment to quality performance improvement strategies. These commenters asserted that the quality performance thresholds should provide equal financial opportunity and incentives to all hospital participants.

The commenters claimed that setting quality performance thresholds at the level of “worse than national rate” as displayed on the Hospital Compare Web site would reduce confusion among the public with an interest in hospital performance under the CJR model, and strike an appropriate balance between encouraging hospital to focus on quality performance and providing hospitals with a fair opportunity to receive reconciliation payments if episode savings are achieved. A commenter reported that nationally there are 22 hospitals with performance on the THA/TKA Complications measure (NQF #1550) or the THA/TKA Readmissions measure (NQF #1551) that is “worse than national rate,” and only one hospital that is “worse than national rate” on both measures.

Response: We appreciate the support of some commenters for our proposal to set performance thresholds for reconciliation payment eligibility at the 30th percentile based on the national distribution of measure results, as well as the concerns expressed by some commenters about using relative performance to assess participant hospital episode quality performance in the CJR model. We continue to believe that relative measure performance is the most appropriate way to incorporate quality performance into the CJR model because we do not have sufficient information about hospital performance to set and use an absolute performance result on each measure. We believe that hospitals nationally are working to improve their performance on the quality measures proposed for the CJR model on an ongoing basis and, thus, while we expect that CJR participant hospitals will have a heightened focus on improvement on these measures as a result of the financial incentives resulting from episode payment, we are not yet certain in this model test what performance outcomes can be achieved under best practices. Therefore, we will not set absolute performance results as quality thresholds for reconciliation payment eligibility under the CJR model. We continue to believe that relative measures of quality performance are most appropriate for the CJR model as hospitals continue to make progress nationally on improving patient outcomes.

Furthermore, we will not make THA/TKA voluntary PRO and limited risk variable data submission mandatory and increase the incentives around their collection in the CJR pay-for-performance methodology as recommended by a commenter. This measure remains under development, and we want to encourage robust hospital reporting to speed measure development, but the measure is not yet ready to have its results incorporated in the CJR model methodology in the manner recommended by the commenter. We refer readers to section III.D.3.a. of this final rule for further discussion of our future plans to incorporate PRO measure results in the CJR pay-for-performance methodology.

We appreciate the suggestions of many commenters that we utilize outcome measure thresholds of “worse than national rate” as displayed on the Hospital Compare Web site to set the thresholds for reconciliation payment eligibility. For purposes of the Hospital Compare Web site, we made a specific choice around categorizing hospitals to performance categories for public display of hospital measure results in order to display a high level of statistical certainty about differences in hospital quality performance that would be reviewed by beneficiaries and other members of the public. Specifically for the Hospital Compare Web site, to assign hospitals to performance categories, the hospital's interval measure estimate is compared to the national rate. If the 95 percent interval estimate includes the national observed rate for that measure, the hospital's performance is in the “no different than national rate” category. If the entire 95 percent interval estimate is below the national observed rate for that measure, then the hospital is performing “better than national rate.” Finally, if the entire 95 percent interval estimate for the hospital is above the national observed rate for that measure, the hospital's performance is “worse than national rate.”

Regarding the commenter who suggested that an individual hospital's performance on a measure should be assessed within confidence intervals as the measure was originally specified, tested, and endorsed by the NQF, we note that the THA/TKA Complications measure (NQF #1550) was not endorsed by the National Quality Forum for its use with an interval estimate, since NQF endorses measure specifications and not the use of measures in various programs Start Printed Page 73372or models. We acknowledge that CMS uses outcome measure ratios and rates in different ways that may lead to some confusion for stakeholders. We also want to clarify that during measure development of the THA/TKA Complications measure (NQF #1550) and the THA/TKA Readmissions measure (NQF #1551), these measures were developed and tested to yield risk-standardized ratios, which are multiplied by the national rate and reported as risk-standardized rates in Hospital Compare, and that the 95 percent interval estimate is specifically used to display the measure for public reporting on the Hospital Compare Web site. We chose to use rates on the Hospital Compare Web site because we believe that presentation of a rate on the Hospital Compare Web site is better understood by consumers than a measure result expressed as a predicted-to-expected ratio. For purposes of the CJR model, we will also use risk-standardized rates for the THA/TKA Complications measure (NQF #1550) as discussed in section III.D.2.a. of this final rule. We discuss our final decision not to adopt the THA/TKA Readmissions measure (NQF #1551) for this model in section III.D.2.b. of this final rule.

We note that “worse than national rate” is the quality performance threshold for reconciliation payment eligibility recommended by many commenters as the statistically certain measure of poor hospital quality performance, yet almost every hospital in the country already exceeds this level on the THA/TKA Complications measure (NQF #1550) and the THA/TKA Readmissions measure (NQF #1551). Nationally, we estimate that only 29 hospitals currently perform at “worse than national rate” on one or more of these measures, a number that is similar to the estimate provided by a commenter. Thus, based on current measure performance only a very small number of hospitals would fail to meet the quality performance thresholds for reconciliation payment eligibility recommended by many commenters. We do not believe that adopting “worse than national rate” as the threshold for reconciliation payment eligibility, or applying no threshold as recommended by some commenters if a hospital “worse than national rate” submits a corrective action plan to CMS, would further encourage quality improvement or maintenance of high performance for participant hospitals in the CJR model, beyond the incentives that already exist in CMS programs.

Either incorporating a “worse than national rate” threshold or applying no threshold would essentially eliminate pay-for-performance under the CJR model, which would not be consistent with our final decision discussed in the prior response to public comments to incorporate a pay-for-performance methodology in the CJR model beginning in performance year 1.

Regarding the recommendations to use interval estimates to identify hospitals with performance “worse than national average” as the most equitable approach to identifying statistically valid poor hospital performance on quality measures, we have previously explained our position on the use of interval estimates when determining payment outcomes for hospital performance on measures. Specifically for the HRRP where we use point estimates for quality measure performance, we acknowledged outcome measures of risk-standardized condition-specific readmission rates to be statistical measures (77 FR 53394). We also recognized that statistical measures will include some degree of variation and stated that other Medicare programs use similar statistical measures as part of their programs, so any consideration of the use of interval estimates with respect to the HRRP may have implications for other programs (77 FR 53394). Despite this reality, we finalized the HRRP methodology for quality measure performance (76 FR 51673), which results in the use of a point estimate for a hospital's excess readmission ratio (77 FR 53394), and we use point estimates in other CMS programs that rely upon statistically-based outcome measures, such as the HVBP Program. (76 FR 26504). We note that over the past several years the HRRP has shown that use of point estimates in the program has still led to improvement in hospital readmission rates.[42 43] We, therefore, continue to believe that quality performance can be assessed by measure result point estimates that do not rely on the statistical certainty of interval estimates which may fail to identify real, clinically meaningful differences in hospital measure performance.

However, we agree with the commenters that our proposal to set performance thresholds for reconciliation payment eligibility at the 30th percentile does not reflect the statistical certainty of intervals around hospital measures performance results and may not adequately account for the variation that occurs in risk-standardized rates like the THA/TKA Complications measure (NQF #1550) and the THA/TKA Readmissions measure (NQF #1551) proposed for use in the CJR model. We also agree with the commenters that setting required measure performance thresholds for reconciliation eligibility may provide insufficient quality and cost episode improvement incentives for some participant hospitals in the CJR model. We estimate that based on their current quality measure performance one-third of participant hospitals would not be eligible for reconciliation payments under our proposed thresholds for the three required quality measures, even if those hospitals achieved savings beyond the target price. While our estimate is lower than the estimate of more than 50 percent of participant hospitals that was provided by some commenters, we agree with the commenters that the proposed methodology would not provide significant quality and cost episode improvement incentives for a substantial percentage of participant hospitals in the CJR model.

We continue to believe there are real, clinically meaningful differences that are important to Medicare beneficiaries in hospital performance on the THA/TKA Complications measure (NQF #1550) finalized for this model, as well as opportunities for improvement, which are not recognized by the statistical certainty approach that we use for the Hospital Compare Web site but can be appropriately recognized by assigning hospitals to measure performance percentiles, such as we proposed for the CJR model. We also believe it is appropriate to make different choices for estimating measure performance for model or program payment policies, depending on the context. For example, in the CJR model where we proposed to use quality performance in the payment methodology of a model specifically focused on quality outcomes directly addressed by the proposed measures, we believe a different approach to estimating performance differences than the statistical certainty approach used on the Hospital Compare Web site would allow us to observe and reward real quality performance incentivized by episode payment that otherwise would Start Printed Page 73373be unrecognized. Therefore, we continue to believe that assigning hospital measure results to a performance percentile in comparison with the national distribution is an appropriate strategy to categorize and recognize hospitals achieving different levels of quality performance on the measures. We note that assigning hospitals to performance percentiles based on their measure result point estimates, and then using deciles of performance in a pay-for-performance model payment methodology that does not use hospital performance percentiles as thresholds, would help account for some of the statistical variation that could occur in measure result point estimates and reduce the likelihood that we would consider variation to be a real change in measure performance. Therefore, we are finalizing our proposal discussed in section III.C.5.b.(5)(b) of this final rule to assign each participant hospital's measure point estimate to a performance percentile based on the national distribution of measure results. However, because the statistical uncertainty in measure results increases the challenge of determining the most equitable performance threshold, below which the level of performance is no longer in the best interest of the beneficiary, as well as our interest in providing quality and cost episode improvement incentives for all participant hospitals under the CJR model, we are not finalizing our proposal to set performance percentile thresholds for reconciliation payment eligibility in the CJR model. Because we are not using performance percentile thresholds for reconciliation payment eligibility in the CJR model's final pay-for-performance methodology, we will neither be setting nor changing such thresholds in the context of the model's payment methodology over the model's performance years. We will be adopting the composite score methodology, as discussed in the following response to comments.