Centers for Medicare & Medicaid Services (CMS), HHS.
In accordance with court rulings in cases that challenge the federal fiscal year (FY) 2004, 2005, and 2006 outlier fixed-loss threshold (FLT) rulemakings, this document provides further explanation of certain methodological choices made in the FLT determinations for those years.
June 6, 2019.
Start Further Info
FOR FURTHER INFORMATION CONTACT:
Don Thompson, (410) 786-6504.
End Further Info
Start Supplemental Information
On May 19, 2015, in District Hospital Partners v. Burwell, 786 F.3d 46 (D.C. Cir. 2015), the Court of Appeals for the District of Columbia Circuit held that the FY 2004 fixed-loss threshold (FLT) was inadequately explained in the federal fiscal year (FY) 2004 hospital inpatient prospective payment systems (IPPS) final rule. The court of appeals ordered the district court to remand to CMS for further explanation of the handling of data pertaining to 123 hospitals the agency had identified as likely to have engaged in “turbocharging,” that is, manipulating their charges to obtain greater outlier payments. The United States District Court for the District of Columbia then remanded to the Secretary in accordance with the decision of the Court of Appeals. Order, Dist. Hosp. Partners, L.P. v. Burwell, Civil Action No. 11-0116 (ESH) (D.D.C. August 13, 2015).
On September 2, 2015, the District Court issued an order in a separate case, Banner Health v. Burwell, No. 10-1638 (ECF Nos. 149 and 150), 126 F. Supp. 3d 28 (D.D.C. 2015), remanding for additional explanation of the FLT from the FY 2004 final rule consistent with the D.C. Circuit's decision in District Hospital Partners. The court stated that the agency should explain further why it did not exclude data from the 123 hospitals from the outlier charge inflation calculation used to produce estimates of future Medicare payments for FY 2004.
In the January 22, 2016 Federal Register (81 FR 3727), we published an additional explanation in response to these court orders. In the October 14, 2016 Federal Register (81 FR 70980), we published a minor, non-substantive correction to the January 2016 document.
In Banner Health v. Price, 867 F.3d 1323 (D.C. Cir. 2017), the court of appeals reviewed the January 2016 document and found that the agency still had not adequately explained why the agency, in the FY 2004 rulemaking, did not exclude the charge data from the 123 hospitals it had identified as likely turbochargers when calculating the charge inflation factor used to transform historical charges into future charges for purposes of the agency's projections. The court of appeals also found that the agency had not adequately explained why it did not apply a downward adjustment to hospitals' cost-to-charge ratios when determining the FLTs for FYs 2004, 2005, and 2006, an issue not addressed in the Court of Appeals decision in District Hospital Partners. The court in Banner Health ordered the district court to remand to CMS to provide additional explanation on these two points. The district court issued a remand order on April 12, 2018. The district court also entered a similar order with respect to the FY 2004 determination in another case, District Hospital Partners, L.P. v. Azar, 320 F. Supp. 3d 42 (D.D.C. 2018).
We are issuing this document to provide the additional explanation required by these decisions.
II. Provisions of the Explanation
A. Inclusion of Data Pertaining to 123 Hospitals Identified as Likely Turbochargers in the Calculation of Estimated Charge Inflation for FY 2004
The first issue pertains to the use of data pertaining to 123 hospitals whom we described in a March 5, 2003 proposed rule (68 FR 10420), as hospitals likely to have engaged in turbocharging. We chose to calculate the FY 2004 charge inflation adjustment using data that incorporated data Start Printed Page 26361pertaining to the 123 hospitals, instead of choosing to omit data pertaining to those hospitals.
As we discussed in our earlier publications, the 123 hospitals were identified through an analysis of Medicare Provider Analysis and Review (MedPAR) file data from FY 1999 to FY 2001. We singled out hospitals whose percentage of outlier payments relative to total diagnosis-related group (DRG) payments increased by at least 5 percentage points over that period, and whose case-mix (the average DRG relative weight value for all of a hospital's Medicare cases) adjusted charges increased at a rate at or above the 95th percentile rate of charge increase for all hospitals over the same period. We note that we conducted this analysis primarily for the purpose of assessing and diagnosing the problem of turbocharging, not for the purpose of making adjustments to our projections for the FY 2004 rulemaking.
We identified the 123 hospitals based on data from the interval from FY 1999 to FY 2001. Our charge inflation calculation for FY 2004 was based on data covering a more recent interval, from FY 2000 to FY 2002. We were attempting to project charge increases over a third period, from FY 2002 to FY 2004.
The hypothesis underlying the suggestion that the 123 hospitals should have been omitted is that charge inflation for those 123 hospitals was likely to begin slowing in FY 2004 in response to the adoption of the June 9, 2003 Outlier final rule (68 FR 34494), while charge inflation for other hospitals would remain in line with historical patterns between FY 2002 and 2004. Consequently, according to this hypothesis, an estimate computed from FY 2000 to 2002 charge data that included the 123 hospitals would likely overstate FY 2004 hospital charges for the entire population of hospitals. But this hypothesis depends on assumptions that, at the time of the FY 2004 rulemaking, we did not find appreciably more credible than the alternative assumptions we ultimately relied upon.
The hypothesis that the 123 hospitals identified in our analysis should have been dropped from the charge inflation computation treats the removal of the 123 hospitals as synonymous with accounting for turbocharging. It presumes that removing the 123 hospitals from the measure of charge inflation would have accounted for the end of turbocharging, without otherwise introducing error or bias, and that, conversely, including the 123 hospitals introduced systematic error. But that assumes both that all of the 123 hospitals were in fact engaged in turbocharging, and that the population of turbocharging hospitals remained for the most part unchanged over all three intervals from FY 1999 through the end of FY 2003—that is, that those 123 hospitals continued to engage in turbocharging after FY 2001, that they did not materially increase their rate of turbocharging during that period, and that no other hospitals started to turbocharge or otherwise increase their rate of charge inflation. We did not feel sufficiently confident that such an assumption would enhance the accuracy of the outlier threshold calculation to incorporate it into our projections for FY 2004.
While our analysis confirmed that turbocharging was a problem, and that rule changes were warranted along the lines of the changes we adopted in June 2003, we did not otherwise have a confident grasp on which hospitals were turbocharging at what times. Our analysis suggested that the 123 hospitals that we identified were likely engaging in turbocharging during the FY 1999 to FY 2001 interval, but it did not tell us whether the population of turbocharging hospitals remained unchanged through the end of FY 2003, with all 123 hospitals continuing to engage in turbocharging and no other hospitals starting to turbocharge.
There was also good reason to question the assumption that the population of turbocharging hospitals and the behavior of turbocharging hospitals did not change between FYs 2001 and 2003. Industry knowledge of turbocharging may have become more widespread late in calendar year (CY) 2002 after publication of an investment analyst report on the subject. As we previously explained in our March 2003 and June 2003 documents (68 FR 10426 and 10427 and 68 FR 34505, respectively), we believed that it was possible that, before the June 2003 final rule took effect, hospitals that had not previously engaged in turbocharging would take advantage of this new knowledge and increase their charges to catch up to the charging practices of their competitors. Likewise, we had reason to believe that turbocharging hospitals, in anticipation of CMS's regulatory action curbing the effects of turbocharging, would accelerate their turbocharging, either so that they could gain as much as they could from the practice before CMS's regulatory changes took effect or because the hospitals now had less reason to keep turbocharging limited to avoid detection. For these reasons, HHS could not necessarily count on the assumption that aggregate charge inflation between FYs 2002 and 2004 would be significantly lower than predicted by the FY 2000 to FY 2002 data.
In sum, in evaluating how to handle the 123 hospitals in estimating charge growth, we were faced with choices among various uncertain assumptions. We understand the intuitive appeal behind the suggestion that we could have imputed the phenomenon of turbocharging strictly and exclusively to those 123 hospitals, and accordingly assumed that dropping those 123 hospitals' charge data from the charge inflation estimate would remove a source of distortion. But that suggestion itself rests on a set of assumptions. Ultimately, we were faced with a choice between those assumptions and the alternative assumption that, by and large, charge inflation between FYs 2000 and 2002 would adequately predict charge inflation between FYs 2002 and 2004 overall. We did not see reason to conclude that those other assumptions were superior.
We note also that there was only a very limited time interval between the finalization of the June 2003 rule and the publication of the FY 2004 final rule on August 1, 2003, so we had very little time to analyze the potential impact of the June 2003 rule, as finally adopted, on our projections. In addition, the June 2003 rule did not take effect upon publication. Instead, some parts of the rule were to take effect August 8, 2003, and the rest were to take effect October 1, 2003. Consequently, at the time of the FY 2004 final rule, we did not yet have any actual data on hospital charging behavior under the June 2003 rule. We did take several measures designed to adapt the FY 2004 estimates in light of the adoption of the final 2003 rule, and those measures resulted in a significantly lower fixed-loss threshold. But the timing of our efforts constrained our ability to explore additional avenues of analysis we might have otherwise explored.
B. Adjustments to Cost-to-Charge Ratios To Simulate Updates When More Recent Cost Reports Are Tentatively Settled
The court rulings also call for additional explanation of a second issue with respect to each of the FY 2004, 2005, and 2006 IPPS rulemakings. Specifically, the court questioned why, in simulating future DRG payments and outlier payments, we did not apply a downward adjustment to hospitals' cost-to-charge ratios to account for the possibility that, after a more recent cost report is tentatively settled during the coming fiscal year, a given hospital's outlier payments will be calculated Start Printed Page 26362based on an updated, and possibly lower, cost-to-charge ratio.
1. FY 2004
We acknowledge that, by the time of the FY 2004 rulemaking, we had reason to believe that the posited phenomenon was real. The cost-to-charge ratio used to compute a hospital's outlier payments was likely to change at some point during the year once a new cost report was tentatively settled. Furthermore, we had reason to believe that, by and large, a given hospital's updated cost-to-charge ratio would likely be lower than its earlier cost-to-charge ratio, because we had long observed that hospital charges generally increased faster than costs. We also acknowledge that the methods we employed did not include an adjustment to account for this specific phenomenon, though they did account for other effects associated with the general phenomenon of charges increasing faster than costs and the general pattern of decline in cost-to-charge ratios. Our reasons for not incorporating such an adjustment relate to the uncertainty and complexity associated with the task of devising and implementing such an adjustment.
The problem of projecting changes in cost-to-charge ratios over time is qualitatively different from the problem of estimating charge inflation over time. Hospital charges—like hospital costs—are a simple scalar quantity, reflecting tangible real-world activity and measured in dollar values greater than zero. Measuring and projecting changes in dollar quantities of this kind is a relatively common problem, both in the administration of the Medicare program in particular and in business- and finance-related fields more generally. Calculating projected future figures by calculating an estimated percentage change from aggregate figures, and then applying that estimated percentage change to a past measurement, is a familiar approach to that problem.
With respect to outlier threshold projections specifically, at the time of the FY 2004 rulemaking in 2003, we had a great deal of experience estimating changes in quantities of this kind using inflation factors computed from changes in aggregate costs or charges for all hospitals. From 1993 to 2001 (the IPPS rules for FYs 1994 to 2002), we had incorporated a measure of cost inflation to account for year-to-year changes in hospital costs. In 2002 (67 FR 50124), we began accounting for inflation based on year-to-year changes in charges instead of costs. This was not a drastic leap, given that charges and costs are similar quantities measured in the same units.
A cost-to-charge ratio is different in kind. A cost-to-charge ratio does not correspond to a tangible real-world dollar quantity; instead, it is a unitless measure that represents the proportional relationship between two quantities (costs and charges). Charges and costs are virtually always positive values, and charges virtually always exceed costs. Consequently, cost-to-charge ratios virtually always fall between 0 and 1 (instead of ranging from 0 on up as costs and charges do). Within that range between 0 and 1, there is considerable variation in cost-to-charge ratios among individual hospitals, among different types of hospitals, and among geographic areas. This variation is evident in the data we typically make available in connection with our annual IPPS rulemaking (including the impact files and Tables 8A and 8B published in the Federal Register).
As discussed previously, computing an update factor from aggregate figures and applying that estimated percentage change to a dollar figure is a familiar method of projecting future dollar amounts. But it was not evident at the time of the FY 2004 rulemaking that the same approach would translate well to the task of projecting updates to cost-to-charge ratios. If we knew that all hospitals' cost-to-charge ratios were fairly uniform and tended to move in similar ways over time, then we could be fairly confident that applying a uniform update factor based on aggregate changes in costs and aggregate changes in charges would be a satisfactory way to compute projected cost-to-charge ratios. But, as noted previously, we knew there was substantial variation in cost-to-charge ratios across hospitals. We also did not have a solid understanding of whether there was variation across hospitals in how cost-to-charge ratios change over time. Given these factors, at the time of the FY 2004 rulemaking, it was not yet clear to us that it would be appropriate to compute a uniform adjustment factor from aggregate changes in costs and aggregate changes in charges and then apply that same uniform adjustment factor to the cost-to-charge ratios of all hospitals across the board.
At the time of the FY 2004 rulemaking, we also had not yet developed any more complex method that might avoid some of the potential pitfalls of a uniform adjustment factor. A more complex method piling adjustments on top of adjustments could introduce uncertainties of its own, especially when done in the limited time we have to project the annual outlier threshold each year. It is incorrect to assume that adding to the complexity of a simulation method, or increasing the number of factors it purports to take into account, will necessarily improve results.
Even if a clearly sound technique had been available to us for estimating updates to hospitals' historical cost-to-charge ratios, applying such a technique in FY 2004 would have involved an additional complication. As explained in our August 1, 2003 document, (68 FR 45476 through 45477), to account for our change from the use of settled cost reports to the use of tentatively settled cost reports, we elected not to employ actual historical hospital cost-to-charge ratios in estimating FY 2004 payments. Instead, for most hospitals, we used cost and charge data from the most recent cost reporting year to compute estimated cost-to-charge ratios, and we used a different method to calculate estimated cost-to-charge ratios for 50 hospitals identified as likely to have their cost reports reconciled. Thus, even if we had had a method for projecting future cost-to-charge ratios (CCRs) from historical CCRs, we would have had to further modify that method for use with the estimated CCRs we computed for FY 2004.
Perhaps it might have been acceptable to incorporate a cost-to-charge ratio adjustment despite all these uncertainties (and we have done so in more recent years). But at the time of the FY 2004 rulemaking, we did not believe the case for such an adjustment was so compelling as to make such an adjustment essential.
Our decisions are also affected by the limited time we have to devise and implement adjustments to our methods in each year's annual outlier rulemaking. At the time of the FY 2004 rulemaking, we had recently made significant changes to our outlier policies in the June 2003 rule, and we recognized that those changes would have a significant effect on Medicare outlier payments. In making adjustments to our methods, we chose to focus our efforts on those issues we judged most likely to have the most significant relative impact on our projections, while deferring fuller analysis of other issues we judged less likely to have a significant impact, including the effect of updates to CCRs.
We strive to make the best possible estimates, but estimation, by definition, involves approximation, and perfect accuracy is unattainable in our payment projections. Adding additional layers to an estimation technique does not necessarily improve the estimates. And adding complexity to an estimation Start Printed Page 26363method can simply create an illusion of accuracy instead of actual improvements in accuracy.
In light of all these complexities, it was not evident to us in the FY 2004 calculation that any particular adjustment to cost-to-charge ratios would improve our projections. Since we believe we acted appropriately and in accordance with statutory requirements, we are not recalculating the FY 2004 threshold.
2. FY 2005
In our FY 2005 projections, we again chose not to introduce a new adjustment to attempt to account for the updating of cost-to-charge ratios during the year as new tentative cost reports were settled. Most of the factors discussed previously were still present: The fundamental differences in the nature and properties of charges and cost-to-charge ratios; the complexity of simulating the updating of cost-to-charge ratios through either application of a uniform update factor or a more complex adjustment; and our lack of experience with that task.
Also, at the time of the FY 2005 rulemaking, we were still focusing our efforts on the task that we believed had the most significant potential impact on our projections: Monitoring the effects of the June 2003 rule changes and related changes in hospital behavior. We again chose to defer closer examination of the possibility of an adjustment to capture the effect of updates to cost-to-charge ratios.
Also, again, it is important not to overestimate the likely impact of updates to cost-to-charge ratios on the overall robustness of our projections. First, the effect typically comes into play only for part of the year. In our FY 2005 projections, we did not use estimated cost-to-charge ratios as we had done in the FY 2004 rulemaking. Rather, for the FY 2005 final rule, we used CCRs from the March 2004 update of the Provider Specific File, the latest data available (the proposed FY 2005 IPPS rule refers to the same data as the “April 2004” update (69 FR 49277)). CCRs are typically in use for 1 year or more, so, for many hospitals, the CCR in the March 2004 update of the Provider Specific File would be the same CCR used for payment at the beginning of FY 2005, which began in October 2004.
Also, the effect of updates to cost-to-charge ratios is just one of many factors—many of them highly unpredictable—that affect our projections. We note that several commenters on the proposed FY 2005 IPPS rule (69 FR 49276 and 49277) advocated for adjustments to account for CCR updates. Three commenters in particular provided us with analyses that purported to include such adjustments. One of these commenters advocated for a FY 2005 threshold of $26,600, another commenter suggested a threshold of $28,455, and a third advocated for a threshold “no higher than $27,000.” In other words, each of these three commenters purported to incorporate adjustments designed to account for the effect of updated CCRs, among many other factors, yet each arrived at a fixed-loss threshold estimate considerably higher than the $25,800 level we ultimately set.
Because we believe we acted appropriately and in accordance with statutory requirements, we are not recalculating the FY 2005 threshold.
3. FY 2006
The factors discussed previously were all still present for FY 2006: (1) The fundamental differences in the nature and properties of charges and cost-to-charge ratios; (2) the complexity of simulating the updating of cost-to-charge ratios; and (3) our desire to focus on monitoring the aftermath of the 2003 rule changes.
While we carefully analyzed comments suggesting we make a separate adjustment to the CCRs, we again declined to do so, noting that the CCRs we were using from the March 2005 Provider-Specific File were the most recent available, were the CCRs that in many instances Medicare contractors would be using to make outlier payments in FY 2006, and were approximately 3 percent lower than the CCRs used in the FY 2006 proposed rule (70 FR 47494).
As had been the case in FY 2005, two commenters submitted recommendations based on an analysis that purported to account for updates to CCRs, and those recommendations were in turn endorsed by many other comments. These commenters advocated for a threshold of $24,050, higher than the $23,600 level that we computed. This lent further support to our decision to defer closer study of the effect of updates to cost-to-charge ratios.
Because we believe we acted appropriately and in accordance with statutory requirements, we are not recalculating the FY 2006 threshold.
III. Collection of Information Requirements
This document does not impose information collection requirements, that is, reporting, recordkeeping or third-party disclosure requirements. Consequently, there is no need for review by the Office of Management and Budget under the authority of the Paperwork Reduction Act of 1995 (44 U.S.C. 3501 et seq.).
End Supplemental Information
Dated: May 14, 2019.
Administrator, Centers for Medicare & Medicaid Services.
Dated: May 28, 2019.
Alex M. Azar II,
Secretary, Department of Health and Human Services.
[FR Doc. 2019-11796 Filed 6-3-19; 11:15 am]
BILLING CODE 4120-01-P