In the Deficit Reduction Act of 2005, Congress gave the Centers for Medicare and Medicaid Services (CMS) the statutory authority and obligation to implement a “coding intensity adjustment” for Medicare Advantage (MA) plans, to adjust for differences in patterns of diagnosis coding between MA and traditional Medicare.
Under both the Obama and Trump administrations, the coding intensity adjustment implemented by CMS has consistently been substantially smaller than justified by the data. The fiscal problem created by inadequate adjustment for coding intensity is large. Under reasonable assumptions about the rate of growth of MA coding, CMS will overpay MA plans by $200 billion over the next decade if the coding intensity adjustment remains at 5.91 percent, its current level.
Why does this overpayment persist? The annual decision about the size of the coding intensity adjustment is made by political appointees at CMS, the Department of Health and Human Services, and the White House, and the political benefits from implementing a smaller than empirically justified coding intensity adjustment far outweigh the potential political gains from protecting the taxpayers and creating a stable and level playing field between MA and traditional Medicare. To fix this problem, Congress should remove discretion from CMS and specify in statute an even-handed method of computing coding intensity. A statutory formula for measuring coding intensity would bring MA payment more in line with payment for other providers, in which Congress, not CMS, decides how much providers will be paid.
Below, I offer suggestions regarding what a statutory coding intensity adjustment formula should look like. Of course, Congress, like CMS, is subject to political pressures that will make enacting such a formula difficult. But unlike CMS, Congress could use the $200 billion in avoided MA overpayments to fund other priorities, offering countervailing political incentives for legislators to save taxpayer money and level the playing field between MA and traditional Medicare by adopting a statutory measure of relative coding intensity.
The Problem
CMS pays MA plans based on the health status of the beneficiaries who enroll, paying more for older and sicker enrollees and less for younger and healthier enrollees. This “risk adjusted” payment mechanism is intended to encourage MA plans to develop systems of care that are attractive to beneficiaries most in need and to prevent MA plans from profiting by attracting only good risks.
The health status of MA enrollees is measured using diagnostic information submitted by the MA plans themselves. As a result, MA plans have strong incentives to identify and report as many diagnoses as can be supported by the medical record, incentives that are not present in traditional Medicare. Although some MA plans may have used fraudulent strategies to inflate enrollee risk scores, many legitimate strategies are available to MA to increase risk scores. The incompleteness of diagnostic coding in traditional Medicare provides ample opportunities for MA plans to increase the identification and reporting of diagnoses. For example, in traditional Medicare, approximately 40 percent of beneficiaries coded with quadriplegia in a 12-month period do not have a diagnosis of quadriplegia appear on any claim in the subsequent 12 months.
Discretion For CMS
Recognizing the likelihood that MA plans will report diagnostic information differently than is reported in traditional Medicare, in the Deficit Reduction Act of 2005, Congress gave CMS the authority to implement a “coding intensity adjustment” to adjust for differences in coding patterns between MA and traditional Medicare. In 2010, CMS implemented a 3.41 percent coding intensity adjustment, reducing MA risk scores by that amount. The Affordable Care Act, and subsequently the American Taxpayers Relief Act of 2012, created a schedule of minimum adjustments, starting at 4.71 percent in 2014, increasing to 5.91 percent in 2018.
CMS retained the authority and obligation to implement adjustments larger than the statutory minimum if the data indicated that a larger adjustment was needed. However, it has not yet done so, despite strong evidence that a larger adjustment is needed to compensate for differences between MA and traditional Medicare in coding patterns. The measured risk of MA enrollees relative to traditional Medicare increased from 95.0 percent in 2007 to 106.2 percent in 2015 (exhibit 1).
Exhibit 1: Average risk score of Medicare Advantage enrollees as a percentage of average risk score of traditional Medicare beneficiaries, 2007–15
Source: Richard Kronick and Pete Welch, 2014; Paul Jacobs and Richard Kronick, 2018.
MA enrollment grew substantially from 2007 to 2015, and it is theoretically possible that the increase in relative risk reflects real changes in the relative health of MA enrollees. However, analyses of other data sources—including mortality rates, survey data from the Medicare Current Beneficiary Survey, data on switchers from traditional Medicare to MA, and prescription drug use data—show conclusively that there was little, if any, change in real relative risk over this time period. The 11-percentage-point increase in measured relative risk from 2007 to 2015 appears largely to be the result of differences in coding patterns between MA and traditional Medicare.
Perverse Political Incentives
The main reason that CMS has not implemented a larger coding intensity adjustment is that there is little political gain from doing so, and the decision about the coding intensity adjustment is, ultimately, made by political appointees. If CMS were to implement an adjustment that is larger than the statutory minimum, MA plans would make their displeasure known to members of Congress. Furthermore, an adjustment larger than the statutory minimum might result in higher MA premiums or fewer extra benefits for enrollees, raising the likelihood of constituent dissatisfaction.
In even numbered years, information on MA premiums and benefits is made public soon before the November elections. Members of Congress (and, every four years, the president) are concerned that voters, likely encouraged by information fed to them by MA plans, may blame incumbents for increased premiums or reduced benefits. This is not a merely theoretical argument: Federal legislators have repeatedly exerted strong bipartisan pressure on CMS in response to industry and constituent concerns about MA payment policy.
The political cost of more aggressive action by CMS on coding intensity is clear. The political benefits are much less clear: The deficit would be slightly smaller but not enough to change interest rates or anything else that voters would notice. Given this imbalance between political costs and benefits, it is not surprising that CMS has stuck with the statutory minimum, despite strong evidence that larger adjustments are warranted.
Proposed Solution: Specify The Coding Intensity Adjustment Formula By Statute
This problem could be solved by removing discretion from CMS when calculating the coding intensity adjustment, which Congress could do by writing a method for calculating the adjustment into statute. This approach would make payment for MA plans more like payment for other providers—Congress typically tells CMS how to set payment rates, leaving relatively little discretion to CMS.
MedPAC’s Method
There are several reasonable options for specification of a formula to calculate the coding intensity adjustment. For example, Congress could direct CMS to adopt the approach used by the Medicare Payment Advisory Commission (MedPAC) in the past few years to estimate the magnitude of differential MA coding. That method compares the rate of growth in risk scores for MA enrollees with the rate of growth in risk scores for similar beneficiaries in traditional Medicare. For 2015, MedPAC estimated coding intensity increased MA risk scores by 10 percent. The MedPAC estimates for 2016 and 2017 are slightly lower—8 percent and 7 percent respectively—largely because of changes in the model used to calculate beneficiary risk scores. However, further changes in that model in 2018 and subsequent years are likely to increase the estimated coding intensity effect.
The MedPAC method is reasonable but requires complex calculations, would be difficult to audit, and is somewhat difficult to explain. It would be extraordinarily difficult to write into statute in enough detail to remove discretion from CMS in implementation.
A Better, More Practical Approach
An alternative, and, in my view, preferable approach would be to require CMS to calculate the coding intensity adjustment using a method that CMS floated as a trial balloon in the Advance Notice for 2016 MA rates. First, CMS would calculate the relative risk of MA using the diagnostic risk adjustment system, which, as noted above, was 106.2 percent in 2015.
Second, CMS would calculate the relative risk of MA using information on the age, gender, Medicaid status, and institutional status of enrollees—that is, using the Adjusted Area Per Capita Cost (AAPCC) system that was used prior to the introduction of diagnosis-based risk adjustment. In 2015, MA relative risk using these four factors was 97 percent. That statistic is less than 100 percent primarily because MA enrollees were less likely than traditional Medicare beneficiaries to be institutionalized, and institutionalized enrollees have high risk scores.
Third, CMS would divide the second number by the first, and subtract the quotient from 1.0. The result, which would have been 8.7 percent in 2015, would be the coding intensity adjustment.
This is a simple approach to explain and write into statute. It assumes that MA enrollees are no healthier and no sicker than demographically similar traditional Medicare beneficiaries. This assumption of no favorable or adverse selection is likely still a bit generous to MA—although the extent of favorable selection for MA appears to have declined over the past decade, there is no evidence to suggest that MA enrollees are, on average, sicker than demographically similar traditional Medicare beneficiaries, and some evidence to suggest they are a bit healthier.
The method of calculating coding intensity suggested above would return to the “budget neutrality” method that CMS, under the leadership of Tom Scully, implemented when the CMS-HCC (hierarchical condition category) risk adjustment model began to be phased in in 2004. At that time, MA relative risk, using the CMS-HCC model, was well below 1.0, and MA revenues would have been reduced if the CMS-HCC model had been implemented without adjustment. To avoid this outcome, CMS implemented a “budget neutrality adjustment,” increasing CMS-HCC scores by the ratio suggested above. In the Deficit Reduction Act of 2005, Congress directed CMS to phase out the budget neutrality adjustment by 2008. Ironically, in 2008, the measured relative risk of MA using the CMS-HCC model became greater than the measured relative risk using the AAPCC model, so if budget neutrality had been left in place, it would have resulted in decreases in payments to MA.
One potential substantive disadvantage to this approach is that at some point in the future it is theoretically possible that MA enrollees will, on average, be sicker than demographically similar traditional Medicare beneficiaries, and that the formula will unfairly penalize MA contracts. At the risk of giving CMS discretion that might be abused, CMS could be given the authority to use other data—such as data on mortality rates, prescription drug use, and the Medicare Current Beneficiary Survey—to analyze whether there is adverse selection against MA. Under this option, if CMS certifies that there is strong evidence of adverse selection against MA, then it would have the authority to adjust the statutory formula to reflect that adverse selection. Political influence on this decision could be lessened by requiring the CMS Office of the Actuary (OACT), not the agency’s Center for Medicare, to make the certification. OACT has traditionally been less responsive than other parts of CMS to the preferences of political appointees.
A second potential concern about the proposed approach, as well as about the status quo adjustment of 5.91 percent, is that there is substantial variation across MA plans in how aggressively they have increased risk scores. Any industrywide adjustment will pay too much to plans that have been most aggressive in coding, too little to plans that have been least aggressive, and leave in place the incentive to code as many diagnoses as legally permissible. Consideration could be given to contract-specific adjustments, but it is far more important to first get the coding intensity adjustment, on average, correct. After first fixing the $200 billion problem, Congress and CMS could then consider whether and how to implement contract-specific adjustments.
Summing Up
Under reasonably conservative assumptions about the rate of growth of coding intensity, using the budget neutrality method to calculate coding intensity would likely result in approximately $200 billion in savings to Medicare over the next decade compared to current policy in which the coding intensity adjustment remains at 5.91 percent. Just as CMS is reluctant to anger the MA industry and its enrollees, members of Congress will be reluctant as well. However, unlike CMS, which does not get to spend any of the savings generated by more appropriate actions on coding intensity, Congress can choose to use the $200 billion in savings for additional spending or tax cuts, providing reason to hope that legislators will, at some point, do the right thing.
