GAO Says Medicare Overpays MA Plans Because of Incomplete Data and Faulty Assumptions

Although Medicare Advantage plans have seen lower premiums and higher enrollment, the Government Accountability Office (GAO) recently reported that the measures CMS uses to estimate the costs of covering Medicare beneficiaries are inaccurate, resulting in higher payments to Medicare Advantage (MA) plans than Medicare would make to cover the same people through traditional, fee-for-service (FFS) Medicare.  What’s more, GAO says, CMS’ choice not to use the most recent available data compounds the inaccuracy and the excess expenditures.

CMS projects expenditures for Medicare beneficiaries based on enrollment and on their claims for the previous year.  Beneficiaries are assigned a risk score comparing their likely expenditures to the average Medicare beneficiary. However, MA plans also report diagnosis codes for their members.  For several years, at least since 2004, the MA plans have reported these codes more and more comprehensively, whether or not the diagnoses significantly affected the beneficiaries’ use of services that year. As a result, their members appear sicker than FFS beneficiaries.

The apparent increase in risk translates into higher projected costs, resulting in higher capitated payments to MA plans. The risk scores of MA plan members have risen faster than the scores of FFS beneficiaries, raising concerns that the coding alone increases Medicare payments to MA plans. According to GAO, FFS providers have no incentive to list a diagnosis code for every condition a beneficiary has because their payments are based on services furnished, not an estimate of future need.

To compensate for the potential excess payments, CMS must estimate the extent to which the diagnostic coding alone has increased beneficiaries’ risk scores. The Deficit Reduction Act of 2005 (DRA) (P.L. 109-171) directed CMS to make adjustments in 2008, 2009 and 2010 to the extent that it could identify the artificial increase in the risk scores.  GAO says the agency did not do so until 2010. That year, CMS determined that the coding differences accounted for 3.4 percent of the risk scores for MA plan members and adjusted MA payments accordingly. The adjustment saved Medicare an estimated $2.7 billion.

GAO, however, estimated that the coding differences accounted for 4.8 to 7.1 percent of the MA plan members’ risk scores. The two estimates are the result of different assumptions about the effects of the coding differences over time. GAO found that the impact of the MA plans’ coding grew between 2005 and 2008. Therefore, it estimated 2010 costs two ways—one, assuming that the differences between the risk scores continued to grow between 2008 and 2010 and the other, assuming that the differences were the same in 2010 as in 2008.

According to GAO, CMS’ estimates are less accurate for three reasons. First, CMS did not consider data from 2008; this information was not available when CMS initially projected rates for 2010 but was out before CMS published the final rates. Second, the effects of the increased risk scores are cumulative, so that the gap between MA and FFS beneficiaries’ scores widens every year. The growth is less apparent when the most recent available data is not included.

According to GAO, CMS also did not update its estimates for 2011 and 2012 with more recent data, even though it was available. Instead, it proposed to make the same adjustment in 2011 and 2012 that it made in 2010, a decrease of 3.4 percent.The Health Care and Education Reconciliation Act (HCERA) (P.L. 111-152) requires CMS to continue adjusting for diagnostic coding until it begins to use data that include cost and use as well as diagnostic coding.  In addition, for 2014, HCERA directs CMS to reduce the risk scores of MA plan members by at least another 1.3 percent

The third reason that GAO’s estimates differed from CMS’ is that CMS compared the MA and FFS populations only on the basis of age and mortality. GAO, however, examined other characteristics of the populations that would affect their use of health care services, specifically, gender, concurrent Medicaid enrollment, location of residence and whether the initial enrollment in Medicare was based on age or disability. Individuals who initially enrolled because of disability rather than age would be expected to need more health services. It was estimated that CMS’ decrease of 3.41 percent saved Medicare $2.7 billion in 2010. However, GAO estimates that Medicare should have saved an additional $1.2 billion to $3.1 billion in 2010. If the differences in risk scores continue to grow, the excess payments will, too.

GAO recommended that CMS use the most recent data available, account for the cumulative impact of the increased score from year to year and incorporate the additional factors, gender, original basis for Medicare eligibility, health status, location of residence and concurrent Medicaid enrollment.  According to GAO, CMS responded that it found GAO’s analysis “informative”  and asked questions about some aspects of the methodology.

GAO’s criticism of CMS’ methodology seems logical. Still, making economic predictions is always risky.  If only they could assume a can opener.