A controversy that has dogged the Medicare risk contracting programs from the start is whether the participating health plans attract beneficiaries with fewer health needs and whether the payments to them adequately or excessively compensate for enrollees anticipated medical costs. A new study in Health Services Research attempts to compare the current risk adjustment methodology used by CMS with other methods of assessing beneficiary health and health needs. (HSR Article) Because it has been widely recognized that MA plans do code member diagnoses more aggressively, Congress has mandated that CMS reduce payments by a coding intensity adjustment, which currently is around 6%. The notion has persisted that MA enrollees are healthier than those remaining in the FFS arm of the program, even though health plans actually probably have an incentive to seek out sicker beneficiaries, for which they will get higher payments. These authors used prescription drug data to estimate relative health risks for beneficiaries and compare that to the standard CMS risk adjustment methodology. It has long been recognized that medication use can predict overall health risk and health care use. Because Part D drug plans are not paid based on diagnoses derived from drug data, they should not have an incentive to overcode those diagnoses, and in any event, most drugs have fairly limited diagnostic indications.
Out of the 20 disease groups with the highest spending, MA members had lower prevalence rates than did FFS ones in 18 of the groups, with an average prevalence of about 91% that for FFS beneficiaries. This would suggest that MA members should have lower health spending. The ratio of MA to FFS risk scores, calculated using the drug-based model, was in fact about .93, so it was fairly close to the coding adjustment currently being used to reduce payments the MA plans would otherwise receive. The average risk score for traditional Medicare beneficiaries was relatively flat over the period 2008 to 2015, while rising about 4 percentage points for MA members according to the model. But as reported to CMS by the MA plans, that average risk score rose 9.8 percentage points, which again may reflect more aggressive efforts to identify and document health issues. So the authors conclude that while the health risk of MA and FFS beneficiaries is converging, the MA plans are reporting the rising risk of beneficiaries faster than it is actually occurring, in fact about two and a half times faster. But again the coding adjustment actually seems to be keeping up with this behavior fairly well, although it may still mean that Medicare is overpaying the MA plans by as much as $16 billion. This debate is likely to continue for some period of time. Use of encounter data for risk-scoring will eliminate some of the game-playing, but not all of it.