Out of concern that Medicare Advantage plans were preferentially seeking out healthy beneficiaries, in hopes of making more profit, CMS switched to a payment system that relies heavily or risk scores for each beneficiary, which supposedly reflect their likely health costs. As might be expected, now the concern is that the risk scores are being manipulated in some manner to get more revenue than is appropriate to pay for care for the beneficiaries actual health needs, which is not a trivial concern, since Medicare Advantage probably will account for almost 35% of beneficiaries in 2018 and I suspect will constitute at least half of all beneficiaries in a few years. A working paper from the Congressional Budget Office suggests some truth to this concern. (CBO Paper) The researchers used data from 2008 to 2013 and compared the risk scores for beneficiaries who switched from traditional Medicare to Medicare Advantage with those who remained in the fee-for-service arm. They further looked at the effect of a beneficiary staying with the same MA plan over time.
Risk scores have a demographic component but also a component based on the beneficiaries health status as reflected by the hierarchical condition categories methodology. The data that feeds the HCC calculation are diagnoses codes that are collected and reported to CMS by the plans, and, for now, to a lesser extent actual encounter data from visits to providers. So as you can imagine the plans might have an incentive to maximize the number and intensity of diagnoses codes. And apparently they have been doing a good job at that. Annual wellness visits become an opportunity to ensure that every possible health condition is identified and recorded. The same incentive doesn’t generally apply in the fee-for-service Medicare world, creating a gap that raises payments, since the average fee-for-service cost for a beneficiary is the unadjusted payment base for MA and that average fee-for-service cost is designated to average out to a risk score of 1. A higher risk score means a greater adjustment to that base.
Prior research found a coding intensity difference of 4% to 10%. The CBO study used people who were in FFS in all of 2008 and then tracked those who switched in 2009 or later and stayed in MA versus those who remained in the FFS program. Under this recent CBO analysis it appears that although risk scores for those beneficiaries who became MA members were initially about a tenth of a point lower than those who remained in FFS, their risk scores have grown by 42% over 2009 to 2013, versus 29% for the stayers. The difference in average risk scores expanded over time, from 5% in 2008 to 8% in 2013. In addition, the longer the beneficiary was in MA, the more the coding gap grew. CMS already adjusts payments downward by over 5% to account for this coding effect. If 100% encounter data is used for risk-scoring, then this adjustment will go away, although I suspect plans and providers will still be more vigilant about full documentation for MA members. One obvious observation is that it is hard to design a reimbursement system that doesn’t get gamed in some manner.