The Medicare Advantage option for beneficiaries continues to show rapid growth; after this year’s open enrollment period, it appears that over a third of all Medicare recipients will be enrolled in an MA plan. CMS’ method for paying these plans has evolved and CMS now attempts to pay no more than it would cost CMS if the beneficiary stayed in FFS Medicare. The heart of this strategy is a risk adjustment methodology. Because this methodology feeds off information regarding the patient’s health status and health needs, MA plans have naturally focused on ensuring that this information is as comprehensive as possible. Several new reports discuss the data used for this risk assessment. (Milliman Paper) (Avalere Paper) (GAO Report) Originally CMS relied on specific forms reporting a beneficiaries health status. So the plans began hiring firms to go out and do “assessments” designed to generate as many diagnosis codes as they could. So CMS countered by reducing plan payments by an upcoding adjustment and by requiring that diagnosis codes show up in an actual provider visit and/or that encounter data be used in the risk score calculation. For 2017, and proposed for 2018, encounter data would be weighted 25% in the calculation. Eventually CMS hopes to move to use of all encounter data for the risk score, which makes total sense and would eliminate the nonsense of assessments, especially in the home, solely for the purpose of jacking up reimbursement. But there are issues with the shift and with the quality of encounter data.
Avalere’s contribution is just a one-page summary of a report to be issued later this month. But that page summarizes plans’ concerns about use of encounter data. According to Avalere, 35% to 40% fewer HCC diagnoses were reflected in encounter data, leading to 16% lower risk scores for 2016 and consequently lower payments to plans. The other way to look at this is that the current assessment-heavy approach is leading to a lot of diagnoses being reported that the patient isn’t actually being treated for. Milliman’s short paper comes to a similar conclusion regarding the impact of encounter data use, but with what seem to be more realistic numbers. Milliman estimates about a 4% lowering of risk scores, although it found that for 87% of members the risk score would be the same under the current process and the one using solely encounter data. Milliman also notes a number of technical issues in the encounter data submission and reconciliation process with CMS. Those technical issues are the focus of the General Accounting Office report. The current report is an update on an earlier one noting significant shortcomings in CMS’ process to validate encounter data. This report says CMS hasn’t made much progress. Part of what GAO is recommending is a periodic review of medical records to verify diagnoses reported in encounter data. Seems like a random audit of a sample of submissions would be a good idea.
Avalere and Milliman are both high quality groups, but they typically work for health plans. Avalere’s paper feels a little like advocacy, Milliman’s is more neutral and problem-solving oriented. Assuming that risk-adjusted payments are the appropriate approach for CMS (an assumption which should be tested), using real-life diagnosis codes derived solely from the actual receipt of health services is the only logical method for getting the data to feed into your risk-scoring algorithms. The health plans will need to ensure that they have a robust combination of claims databases and encounter tapes from capitated providers. Harking back to yesterday’s post regarding waste in health systems, this area is a bit of an example. Health plans have spent a lot of money solely on jacking up risk scores and CMS probably has overpaid as a result. CMS has a habit of creating regulatory schemes that lead to unintended consequences that raise both administrative and health spending. Might be time to revisit some of those approaches.