Abstract
Payments to Medicare Advantage (MA) plans are adjusted by a risk-score model that is calibrated on diagnostic and demographic data from traditional Medicare beneficiaries and then applied to MA beneficiaries. If MA plans capture more diagnostic codes than traditional Medicare, they receive payment that is higher than the amount that would be spent in traditional Medicare. Although most previous research has focused on the coding practices of MA plans, less attention has been paid to the completeness of coding in traditional Medicare. We analyzed 2017-19 traditional Medicare claims data and MA encounter data to compare the persistence of diagnostic coding for sixteen chronic conditions. Our primary analysis found that the lack of persistent coding of these conditions in traditional Medicare accounted for 2.85 percentage points, or 22.3 percent, of the 2020 traditional Medicare/MA risk-score gap, translating to $8.1 billion in Medicare spending. In our most conservative sensitivity analysis, this discrepancy accounted for 9.8 percent of the total gap, although this was likely an underestimate, as it excluded acute conditions and incident chronic conditions. To resolve this discrepancy, a comprehensive approach addressing coding practices in both MA and traditional Medicare may be required.