Following the Money: do Medicare Advantage Revenues from Coding Intensity Increase Benefit Generosity?

Since the introduction of Medicare Advantage (MA), policymakers have voiced concerns about how to appropriately set payments for MA plans. In particular, because MA plans are paid based on beneficiary diagnoses and they are able to code diagnoses more comprehensively than beneficiaries in Traditional Medicare (TM), there is widespread agreement that taxpayers pay more for beneficiaries enrolled in MA plans (GAO, 2012; Kronick and Welch, 2014; MedPAC, 2018).

Researchers have clearly documented this pattern of coding intensity among MA plans, but little is known about whether these increased payments affect plan benefits or premiums. When an MA plan receives revenues in excess of its costs, it can retain revenues as profits as long as doing so does not run afoul of minimum loss ratio or other regulations. Alternatively, any additional revenues could help plans offer additional benefits or lower premiums in the hopes of attracting and retaining enrollees.

This paper tests whether MA plans change benefits, cost-sharing, or premiums when their payments exceed what would be commensurate with an independent measure of the underlying risk of their beneficiaries. We identify the health risk of Medicare beneficiaries using a technique developed by Jacobs and Kronick (2018) that imputes health risk for MA enrollees using their Medicare Part D claims data, which is independent of the medical diagnoses plans report for payment purposes. We limit our sample to MA beneficiaries enrolled in Part D from 2007 through 2015 and define coding intensity as higher administratively-assigned risk scores holding constant beneficiaries’ prescription drug-based risk scores. We used administrative data on MA enrollment and Part D claims as well as publicly-available data on MA plan costs, benchmarks, and plan characteristics. Because MA beneficiaries likely choose plans in ways that depend on their health status, we compute all of our results at the county level to address the potential endogeneity of individual health risk. Additionally, because plans likely set benefits and premiums to reflect the preferences and health status of beneficiaries in particular areas, we include county and year fixed effects.

Our preliminary estimates suggest a weak relationship between MA plan benefits and higher levels of coding intensity. A $1 increase in monthly MA revenues from increased coding intensity, holding health risk constant, was associated with a reduction in both expected beneficiary cost-sharing (-$0.10) and plan premiums (-$0.03). Models with lagged health risk measures did not produce meaningfully different estimates. These early results suggest that plans may be profiting from their coding intensity efforts and are consistent with previous evidence suggesting that competition in the MA marketplace is relatively weak (Song et al. 2013). Policymakers have increasingly relied on risk adjustment to help stabilize risk pools and establish more competitive choice environments for health plans. Understanding how plans utilize revenues to provide health benefits and set premiums is instructive for both the MA marketplace as well as other managed competitions settings, including the Marketplaces and some Medicaid managed care settings.