UC San Diego

In Chapter 1, utilizing Affordable Care Act reforms to a federal program whose reward structure is based on care supplied to multiple patient payer categories, I examine how hospitals adjust payer group care volumes in response to own- and cross-price changes. Constructing simulated instruments to estimate the reform-induced marginal revenue shocks, I find that hospitals respond to positive price changes with significant positive volume adjustments (or, conversely, to negative price changes with negative volume adjustments). I estimate that a $100 increase in reimbursement per payer group patient day causes hospitals to increase non-low income Medicare service volume by 2.9%, and Medicaid service volume by at least 5%. My results also suggest that a price shock to a single payer group causes spillover supply adjustmentsfor other patient groups, both private and public. These estimated cross-price elasticities are larger for non-profit hospitals than for for-profits, consistent with models of hospital behavior in which non-profits more readily cross-subsidize care across payer groups in response to a single group’s positive price shock.

In Chapter 2, we analyze prospective changes to the Medicare Advantage (MA) risk adjustment payment system and evaluate how effectively they normalize MA risk scores relative to Traditional Medicare (TM) and reduce variation in contract risk scores caused by differential coding. We also analyze the effects of a risk adjustment reform the Centers for Medicare & Medicaid Services (CMS) began implementing in 2024 and find that it will have a substantial impact on MA payments: if it had been fully implemented in 2021, the average MA risk score would have been 112% of TM’s instead of 120%. The contract-level equity and efficiency problems created by differential MA coding would be solved almost completely if CMS went further and excluded 10 additional Hierarchical Condition Categories from its model. With these excluded, the average MA risk score would be very similar to TM’s, and there would be very little relationship between contract-level coding intensities and risk scores.

In Chapter 3, we analyze differences between MA and TM patterns of diagnostic coding, estimate the effects of differential coding on MA contract risk scores, and quantify the additional payments each contract received through those effects. We aggregate our contract-level results to each corresponding MA sponsor. In 2021, the average MA risk score was 0.189 higher than if MA coding patterns were identical to TM’s. The average risk score at UnitedHealth was 0.284 higher than if its plans coded identically to TM, much greater than the industry average, and significantly greater than Kaiser Permanente’s 0.036. MA plans received an estimated $33 billion in additional payment in 2021 through differential coding, $13.9 billion of which went to United. Any MA payment policy reform which targets differential coding will have a significantly larger effect than average on United, virtually no effect on Kaiser, and intermediate effects on other large sponsors such as Humana and Anthem/Elevance.