Grading Risk-Adjusted Health Plan Payments: Fit, Power, Balance

Risk adjustment formulas are based partly on enrollee traits, such as age and gender, but also on health care events like a doctor visit or hospital stay during which a condition is diagnosed.  The diagnostic information contained in claims from health care events account for much of the ability of risk adjusted payments to match enrollee costs without conditioning on actual realized expenses.

The use of risk adjustment to discourage cream skimming is in tension with another, distinct goal of payment systems: to create incentives for the insurer to properly internalize each dollar of utilization that the enrollee generates. Even partially tying payments to realized costs reduces such incentives to supply.  Risk adjustment in the US Medicare system and in major systems in Europe is based on past diagnoses.  Recently, however, in connection with health care reform, the federally proposed risk adjustment formula adds diagnoses to the formula, and significantly, the proposed formula is based on concurrent health care encounters, not past.  Furthermore, extensive cost-sharing features (e.g., reinsurance) are also part of Exchange plan payment. Tying the risk adjusted payment to an instance of utilization creates incentives to make visits and hospitalize patients – a set of incentives that has been recognized but not quantified in the literature. 

This paper characterizes the properties of risk-adjusted plan payment systems, analytically and empirically, with the capitation payment formula proposed for the new Exchanges as part of the Affordable Care Act (ACA) serving as the principle application.   Specifically, we study the fit, power and balance of payment systems.  “Fit” describes how well variation across enrollees in plan costs is explained by variation in payments and is usually reported as an R-squared from a risk adjustment regression. But this statistic fails to capture the full relationship of payments to costs.  We generalize the measure to include the fit of the entire payment system which consists not only of risk adjustment but of cost-sharing features such as reinsurance.  “Power” is meant in the sense of the power of a contract: it is related to how expenditure by plans is compensated on the margin by the payer.  “Balance” assesses the differences in power across various types of services, such as mental health care.

The major precursor of our paper is McClellan (1997) which assessed the de facto incentives in Medicare’s Diagnosis-Related Group (DRG)-based “Prospective Payment System” for paying hospitals.  In our terms, McClellan showed that the power of the DRG system was 45%.  We are unaware of any research applying McClellan’s ideas and methods to health plan payments.    

Using two years of claims data from an employed population, we assess the de facto incentives in the plan payment systems used to pay Exchange plans under the ACA.  The Exchange payment system is particularly complex so we take it apart to assess the partial contribution of some of its key features, such as the decision to pay plans with a concurrent rather than prospective risk-adjustment formula.