Grant Description: To what extent do payment systems lead to misaligned incentives for quality improvements in specific service areas? This project examined a methodological model known as “shadow-pricing” to assess risk selection. A shadow price is a measure of the relative profitability of a health care service, taking into account patterns of spending across all services. The higher a service’s shadow price, the lower the profitability of that service, and the greater the incentive for a health plan to reduce expenses on that service. Building on a model developed by Frank Glazer and Thomas McGuire (2000), the researchers tested a tool designed for purchasers to improve payment incentives. Specifically, the researchers (1) expanded the theoretical foundation for the shadow-price approach to payment incentive assessment; (2) assessed selection incentives under the current payment policies of the Massachusetts State employee insurance program, demonstrating the customizability of the shadow price approach; and (3) simulated and analyzed alternative payment methods to identify the most effective strategies for reducing selection incentives. The researchers hypothesized that their simulation would reveal that purchasers will gain substantial benefits from implementing a combination of risk adjustment and blended payment, possibly in conjunction with high-risk pooling. The objectives of the project was to develop a technique for purchasers to use to identify how a payment system leads to misaligned incentives for quality improvements and to provide purchasers with a tool to allow them to assess the incentives for selective quality distortions within their own plans.
Policy Summary: This grant explored tools purchasers can use to identify health plan incentives to compete to avoid the sick rather than provide quality care. The researchers used three different metrics of selection incentives to estimate how an insurer would want to distort service offerings to attract profitable enrollees (“risk selection,” “cream skimming,” or “cherry picking”). They applied these metrics to indemnity claims and managed care encounter data on both medical and pharmacy spending for 2001 and 2002 from the Group Insurance Commission (GIC) of the Commonwealth of Massachusetts—one of the largest purchasers in New England. The results revealed strong financial returns to risk selection, as indicated by all three selection indices as well as by the double-digit profits an insurer could earn if it could exclude unprofitable patients. For this population, services most vulnerable to stinting were cardiac care, diabetes care and mental health and substance abuse services. In other words, insurers had incentive to ‘dump’ enrollees with expensive heart conditions, diabetes or mental health and substance abuse problems (and to ‘cream’ enrollees with skin problems or conditions of the eyes, ears, nose and throat). Risk adjustment and mixed payment considerably mitigated these selection incentives. The advantages of mixed payment were evident when the researchers calculated returns to risk selection over a full range of supply-side cost sharing. Doubling the fraction of costs borne by plans or providers more than doubled the rewards to risk selection. This confirmed that employers and other payers can use mixed payment—with even a small amount of risk sharing—to significantly reduce insurers’ strong financial temptation to invest in risk selection. As a tool for measuring selection incentives, these three metrics—two developed previously by researchers Ellis, Frank, Glazer and McGuire, and one that the researchers proposed—have at least four advantages. First, they are more evidence-based than case-by-case analysis or deductive reasoning regarding all-encompassing categories (e.g., overuse with fee-for-service, underuse with capitation). Second, analysts can tailor the measurement to specific patient populations (e.g., by using the expenditure and diagnosis patterns of that population). Third, purchasers can analyze various alternative payment methods prior to implementation to examine alignment of payment incentives with quality improvement. This may contribute to better contracting and clearer targeting of quality assurance programs or pay-for-performance initiatives. Fourth, since the metrics measure incentives, not actual behavior, they do not raise issues of legal liability. They should be complemented by analysis of actual selection behavior, to document the extent to which plans respond to the financial temptation to risk select.