Utilization Thresholds in Risk Adjustment Systems

Risk adjustment systems, that reallocate funds among competing health insurers, often use risk adjustors that are based on utilization documented in medical claims. The level of utilization that triggers an adjustor, i.e. the utilization threshold, is frequently chosen implicitly and uniformly. I empirically study utilization thresholds in the setting of the U.S. Marketplaces. Simulat-ing alternative levels of thresholds for adjustors, I find thresholds that improve the prediction fit, by up to 9.6% in some disease groups. Using newly-defined measures for the incentives to game the system, I show that for some thresholds a tradeoff between fit and gaming-incentives does not exist. To guide a choice of multiple utilization thresholds, I employ a regression tree algorithm that considers both fit and gaming incentives.