Ex ante evaluation of risk adjustment models for prospective provider payment: a conceptual framework and empirical application.

Alternative payment models (APMs) aim to improve efficiency and fairness in healthcare by shifting financial responsibility from payers to providers. Given their prospective nature, APMs require effective risk adjustment (RA) to prevent risk-selection incentives. RA design comes with complex trade-offs between risk selection, cost control and gaming. In the light of these trade-offs, thorough ex-ante evaluation of RA models is crucial. Traditionally, RA-model evaluation in the context of APMs has heavily relied on statistical metrics like R-squared. While useful for assessing model fit, these metrics often fail to capture the full spectrum of relevant incentives. This study therefore addresses the question: "What do meaningful incentive metrics for ex-ante evaluation of RA models look like in the context of prospective APMs for healthcare providers?" We conducted a literature review and consulted experts to synthesize existing work on RA evaluation. This informed the development of a conceptual framework for defining incentive metrics, distinguishing among risk-selection, cost-control, and gaming incentives. We applied our framework in a simulation of prospective payments to primary care practices (PCPs) in the Netherlands, using 2019 claims data from 346 PCPs (N = 1.4 M patients). The analysis focused on selection incentives, comparing traditional statistical metrics with metrics derived from our framework. Results show that statistical metrics like R-squared fall short in assessing selection incentives compared to our incentive metrics. This highlights the need for tailored incentive metrics for the ex-ante evaluation of RA models that are grounded in a thorough understanding of relevant provider behaviors in the light of APM goals.