Robust Risk Adjustment in Health Insurance

Abstract

Objectives.  To create risk scores in a manner that incorporates uncertainty in the models used to redistribute payments to payers based on the risk of their enrollee population, to offset the cost of some payers' providing health insurance to relatively high-risk individuals.

Study Design.  We propose a methodology based on worst-case optimization over an appropriate uncertainty set.

Data Sources.  We use Federal Register data to create a realistic case study.

Principal Findings.  Our approach to compute robust risk scores involves solving a series of linear problems and thus can be done using standard analytical tools such as Excel Solver.

Conclusions.  There is an important need for the "robustification" of risk scores so that health payers can be properly compensated for the risk they take in providing insurance to relatively high-risk individuals.  Our model presents one possible way to do so in a tractable manner.