Using Fuzzy Diagnoses for Risk Adjustment

Risk adjusted capitation payments attempt to pay health plans fairly for their enrollees and eliminate the incentive for health plans to attract only profitable enrollees. Diagnoses are widely used for risk adjustment, and recent theoretical papers have derived Optimal Risk Adjustment formulas that focus on plan-level rather than individual level prediction. This paper examines the selection incentives of imperfect risk adjustment when plans differ in how diagnoses are reported, so that diagnoses are uneven signals of health status. A new optimal risk adjustment formula is developed that corrects for “fuzzy” diagnoses, and shown to be similar to the formula derived by Glazer and McGuire in being relatively robust to large amounts of signal noise. Dedication: I would like to dedicate this paper to my colleague and friend Robert W Rosenthal, who died unexpectedly and will be greatly missed. I would like to thank Ana Lorena Prieto and seminar participants at Hong Kong University of Science and Technology for helpful comments on the paper. Research support for this paper is provided by the Agency for Healthcare Research and Quality grant number R01-HS10620-01A.