Patient willingness to trade predictive accuracy for equity and the effects of removing race from clinical risk prediction

Ascertain, via empirical studies of patient populations, the extent to which patients are willing to accept reductions in the accuracy of medical risk predictions to mitigate racially defined disparities, and quantify how removing race as a predictor in clinical risk assessment changes the magnitudes of specified disparity measures across racial groups.

Background

The paper contrasts a utilitarian clinical decision-making framework—which recommends using all predictive covariates, including race, to maximize expected patient welfare—with a growing movement in medicine that advocates removing race from risk prediction tools on non-utilitarian equity grounds.

Within this debate, the author notes a lack of empirical evidence about two key issues: whether patients themselves would accept less accurate predictions to mitigate racial disparities, and how the omission of race from prediction models actually changes different disparity metrics. The author highlights these as pressing empirical gaps that bear directly on policy design and evaluation.