Penalized Fair Regression for Multiple Groups in Chronic Kidney Disease

TL;DR

This paper introduces a penalized fair regression framework for multiple groups, specifically applied to health care data on chronic kidney disease, improving fairness without sacrificing accuracy.

Contribution

It proposes a novel regression method with unfairness penalties for multiple groups, including automatic penalty weight selection, and demonstrates its effectiveness in health care data.

Findings

Achieves a better fairness-accuracy trade-off than existing methods.

Improves fairness for multiple race and ethnicity groups in health care.

Maintains overall model fit while reducing bias.

Abstract

Fair regression methods have the potential to mitigate societal bias concerns in health care, but there has been little work on penalized fair regression when multiple groups experience such bias. We propose a general regression framework that addresses this gap with unfairness penalties for multiple groups. Our approach is demonstrated for binary outcomes with true positive rate disparity penalties. It can be efficiently implemented through reduction to a cost-sensitive classification problem. We additionally introduce novel score functions for automatically selecting penalty weights. Our penalized fair regression methods are empirically studied in simulations, where they achieve a fairness-accuracy frontier beyond that of existing comparison methods. Finally, we apply these methods to a national multi-site primary care study of chronic kidney disease to develop a fair classifier for end-stage renal disease. There we find substantial improvements in fairness for multiple race and ethnicity groups who experience societal bias in the health care system without any appreciable loss in overall fit.