JMIR Publications

ABSTRACT

Background:

Opioid Use Disorder (OUD) is a critical public health crisis in the US, affecting over 5.5 million Americans in 2021. Machine learning (ML) has been used to predict patient risk of incident OUD. However, little is known about the fairness and bias of these predictive models.

Objective:

The aims of this study are two-fold: (1) to develop an ML bias mitigation algorithm for sociodemographic features, and (2) to develop a fairness-aware weighted majority voting (WMV) classifier for OUD prediction

Methods:

We used the 2020 national survey on drug and health (NSDUH) data to develop a neural network (NN) model using SGD (NN-SGD) and Adam (NN-Adam) optimizers and evaluated sociodemographic bias by comparing area under curve (AUC) values. A bias mitigation algorithm, based on equality of odds (EO), was implemented to minimize disparities in specificity and recall. Finally, a WMV classifier was developed for fairness-aware prediction of OUD.

Results:

Our bias mitigation algorithm significantly reduced bias in NN-SGD by 21.66% for gender, 1.48% for race, and 21.04% for income, and in NN-Adam by 16.96% for gender, 8.87% for marital status, 8.45% for working condition, and 41.62% for race. The fairness-aware WMV classifier achieved a recall of 85.37% and 92.68%, and an accuracy of 58.85% and 90.21% using NN-SGD and NN-Adam, respectively.

Conclusions:

The application of the proposed bias mitigation algorithm shows promise in reducing sociodemographic bias, with the WMV classifier confirming bias reduction and high performance in OUD prediction.