Bayesian index models for heterogeneous treatment effects

TL;DR

This paper introduces a Bayesian single-index model to estimate heterogeneous treatment effects, enabling the creation of a treatment benefit index for patient stratification in precision health, demonstrated through a COVID-19 study.

Contribution

It develops a flexible Bayesian model with a data-driven link function for estimating treatment effects and constructing a treatment benefit index for personalized treatment decisions.

Findings

Effective stratification of patients based on predicted treatment benefit.

Application to COVID-19 treatment data demonstrates practical utility.

Model captures heterogeneity in treatment responses.

Abstract

The general idea of this article is to develop a Bayesian model with a flexible link function connecting an exponential family treatment response to a linear combination of covariates and a treatment indicator and the interaction between the two. Generalized linear models allowing data-driven link functions are often called "single-index models," and among popular semi-parametric modeling methods. In this article, we will focus on modeling heterogeneous treatment effects, with the goal of developing a treatment benefit index (TBI) incorporating prior information from historical data. This treatment benefit index can be useful for stratifying patients according to their predicted treatment benefit levels and can be especially useful for precision health applications. The proposed method is applied to a COVID-19 treatment study.