Estimation of Treatment Harm Rate via Partitioning

TL;DR

This paper introduces a novel partitioning-based method to estimate treatment harm rate (THR) in randomized controlled trials, effectively utilizing auxiliary covariates and providing consistent bounds without untestable assumptions.

Contribution

It develops a new class of bounds for THR that incorporate covariate information and are estimable without untestable assumptions, advancing causal inference methods.

Findings

Proposed bounds effectively use auxiliary covariates.

Bounds can be consistently estimated without untestable assumptions.

Method performs well in simulation studies and real data application.

Abstract

In causal inference with binary outcomes, there is a growing interest in estimation of treatment harm rate (THR), which is a measure of treatment risk and reveals treatment effect heterogeneity in a subpopulation. The THR is generally non-identifiable even for randomized controlled trials (RCTs), and existing works focus primarily on the estimation of the THR under either untestable identification or ambiguous model assumptions. We develop a class of partitioning-based bounds for the THR based on data from RCTs with two distinct features: Our proposed bounds effectively use available auxiliary covariates information and the bounds can be consistently estimated without relying on any untestable or ambiguous model assumptions. Finite sample performances of our proposed interval estimators along with a conservatively extended confidence interval for the THR are evaluated through Monte Carlo simulation studies. An application of the proposed methods to the ACTG 175 data is presented. A Python package named partbte for the partitioning-based algorithm has been developed and is available on https://github.com/w62liang/partition-te.