Enabling Counterfactual Survival Analysis with Balanced Representations

TL;DR

This paper introduces a unified, theoretically grounded framework for counterfactual survival analysis that effectively handles censored data, improving treatment effect estimation and survival prediction in medical and industrial applications.

Contribution

It presents a novel framework for counterfactual inference with survival data, incorporating a nonparametric hazard ratio metric and demonstrating superior performance over existing methods.

Findings

Outperforms existing methods in survival prediction

Accurately estimates treatment effects in censored data

Introduces a semi-synthetic dataset for evaluation

Abstract

Balanced representation learning methods have been applied successfully to counterfactual inference from observational data. However, approaches that account for survival outcomes are relatively limited. Survival data are frequently encountered across diverse medical applications, i.e., drug development, risk profiling, and clinical trials, and such data are also relevant in fields like manufacturing (e.g., for equipment monitoring). When the outcome of interest is a time-to-event, special precautions for handling censored events need to be taken, as ignoring censored outcomes may lead to biased estimates. We propose a theoretically grounded unified framework for counterfactual inference applicable to survival outcomes. Further, we formulate a nonparametric hazard ratio metric for evaluating average and individualized treatment effects. Experimental results on real-world and semi-synthetic datasets, the latter of which we introduce, demonstrate that the proposed approach significantly outperforms competitive alternatives in both survival-outcome prediction and treatment-effect estimation.