Orthogonal Survival Learners for Estimating Heterogeneous Treatment Effects from Time-to-Event Data

2025-05-19

AI Summary

This paper addresses heterogeneous treatment effect (HTE) estimation under censored survival data. Methodologically, it introduces the first orthogonal survival learner toolbox—a model-agnostic orthogonalization framework that incorporates customizable weighting functions to simultaneously mitigate low-coverage bias arising from both survival and censoring overlap deficiencies. It formulates survival analogues of the DR-learner and R-learner, along with a novel robust variant. Theoretical analysis establishes formal orthogonality guarantees, ensuring validity in both randomized and observational study settings. Empirical evaluation demonstrates substantial improvements in HTE estimation accuracy and robustness across diverse low-overlap survival scenarios, outperforming existing approaches. By enabling reliable causal inference under complex censoring mechanisms, the proposed framework advances personalized, evidence-based decision-making in clinical and epidemiological applications.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Estimating heterogeneous treatment effects (HTEs) is crucial for personalized decision-making. However, this task is challenging in survival analysis, which includes time-to-event data with censored outcomes (e.g., due to study dropout). In this paper, we propose a toolbox of novel orthogonal survival learners to estimate HTEs from time-to-event data under censoring. Our learners have three main advantages: (i) we show that learners from our toolbox are guaranteed to be orthogonal and thus come with favorable theoretical properties; (ii) our toolbox allows for incorporating a custom weighting function, which can lead to robustness against different types of low overlap, and (iii) our learners are model-agnostic (i.e., they can be combined with arbitrary machine learning models). We instantiate the learners from our toolbox using several weighting functions and, as a result, propose various neural orthogonal survival learners. Some of these coincide with existing survival learners (including survival versions of the DR- and R-learner), while others are novel and further robust w.r.t. low overlap regimes specific to the survival setting (i.e., survival overlap and censoring overlap). We then empirically verify the effectiveness of our learners for HTE estimation in different low-overlap regimes through numerical experiments. In sum, we provide practitioners with a large toolbox of learners that can be used for randomized and observational studies with censored time-to-event data.

Problem

• Estimating heterogeneous treatment effects in survival analysis with censored data
• Developing orthogonal survival learners with robust theoretical properties
• Providing model-agnostic tools for personalized decision-making in low-overlap regimes

Innovation

• Orthogonal survival learners for HTE estimation
• Custom weighting functions enhance robustness
• Model-agnostic learners with neural implementations