Adaptive Experimentation for Censored Survival Outcomes

2026-05-18 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

A new framework for adaptive experimentation, called the Adaptive Survival Estimator (ASE), has been developed to efficiently estimate causal effects from survival data that includes right censoring. This framework addresses a gap in existing adaptive experimentation methods, which typically do not account for partially observed event times due to censoring (e.g., patient dropout in clinical trials). The ASE framework derives a semiparametric efficiency bound for the average survival effect curve, leading to a closed-form, efficiency-optimal allocation policy that extends classical Neyman allocation. This policy prioritizes patient strata with high uncertainty from both event and censoring dynamics. ASE accommodates arbitrary machine learning models for nuisance estimation, is guided by this optimal allocation policy, and offers strong theoretical guarantees, including asymptotic normality. Numerical experiments demonstrate consistent efficiency gains compared to uniform randomization and baselines that ignore censoring.

Key takeaway

For research scientists designing clinical trials or other experiments with censored survival outcomes, the Adaptive Survival Estimator (ASE) offers a robust method to improve efficiency. You should consider implementing ASE to optimize treatment allocation and obtain more precise causal effect estimates, especially when dealing with patient dropout or other forms of right censoring, thereby potentially reducing trial duration or sample size requirements.

Key insights

A new adaptive experimentation framework efficiently estimates causal effects from censored survival data using an optimal allocation policy.

Principles

• Prioritize strata with high event and censoring uncertainty.
• Optimal allocation generalizes Neyman allocation for survival data.

Method

The Adaptive Survival Estimator (ASE) sequentially learns an allocation policy and estimates the average survival effect curve, accommodating various ML models for nuisance estimation.

In practice

• Apply ASE for causal effect estimation with censored data.
• Use ASE in clinical trials with patient dropout.

Topics

• Adaptive Experimentation
• Censored Survival Outcomes
• Causal Effect Estimation
• Semiparametric Efficiency Bound
• Optimal Allocation Policy

Best for: AI Scientist, Research Scientist, Data Scientist

Related on AIssential

• Learning Survival Models with Right-Censored Reporting Delays — Jointly modeling event occurrence and report timing addresses right-censored reporting delays for timely risk evaluation.
• Robust Sequential Experimental Design for A/B Testing — Robust sequential experimental design improves A/B testing efficiency by mitigating model misspecification and finite-sample imbalances.
• Code Lifespan Survival Analysis (CLSA): Predicting the Survival of Source Code Lines Using AST-Aware Mining — Predicting individual source code line deletion risk is feasible using static features and survival analysis, revealing time-varying effects.
• A Decision-Theoretic View of Test-Time Training: When, How Far, and Which Directions to Adapt — Test-time training's instability is explained and guided by a decision-theoretic, Bayesian inference framework for adaptive updates.
• Synthetic Designed Experiments for Diagnosing Vision Model Failure — SDRS uses Design of Experiments and ANOVA to diagnose vision model failures and prescribe targeted synthetic data.
• XMSE-Aware Adaptive Empirical Bayes Estimation — An XMSE-aware mixed estimator adaptively balances ML and EB shrinkage to improve performance under kernel misspecification.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.