History-Aware Conformal Prediction Sets for Censored Time-to-Event Outcomes

2026-05-07

AI Summary

This work addresses a key limitation of existing conformal prediction methods, which rely solely on baseline covariates and thus struggle to support clinical decisions informed by individual dynamic histories. To overcome this, the authors propose History-Aware Prediction Sets (HAPS), the first framework integrating longitudinal covariate history into conformal prediction for time-to-event outcomes. HAPS constructs prediction sets conditioned on individuals having survived up to the decision time and employs inverse probability censoring weighting to handle right-censoring and time-dependent confounding, thereby guaranteeing conditional coverage. Two doubly robust extensions are further introduced to reduce reliance on correct specification of the censoring model. Empirical results demonstrate that HAPS reduces median prediction interval length by up to 75% in simulations and by up to 60% on two public datasets for five-year predictions, while maintaining coverage close to the nominal level.

Abstract

Existing conformal prediction methods for time-to-event outcomes leverage only baseline covariates, producing prediction intervals that are insufficiently informative to facilitate decision making. We propose History-Aware Prediction Sets (HAPS), a conformal framework that constructs prediction sets for individual event times using covariate histories observed up to a decision time, targeting coverage among individuals who have survived to this time. HAPS handles right censoring adjusted for time-varying confounders via inverse probability of censoring weighting. When the censoring weights are consistently estimated, it achieves PAAC (probably asymptotically approximately correct) coverage among survivors. We further propose two doubly robust extensions of HAPS to weaken reliance on consistent estimation of the censoring distribution. In simulations, HAPS and its extensions reduce median prediction interval length by up to 75\% relative to baseline comparators while maintaining close to nominal coverage. On two public benchmark data sets, HAPS reduces the median interval length by up to 60\% for predictions at year 5, compared to the baseline comparators.

Problem

• conformal prediction
• time-to-event outcomes
• right censoring
• time-varying covariates
• prediction sets

Innovation

• History-Aware Prediction Sets
• conformal prediction
• time-to-event
• right censoring
• doubly robust