SPOT-IC: Improving prediction for interval-censored data via survival probability transfer

2026-07-10

AI Summary

This study addresses the challenge of accurately predicting survival outcomes for chronic diseases under wide-interval censoring and limited follow-up, where existing methods often falter. The authors propose a novel transfer learning approach tailored for interval-censored data that enhances target-model performance by leveraging survival probability information from multiple source studies—without requiring access to individual-level source data or imposing constraints on source model types. The method innovatively incorporates a penalty term based on survival probability transfer and enables adaptive multi-source aggregation to mitigate negative transfer. Theoretically, it achieves a faster convergence rate than models using only target data. An efficient EM algorithm, combined with a data-adaptive weighting scheme, integrates multi-source information. Simulations and empirical analysis using data from the Alzheimer’s Disease Neuroimaging Initiative demonstrate substantial gains in prediction accuracy, particularly when informative source studies are available.

Abstract

Accurate prediction with interval-censored data is particularly challenging when censoring intervals are wide and follow-up is limited, as is common in studies of chronic diseases. Although auxiliary information from source studies may improve prediction in a target study, existing transfer learning methods typically impose restrictive assumptions on model or parameter similarity, or require access to individual-level source data. We propose a novel transfer learning method for interval-censored data that allows arbitrary source models and avoids sharing source data. Our approach transfers survival probability information from source studies through a carefully designed penalty and enables efficient computation via a simple EM algorithm. When multiple source studies are available and their informativeness is unknown, we further develop a data-adaptive aggregation procedure that is robust to negative transfer. Theoretical analysis shows that the proposed estimator attains a faster convergence rate than the target-only estimator whenever at least one source study is sufficiently informative. Extensive simulation studies and an application to data from the Alzheimer's Disease Neuroimaging Initiative demonstrate the effectiveness of our approach.

Problem

• interval-censored data
• survival prediction
• transfer learning
• chronic diseases
• censoring intervals

Innovation

• transfer learning
• interval-censored data
• survival probability
• negative transfer
• EM algorithm