Statistical methods for clustered competing risk data when the event types are only available in a training dataset

2025-05-04

AI Summary

This paper addresses clustered competing-risks data where event-type labels are missing in the main study and available only in a labeled training dataset. Method: We propose a cause-specific proportional hazards model with random effects, uniquely embedding predicted probabilities from a supervised classification model into the competing-risks framework. Event-type probability-weighted imputation and weighted penalized partial likelihood estimation enable testable, bias-controlled transfer of supervision from the training set to the main study. Analytic variance is rigorously derived to ensure statistical compatibility under cluster dependence. Results: Monte Carlo simulations confirm unbiased estimation, accurate standard error estimation, and nominal coverage of confidence intervals. Applied to hearing protection research, the method successfully uncovers distinct associations between tinnitus and metabolic, sensory, and mixed-type hearing loss—demonstrating its practical utility in etiologic differentiation.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Security and Privacy: Data transparency and provenance

Abstract

We develop methods to analyze clustered competing risks data when the event types are only available in a training dataset and are missing in the main study. We propose to estimate the exposure effects through the cause-specific proportional hazards frailty model where random effects are introduced into the model to account for the within-cluster correlation. We propose a weighted penalized partial likelihood method where the weights represent the probabilities of the occurrence of events, and the weights can be obtained by fitting a classification model for the event types on the training dataset. Alternatively, we propose an imputation approach where the missing event types are imputed based on the predictions from the classification model. We derive the analytical variances, and evaluate the finite sample properties of our methods in an extensive simulation study. As an illustrative example, we apply our methods to estimate the associations between tinnitus and metabolic, sensory and metabolic+sensory hearing loss in the Conservation of Hearing Study Audiology Assessment Arm.

Problem

• Analyze clustered competing risks data with missing event types
• Estimate exposure effects using cause-specific proportional hazards
• Develop weighted and imputation methods for event type classification

Innovation

• Cause-specific proportional hazards frailty model
• Weighted penalized partial likelihood method
• Imputation approach for missing event types