ADHAM: Additive Deep Hazard Analysis Mixtures for Interpretable Survival Regression

2025-09-08

AI Summary

Survival analysis in clinical decision-making demands both high predictive accuracy and model interpretability—yet most deep learning approaches sacrifice the latter. To address this, we propose the Additive Deep Risk Mixture (ADRM) model, which enables three-level interpretable survival modeling—population-, subgroup-, and individual-level—via conditional latent subgroup structure and covariate-specific additive risk functions. A neural network learns latent subgroup assignments, while post-hoc clustering enhances subgroup interpretability; equivalent subgroups are subsequently merged after training to improve robustness. Evaluated on multiple real-world medical datasets, ADRM achieves state-of-the-art predictive performance (measured by C-index and Integrated Brier Score) while uncovering heterogeneous effects of exposures across subgroups. This yields clinically actionable, interpretable risk attribution and supports stratified intervention strategies.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Graphical Models
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Survival analysis is a fundamental tool for modeling time-to-event outcomes in healthcare. Recent advances have introduced flexible neural network approaches for improved predictive performance. However, most of these models do not provide interpretable insights into the association between exposures and the modeled outcomes, a critical requirement for decision-making in clinical practice. To address this limitation, we propose Additive Deep Hazard Analysis Mixtures (ADHAM), an interpretable additive survival model. ADHAM assumes a conditional latent structure that defines subgroups, each characterized by a combination of covariate-specific hazard functions. To select the number of subgroups, we introduce a post-training refinement that reduces the number of equivalent latent subgroups by merging similar groups. We perform comprehensive studies to demonstrate ADHAM's interpretability at the population, subgroup, and individual levels. Extensive experiments on real-world datasets show that ADHAM provides novel insights into the association between exposures and outcomes. Further, ADHAM remains on par with existing state-of-the-art survival baselines in terms of predictive performance, offering a scalable and interpretable approach to time-to-event prediction in healthcare.

Problem

• Develops interpretable survival model for healthcare outcomes
• Addresses lack of interpretability in neural network survival models
• Identifies covariate-specific hazard patterns through subgroup analysis

Innovation

• Interpretable additive survival model with latent subgroups
• Post-training refinement merges similar latent subgroups
• Maintains predictive performance with scalable interpretable approach