CRISP-NAM: Competing Risks Interpretable Survival Prediction with Neural Additive Models

2025-05-27

AI Summary

Addressing the challenge of modeling competing risks in medical multi-endpoint survival analysis, this paper introduces the Competing-Risk Neural Additive Model (CR-NAM), the first extension of Neural Additive Models (NAMs) to the cause-specific hazard framework. CR-NAM learns independent, nonlinear, and differentiably monotonic effect functions for each covariate under monotonicity constraints, enabling risk decoupling across distinct competing event types and yielding interpretable univariate effect plots. Leveraging a customized survival loss function and end-to-end optimization, CR-NAM achieves state-of-the-art predictive performance across multiple clinical datasets. Crucially, it simultaneously delivers high prediction accuracy and fine-grained, feature-level interpretability—providing clinicians with verifiable, attribution-based risk assessments to support evidence-driven decision-making.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Relational Probabilistic Models
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Competing risks are crucial considerations in survival modelling, particularly in healthcare domains where patients may experience multiple distinct event types. We propose CRISP-NAM (Competing Risks Interpretable Survival Prediction with Neural Additive Models), an interpretable neural additive model for competing risks survival analysis which extends the neural additive architecture to model cause-specific hazards while preserving feature-level interpretability. Each feature contributes independently to risk estimation through dedicated neural networks, allowing for visualization of complex non-linear relationships between covariates and each competing risk. We demonstrate competitive performance on multiple datasets compared to existing approaches.

Problem

• Interpretable survival prediction for competing risks
• Modeling cause-specific hazards with neural networks
• Visualizing non-linear covariate-risk relationships

Innovation

• Extends neural additive models for competing risks
• Preserves feature-level interpretability in predictions
• Visualizes non-linear covariate-risk relationships