Interpretable Fine-Gray Deep Survival Model for Competing Risks: Predicting Post-Discharge Foot Complications for Diabetic Patients in Ontario

2025-11-15

AI Summary

To address the limited interpretability of deep learning models for post-discharge foot complication prediction in diabetic patients under competing risks, this paper proposes CRISPNAM-FG—a novel intrinsically interpretable competing-risks survival model. It integrates the Fine-Gray cumulative incidence function with a neural additive modeling architecture, learning distinct projection vectors for each competing risk to yield human-readable shape functions (feature effect curves) and quantified feature importances—without sacrificing predictive accuracy. Evaluated on multiple benchmark datasets and real-world clinical data from 29 hospitals in Ontario (2016–2023), CRISPNAM-FG achieves performance on par with state-of-the-art deep survival models. Its transparent structure enables clinical decision trust, fine-grained attribution analysis, and regulatory compliance auditing, thereby advancing the real-world deployment of interpretable AI for competing-risks prediction.

Technology Category

• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Interpretability, Explainability, and Transparency

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Model interpretability is crucial for establishing AI safety and clinician trust in medical applications for example, in survival modelling with competing risks. Recent deep learning models have attained very good predictive performance but their limited transparency, being black-box models, hinders their integration into clinical practice. To address this gap, we propose an intrinsically interpretable survival model called CRISPNAM-FG. Leveraging the structure of Neural Additive Models (NAMs) with separate projection vectors for each risk, our approach predicts the Cumulative Incidence Function using the Fine-Gray formulation, achieving high predictive power with intrinsically transparent and auditable predictions. We validated the model on several benchmark datasets and applied our model to predict future foot complications in diabetic patients across 29 Ontario hospitals (2016-2023). Our method achieves competitive performance compared to other deep survival models while providing transparency through shape functions and feature importance plots.

Problem

• Developing interpretable deep survival models for competing risks analysis
• Predicting post-discharge foot complications in diabetic patients
• Addressing black-box limitations in clinical AI implementation

Innovation

• Neural Additive Models structure for separate risk projections
• Fine-Gray formulation for Cumulative Incidence Function prediction
• Shape functions and feature importance plots for transparency