Think as a Doctor: An Interpretable AI Approach for ICU Mortality Prediction

2025-10-11

AI Summary

Existing ICU mortality prediction models suffer from limited interpretability and fail to jointly address three critical clinical requirements: dynamic clinical course identification, demographic heterogeneity modeling, and prognostic awareness—particularly lacking integrated modeling of the latter two. This paper proposes ProtoDoctor, the first framework to inherently embed all three elements. It introduces a prognostic clinical course identification module to capture temporal disease progression; incorporates cohort-specific prototype learning with risk adjustment to model population heterogeneity; and unifies prototype-based reasoning with temporal EHR representation for intrinsic interpretability. A novel regularization method further enhances generalization. Evaluated on multicenter datasets, ProtoDoctor significantly outperforms state-of-the-art models in predictive accuracy. Clinical expert evaluation confirms its explanations align closely with real-world decision-making pathways, demonstrating high credibility and clinical utility.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Reasoning under Uncertainty: Relational Probabilistic 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

Intensive Care Unit (ICU) mortality prediction, which estimates a patient's mortality status at discharge using EHRs collected early in an ICU admission, is vital in critical care. For this task, predictive accuracy alone is insufficient; interpretability is equally essential for building clinical trust and meeting regulatory standards, a topic that has attracted significant attention in information system research. Accordingly, an ideal solution should enable intrinsic interpretability and align its reasoning with three key elements of the ICU decision-making practices: clinical course identification, demographic heterogeneity, and prognostication awareness. However, conventional approaches largely focus on demographic heterogeneity, overlooking clinical course identification and prognostication awareness. Recent prototype learning methods address clinical course identification, yet the integration of the other elements into such frameworks remains underexplored. To address these gaps, we propose ProtoDoctor, a novel ICU mortality prediction framework that delivers intrinsic interpretability while integrating all three elements of the ICU decision-making practices into its reasoning process. Methodologically, ProtoDoctor features two key innovations: the Prognostic Clinical Course Identification module and the Demographic Heterogeneity Recognition module. The former enables the identification of clinical courses via prototype learning and achieves prognostication awareness using a novel regularization mechanism. The latter models demographic heterogeneity through cohort-specific prototypes and risk adjustments. Extensive empirical evaluations demonstrate that ProtoDoctor outperforms state-of-the-art baselines in predictive accuracy. Human evaluations further confirm that its interpretations are more clinically meaningful, trustworthy, and applicable in ICU practice.

Problem

• Predicts ICU mortality using early admission EHR data
• Enhances interpretability for clinical trust and regulatory compliance
• Integrates clinical course identification and demographic heterogeneity modeling

Innovation

• Prototype learning identifies clinical courses for interpretability
• Regularization mechanism achieves prognostication awareness in predictions
• Cohort-specific prototypes model demographic heterogeneity and risks