Interpretable phenotyping of Heart Failure patients with Dutch discharge letters

2025-05-30

AI Summary

Accurate left ventricular ejection fraction (LVEF) phenotyping in heart failure (HF) patients remains clinically challenging due to heterogeneous data sources and the need for both high predictive performance and clinical interpretability. Method: We propose an intrinsically interpretable “white-box” modeling framework—Aug-Linear—that integrates structured clinical data (ICD codes, echocardiography reports) with unstructured Dutch discharge summaries. Leveraging multi-source silver-standard labels and domain-informed feature engineering, we systematically evaluate data modalities and identify discharge summaries as the most discriminative information source for LVEF classification. Contribution/Results: The Aug-Linear model achieves AUC = 0.81 on external validation across two independent centers. Crucially, its model-generated explanations exhibit significantly higher consistency with clinician annotations than black-box models (e.g., BERT, AUC = 0.84) and post-hoc attribution methods (e.g., SHAP, LIME). This work delivers a deployable, trustworthy AI tool for transparent HF phenotyping, jointly optimizing intrinsic interpretability and predictive accuracy.

Technology Category

• Machine Learning: Transparent, Interpretable, Explainable ML
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Computer Vision: Interpretability, Explainability, and Transparency

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Objective: Heart failure (HF) patients present with diverse phenotypes affecting treatment and prognosis. This study evaluates models for phenotyping HF patients based on left ventricular ejection fraction (LVEF) classes, using structured and unstructured data, assessing performance and interpretability. Materials and Methods: The study analyzes all HF hospitalizations at both Amsterdam UMC hospitals (AMC and VUmc) from 2015 to 2023 (33,105 hospitalizations, 16,334 patients). Data from AMC were used for model training, and from VUmc for external validation. The dataset was unlabelled and included tabular clinical measurements and discharge letters. Silver labels for LVEF classes were generated by combining diagnosis codes, echocardiography results, and textual mentions. Gold labels were manually annotated for 300 patients for testing. Multiple Transformer-based (black-box) and Aug-Linear (white-box) models were trained and compared with baselines on structured and unstructured data. To evaluate interpretability, two clinicians annotated 20 discharge letters by highlighting information they considered relevant for LVEF classification. These were compared to SHAP and LIME explanations from black-box models and the inherent explanations of Aug-Linear models. Results: BERT-based and Aug-Linear models, using discharge letters alone, achieved the highest classification results (AUC=0.84 for BERT, 0.81 for Aug-Linear on external validation), outperforming baselines. Aug-Linear explanations aligned more closely with clinicians' explanations than post-hoc explanations on black-box models. Conclusions: Discharge letters emerged as the most informative source for phenotyping HF patients. Aug-Linear models matched black-box performance while providing clinician-aligned interpretability, supporting their use in transparent clinical decision-making.

Problem

• Classify Heart Failure patients using LVEF from discharge letters
• Compare interpretability of black-box and white-box phenotyping models
• Evaluate model performance on structured and unstructured clinical data

Innovation

• Uses Transformer-based and Aug-Linear models
• Combines structured and unstructured data
• Provides clinician-aligned interpretability