A data-driven approach to discover and quantify systemic lupus erythematosus etiological heterogeneity from electronic health records

2025-01-13

AI Summary

Systemic lupus erythematosus (SLE) exhibits high etiological heterogeneity, and its diagnosis is hindered by incomplete, noisy electronic health records (EHRs). Method: We propose an unsupervised framework integrating independent component analysis (ICA) with causal representation learning—first modeling multimodal clinical observations in EHRs as latent exogenous factors in a causal graph to interpretable decompose SLE’s etiological heterogeneity. Contribution/Results: Applied to real-world EHRs, our method automatically discovers 19 probabilistically independent, clinically interpretable latent etiologic sources, serving as patient-level explainable representations. Lightweight supervised validation and multi-dimensional clinical evaluation confirm that these sources significantly improve SLE discrimination (AUC ↑8.2%) and support clinicians in performing etiology-guided decision trade-offs for complex cases.

Technology Category

• Machine Learning: Causal Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Reasoning under Uncertainty: Causality

Application Category

• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Systemic lupus erythematosus (SLE) is a complex heterogeneous disease with many manifestational facets. We propose a data-driven approach to discover probabilistic independent sources from multimodal imperfect EHR data. These sources represent exogenous variables in the data generation process causal graph that estimate latent root causes of the presence of SLE in the health record. We objectively evaluated the sources against the original variables from which they were discovered by training supervised models to discriminate SLE from negative health records using a reduced set of labelled instances. We found 19 predictive sources with high clinical validity and whose EHR signatures define independent factors of SLE heterogeneity. Using the sources as input patient data representation enables models to provide with rich explanations that better capture the clinical reasons why a particular record is (not) an SLE case. Providers may be willing to trade patient-level interpretability for discrimination especially in challenging cases.

Problem

• Systemic Lupus Erythematosus
• Causal Inference
• Medical Record Analysis

Innovation

• Imperfect Medical Records
• Systemic Lupus Erythematosus (SLE) Prediction
• Minimal Labeled Cases