Patients-like-me: A Variational LM--GNN Framework for Explainable Clinical Prediction

2026/08/04 by Xinyu Wang, Yixuan Li, Hanwei Wu +4
Computer Science · #cs.CL #cs.AI

paper · pdf

arxiv created 2026/08/04 · arxiv updated 2026/08/06

Abstract

Language models (LMs) offer strong textual representations for electronic health records (EHRs), but they encode patient sequences in isolation and provide limited explainability. Graph neural networks (GNNs) complement LMs by incorporating inter-patient relationships and enabling reference-patient attribution, yet they rely on high-quality patient representations. We propose Patients-like-me (PLM), a unified LM--GNN framework that integrates local patient semantics with global cohort structure. To train PLM efficiently, we introduce a Variational Expectation-Maximization algorithm that alternates LM and GNN updates under a supervised variational objective. Extensive experiments on MIMIC-III and MIMIC-IV show that PLM consistently outperforms state-of-the-art methods, with improvements generalizing across encoder-only and decoder-only LM backbones. These gains are achieved with only modest additional computational overhead. PLM also provides reference-patient explanations by retrieving influential similar patients, while edge-masking experiments confirm that the highest-ranked references have the greatest impact on model predictions.

Citations

• BioClinical ModernBERT: A State-of-the-Art Long-Context Encoder for Biomedical and Clinical NLP
• Large Language Models are Powerful Electronic Health Record Encoders
• Unveiling Discrete Clues: Superior Healthcare Predictions for Rare Diseases
• Reasoning-Enhanced Healthcare Predictions with Knowledge Graph Community Retrieval
• LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders
• Small Language Models Learn Enhanced Reasoning Skills from Medical Textbooks
• Leveraging graph neural networks for supporting Automatic Triage of Patients
• Can GNN be Good Adapter for LLMs?
• Large Language Model Meets Graph Neural Network in Knowledge Distillation
• Large Language Model Distilling Medication Recommendation Model
• A Survey of Graph Meets Large Language Model: Progress and Future Directions
• GraphCare: Enhancing Healthcare Predictions with Personalized Knowledge Graphs
• SafeDrug: Dual Molecular Graph Encoders for Recommending Effective and Safe Drug Combinations
• Change Matters: Medication Change Prediction with Recurrent Residual Networks
• On Explainability of Graph Neural Networks via Subgraph Explorations
• Predicting Patient Outcomes with Graph Representation Learning
• Parameterized Explainer for Graph Neural Network
• Med-BERT: pre-trained contextualized embeddings on large-scale structured electronic health records for disease prediction
• Learning Hierarchical Review Graph Representations for Recommendation
• StageNet: Stage-Aware Neural Networks for Health Risk Prediction
• AdaCare: Explainable Clinical Health Status Representation Learning via Scale-Adaptive Feature Extraction and Recalibration
• Pre-training of Graph Augmented Transformers for Medication Recommendation
• ClinicalBERT: Modeling Clinical Notes and Predicting Hospital Readmission
• GNNExplainer: Generating Explanations for Graph Neural Networks
• BioBERT: a pre-trained biomedical language representation model for biomedical text mining
• GAMENet: Graph Augmented MEmory Networks for Recommending Medication Combination
• Neural Message Passing for Quantum Chemistry
• GRAM: Graph-based Attention Model for Healthcare Representation Learning
• RETAIN: An Interpretable Predictive Model for Healthcare using Reverse Time Attention Mechanism
• Deepr: A Convolutional Net for Medical Records
• DeepCare: A Deep Dynamic Memory Model for Predictive Medicine
• Variational Inference: A Review for Statisticians
• Doctor AI: Predicting Clinical Events via Recurrent Neural Networks
• Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
• A tutorial on variational Bayesian inference
• Statistical Analysis of Non-Lattice Data