Patients-like-me: A Variational LM--GNN Framework for Explainable Clinical Prediction
Quick Answer
This paper shows that The Patients-like-me (PLM) framework combines language models and graph neural networks to enhance clinical predictions by integrating patient semantics and cohort structures.
Quick Take
Experiments on MIMIC-III and MIMIC-IV demonstrate that PLM outperforms existing methods with minimal computational overhead, while also providing explainability through reference-patient retrieval.
Key Points
- PLM integrates local patient semantics with global cohort structure for better predictions.
- Utilizes a Variational Expectation-Maximization algorithm for efficient training.
- Outperforms state-of-the-art methods on MIMIC-III and MIMIC-IV datasets.
- Provides explainability by retrieving influential similar patients.
- Achieves performance gains with modest additional computational overhead.
DeepSignal Analysis
What happened
The Patients-like-me (PLM) framework integrates language models and graph neural networks to improve clinical predictions by leveraging patient semantics and cohort structures. It was tested on the MIMIC-III and MIMIC-IV datasets, showing superior performance compared to existing methods with minimal computational overhead.
Key evidence
- PLM combines language models and graph neural networks, enhancing clinical predictions by integrating local patient semantics with global cohort structure.
- Experiments on MIMIC-III and MIMIC-IV demonstrate that PLM consistently outperforms state-of-the-art methods, indicating its effectiveness in clinical settings.
- The framework provides explainability through reference-patient retrieval, allowing for influential similar patients to be identified and linked to model predictions.
Why it matters
The integration of language models and graph neural networks in PLM addresses the limitations of existing methods in clinical prediction, particularly in terms of explainability and patient relationship modeling. This advancement could lead to more accurate and interpretable clinical decision-making tools, ultimately improving patient outcomes.
Paper Resources
📖 Reader Mode
~2 min readAbstract: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.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.04193 [cs.CL] |
| (or arXiv:2608.04193v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2608.04193 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Xinyu Wang [view email]
[v1]
Tue, 4 Aug 2026 19:51:52 UTC (10,087 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →TriAgent: Divergence-Aware Committees for Cost-Efficient Financial Sentiment Analysis
TriAgent introduces a cost-efficient multi-agent system for financial sentiment analysis, combining VADER, FinBERT, and Qwen2.5. It achieves an F1 score of ~0.87 with significant savings of $9.3M/year at a 10M-user scale compared to GPT-4o-mini, while also detecting hallucinations with an AUC of 0.90.