Researchers have developed a new Electronic Health Record (EHR) Foundation Model designed for large-scale chronic disease prediction. This model, trained on billions of medical events from over 5 million patients across Taiwan and the United States, utilizes a unified code alignment framework to handle data heterogeneity. The model demonstrates strong scaling capabilities, with versions up to 2.4 billion parameters, and outperforms existing tree-based and general language models on 11 chronic disease prediction tasks. It also shows robust few-shot generalization on the EHRShot benchmark, even with significant distribution shifts, and highlights the benefits of aligned cross-system data for pretraining in data-limited healthcare settings. AI
IMPACT This model's ability to generalize across different patient populations and healthcare systems could significantly improve population health management and chronic disease prediction globally.
RANK_REASON The cluster contains an academic paper detailing a new model and its performance on benchmarks. [lever_c_demoted from research: ic=1 ai=1.0]
- arXiv
- EHRSHOT
- Hao-Ren Yao
- International Statistical Classification of Diseases and Related Health Problems
- IsoFLOP
- Scaling Electronic Health Record Foundation Models for Population Health Management
- Taiwan
- United States
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
