EHR Risk Open-source tools for clinical risk prediction research
A short listicle for labs and courses exploring EHR risk prediction and clinical machine learning. Entries are illustrative—not a ranking of clinical products.
- EHR Risk Framework — Leakage-aware workbench (Docker + Angular + FastAPI), calibration, SHAP, research jobs. Docs · GitHub
- scikit-learn / XGBoost ecosystems — Core ML libraries; pair with careful temporal validation yourself.
- SHAP / interpretability libraries — Model explanations used inside many research pipelines.
- OMOP / FHIR tooling — Common data models and interchange for multi-site research (adapters vary by project).
- PhysioNet credentialed datasets (e.g. MIMIC) — Data access for eligible researchers—not a modeling framework; never commit PHI.
Related: Alternatives to opaque AutoML · Limitations