sunlabuiuc/PyHealth

Provides modular 5-stage pipelines with 33+ pre-built models (RNN, LSTM, Transformer, RETAIN, SafeDrug, etc.) and optimized support for clinical datasets (MIMIC, eICU, OMOP-CDM). Features healthcare-specific data processing 3x faster than pandas for temporal medical codes and EHR records. Integrates PyTorch-based trainers with production-ready metrics across 10+ prediction tasks including mortality, readmission, and drug recommendation.

1,486 stars. Actively maintained with 15 commits in the last 30 days. Available on PyPI.