5-Stage Pipeline in <15 Lines

The same pattern works for any task — swap the dataset and task class to move between mortality, readmission, drug recommendation, or imaging.

from pyhealth.datasets import MIMIC3Dataset, PatientSplit, get_dataloader
from pyhealth.models import Transformer
from pyhealth.tasks import MortalityPredictionMIMIC3
from pyhealth.trainer import Trainer

if __name__ == "__main__":
    # 1. Load data: raw tables are processed once and cached
    dataset = MIMIC3Dataset(
        root="data/mimic3",
        tables=["diagnoses_icd", "procedures_icd", "prescriptions"],
        cache_dir="cache/",
    )

    # 2. Task + patient split: preprocessing is fit on training patients only
    train_ds, val_ds, test_ds = dataset.set_task(
        MortalityPredictionMIMIC3(), split=PatientSplit((0.8, 0.1, 0.1), seed=0)
    )
    train_loader = get_dataloader(train_ds, batch_size=32, shuffle=True)
    val_loader = get_dataloader(val_ds, batch_size=32)
    test_loader = get_dataloader(test_ds, batch_size=32)

    # 3. Train and evaluate
    model = Transformer(dataset=train_ds)
    trainer = Trainer(model=model, metrics=["pr_auc", "roc_auc"])
    trainer.train(train_loader, val_loader, epochs=50, monitor="pr_auc")
    print(trainer.evaluate(test_loader))

Built for Scale — Efficient by Design

Benchmarked on MIMIC-IV at 4 parallel workers. PyHealth 2.0 uses a memory-mapped architecture that dynamically adapts to your hardware — whether you have 2 cores or 64, it scales without manual memory management so you can focus on the ML, not the infrastructure. Pandas values shown at 1 worker only (†) as it could not scale beyond single-threaded execution. PyHealth 2.0 even runs the full in-hospital mortality task on a consumer laptop in roughly 5 GB of memory.

† Pandas measured at 1 worker; failed beyond single-threaded execution. All other methods at 4 workers on MIMIC-IV. Laptop rows are PyHealth 2.0 on a MacBook Pro M2 Pro and an ASUS Zephyrus G16 at 4 workers, benchmarked for in-hospital mortality. Full benchmark in the PyHealth 2.0 paper (ICML 2026).

How PyHealth Standardizes Clinical AI Development

A unified API from raw clinical data to trustworthy, interpretable models — follow the pipeline step by step.

Resources

Everything you need to get started or get involved.

Cite PyHealth

If PyHealth helps your research, please cite the PyHealth 2.0 paper (ICML 2026). If you use the 1.x releases, cite the KDD 2023 paper.

@inproceedings{wu2026pyhealth,
  title={PyHealth 2.0: A Comprehensive Open-Source Toolkit for Accessible and Reproducible Clinical Deep Learning},
  author={John Wu and Yongda Fan and Zhenbang Wu and Paul Landes and Eric Schrock and Sayeed Sajjad Razin and Arjun Chatterjee and Naveen Baskaran and Joshua Steier and Andrea Fitzpatrick and Bilal Arif and Rian Atri and Jathurshan Pradeepkumar and Siddhartha Laghuvarapu and Junyi Gao and Adam Richard Cross and Jimeng Sun},
  booktitle={Forty-third International Conference on Machine Learning},
  year={2026},
  url={https://openreview.net/forum?id=gMLVFN9hl8}
}

@inproceedings{pyhealth2023yang,
  author={Yang, Chaoqi and Wu, Zhenbang and Jiang, Patrick and Lin, Zhen and Gao, Junyi and Danek, Benjamin P. and Sun, Jimeng},
  title={{PyHealth}: A Deep Learning Toolkit for Healthcare Applications},
  booktitle={Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
  pages={5788--5789},
  year={2023},
  doi={10.1145/3580305.3599178}
}