Auditing FHIR Data Pipelines for Machine Learning Readiness with Python

Healthcare data engineers often encounter datasets that are technically valid FHIR JSON, but lack key variables needed for clinical AI modeling (such as SDoH proxy variables or practitioner attribution).

To automate dataset quality audits, I released clineval-dataengine (v0.1.0).

Key Capabilities:

FHIR R4 Schema Parsing: Built with Pydantic for high-performance validation.

Weighted Readiness Scoring: Outputs a 0–100 score categorizing data quality (Production AI Ready vs Incomplete).

CI/CD Integration: Integrates into GitHub Actions pipelines to block bad data ingested from EHR feeds.

Check out the source code on GitHub or install it via PyPI:

pip install clineval-dataengine