ehr-analysis

What problem does it solve?

End-to-end EHR predictive modeling pipelines are complex and time-consuming; this skill provides a structured pattern to load diverse EHR datasets, define clinical tasks, train models, evaluate results, calibrate predictions, and interpret outcomes.

Core Features & Use Cases

• End-to-end EHR predictive modeling pipeline covering dataset loading, task definition, model training, evaluation, calibration, and clinical interpretation.
• Supports datasets such as MIMIC-III, MIMIC-IV, eICU, OMOP-CDM, or custom datasets, with tasks including mortality, readmission, length of stay, and drug recommendation; includes medical code normalization and ontology mapping; supports calibration and interpretability.
• Workflow and governance artifacts include reproducible experiments, task schemas, and result reporting for clinical AI research.

Quick Start

Load an EHR dataset with PyHealth, define a clinical task, train a model, and evaluate on a held-out patient test set.