Add Synthetic EHR Evaluation Example - Will Hunnius Jiesen Zhang

[willhunnius]

Synthetic EHR Data Evaluation Example

Description

This PR adds a comprehensive example demonstrating how to evaluate synthetic Electronic Health Record (EHR) data quality using PyHealth.

Contribution Type

• [x] New example/use case

Based On

Lin et al. (2025) "A Case Study Exploring the Current Landscape of Synthetic Medical Record Generation with Commercial LLMs" - JMLR CHIL 2025 https://arxiv.org/abs/2504.14657

Files Added

• examples/synthetic_ehr_evaluation.py - Main example script

Features

• Fidelity Metrics: KL divergence for distribution matching
• Utility Metrics: TSTR (Train-Synthetic, Test-Real) evaluation
• Privacy Metrics: Membership inference attack evaluation
• Visualizations: Distribution comparisons, ROC curves, summary charts

Usage

evaluator = SyntheticDataEvaluator(target_column="mortality")
results = evaluator.evaluate(real_data, synthetic_data)
print(f"TSTR AUC: {results['utility']['tstr_auc']:.3f}")

Course Information

• Course: CS598 DL4H - Deep Learning for Healthcare
• University: University of Illinois Urbana-Champaign
• Authors: Will Hunnius, Jiesen Zhang

Checklist

• [x] Code follows PEP8 style guidelines
• [x] Added docstrings with Google style
• [x] Example runs without errors
• [x] Notebook is well-documented