Robust-ODAL: Learning from heterogeneous health systems without sharing patient-level data

Area of Science:

• Health Informatics
• Biostatistics
• Distributed Computing

Background:

• Electronic Health Records (EHR) offer rich patient data for health research, but data sharing is restricted.
• Distributed algorithms enable multi-site analysis using aggregated data, overcoming privacy concerns.
• Existing distributed methods often overlook data heterogeneity, introducing bias in outcome-exposure association studies.

Purpose of the Study:

• To develop a privacy-preserving and communication-efficient distributed algorithm for analyzing heterogeneous EHR data.
• To address the bias caused by data heterogeneity in multi-site health research.
• To improve the accuracy of estimation and prediction in distributed health data analysis.

Main Methods:

• Proposed a novel distributed algorithm designed to account for data heterogeneity across clinical sites.
• Conducted systematic simulations based on real-world health data scenarios.
• Applied the algorithm to multiple claims datasets from the Observational Health Data Sciences and Informatics (OHDSI) network.

Main Results:

• The proposed algorithm demonstrated superior performance compared to the existing ODAL distributed algorithm.
• The new method outperformed a standard meta-analysis approach in handling heterogeneous data.
• Simulations and real-world data application confirmed the algorithm's effectiveness.

Conclusions:

• The developed distributed algorithm effectively addresses data heterogeneity in multi-site EHR analysis.
• This approach offers a more accurate and reliable method for distributed health research.
• The findings have implications for improving large-scale observational health studies.