Privacy-protecting, reliable response data discovery using COVID-19 patient observations

Area of Science:

• Health Informatics
• Epidemiology
• Biostatistics

Background:

• Electronic Health Records (EHR) contain vast patient data crucial for public health research.
• Centralized data analysis poses privacy and logistical challenges.
• COVID-19 research requires rapid access to diverse patient information.

Purpose of the Study:

• To develop a privacy-preserving method for analyzing multi-institutional EHR data for COVID-19 research.
• To create a federated network enabling collaborative analysis of EHR data without data centralization.
• To demonstrate the feasibility of answering clinical questions using distributed EHR data.

Main Methods:

• Established a distributed, federated network connecting 12 health systems.
• Harmonized EHR data and developed distributed algorithms for analysis.
• Utilized multivariate, iterative regression models on aggregated data.

Main Results:

• Generated counts, descriptive statistics, and regression models from distributed EHR data.
• Identified associations between medications (ACE inhibitors, ARBs) and lower in-hospital mortality.
• Demonstrated that age, sex, and ethnicity were not significantly associated with mortality after adjustment.

Conclusions:

• A federated network provides a viable alternative to centralized COVID-19 registries.
• Multivariate distributed logistic regression can yield results from EHR data without transferring individual-level data.
• This approach respects institutional privacy and regulatory requirements.