State and local Health Information Exchanges (HIEs) receive data from a high density of healthcare providers in their coverage area, and, in the aggregate, from more than 60% of US hospitals. HIEs exchange patient health information with clinicians, public health agencies, and laboratories and link, analyze, and aggregate that data. However, despite this availability of robust patient-level electronic health data, these datasets are rarely used for research purposes because of technical and privacy related barriers. Privacy-preserving artificial intelligence (AI) and machine learning (ML) techniques can leverage HIE data to conduct complex patient outcomes research studies and further the understanding of COVID-19 and its progression.
This project contributes to the Department of Health and Human Services strategic goal of strengthening and modernizing the nation’s data infrastructure by creating a foundation to use electronic health data from HIEs for patient-centered outcomes research (PCOR).
This project began in 2021 and ends in 2024.
The goal of this project is to create a foundation to use electronic health data from HIEs for PCOR by implementing data standards, APIs, and privacy-preserving machine learning (ML) infrastructure. It will accomplish this by:
- Implementing the United States Core Data for Interoperability (USCDI) and Bulk FHIR API at three HIEs to facilitate interoperable and efficient data access.
- Testing the use of split learning, a privacy-preserving machine learning technique, with HIE data to address a COVID-19 and PCOR-related research question.
- Disseminating resources to support the adoption by HIEs of technologies and methods used in the project and to encourage PCOR researchers to use HIEs and their data for research.
This component process FHIR bulk data based on a cohort and transforms to ML data model
This component executes the split learning model on the data from the fhir-data component