Leveraging FHIR in Federated Learning Environments: A Data Harmonization Framework for Cohort Studies

Area of Science:

• Digital Health
• Cardiovascular Disease Research
• Federated Learning

Background:

• MyDigiTwin aims to prevent cardiovascular diseases using a federated learning platform hosted by Dutch Personal Health Environments (PGOs).
• Ensuring data consistency between PGOs and reference datasets is a key challenge for this federated architecture.
• Privacy preservation is central to the federated learning approach for prediction models.

Purpose of the Study:

• To introduce a novel data harmonization framework for efficient generation of FHIR-based representations.
• To streamline the integration of multiple cohort study data into the MyDigiTwin federated research infrastructure.
• To address the challenge of data consistency in a federated learning environment for cardiovascular disease research.

Main Methods:

• Development of a novel data harmonization framework.
• Generation of FHIR-based data representations from cohort studies.
• Application of the framework for integrating Lifelines cohort data into MyDigiTwin.

Main Results:

• The proposed framework enables efficient harmonization of diverse cohort data.
• FHIR-based representations facilitate data integration into the federated infrastructure.
• Demonstrated applicability in integrating Lifelines data, showcasing potential for broader use.

Conclusions:

• The data harmonization framework is effective for integrating heterogeneous cohort data in federated research.
• This approach supports the MyDigiTwin initiative by ensuring data consistency and enabling privacy-preserving analysis.
• The framework facilitates the development of robust prediction models for early cardiovascular disease detection.