Abstract
BACKGROUND: In electronic health records (EHRs), standardization and interoperability challenges stem from fragmented data across institutions. Federated learning, a distributed machine learning framework through which multiple institutions can collaborate on model development while maintaining patient data privacy, bridges this gap by training shared models while keeping data localized. Therefore, this study focused on the application of federated learning in the biomedical domain, with the aim of addressing statistical challenges, system complexities, and privacy issues. METHODS: Following PRISMA guidelines, the authors conducted a comprehensive literature search across PubMed/Medline, Cochrane/EMBASE, PEDro, SCOPUS, MEDLINE, Web of Science, Embase, and arxiv, covering publications from January 2020 to April 2024. The search included terms such as "electronic health records," "EHR," "electronic medical records," "EMR," "registry/registries," "tabular," "federated learning," "distributed learning," and "distributed algorithms." Data were extracted on cohort characteristics, modeling approaches, and federated learning frameworks. RESULTS: After applying inclusion and exclusion criteria to 58 initial results, we analyzed 15 previously-published articles. According to the results described in those articles, federated learning improved data sharing and analysis in various healthcare environments, enhancing EHR standardization and interoperability. Federated learning models typically matched or surpassed localized models, especially when local data was limited or fragmented, and were particularly effective in predicting rare diseases and handling different data types. The use of federated averaging, personalized models, and heterogeneity-aware aggregation methods effectively managed diverse data, ensuring strong performance. Federated learning also maintained privacy and security by keeping patient data local and implementing advanced security protocols like differential privacy. CONCLUSIONS: Federated learning represents a transformative advancement in health informatics, addressing the critical need for seamless data exchange in the fragmented US healthcare landscape. By improving patient outcomes and operational efficiencies, federated learning paves the way for leveraging big data analytics on a nationwide scale.