Abstract
Privacy protection is paramount in conducting health research. However, studies often rely on data stored in a centralized repository, where analysis is done with full access to the sensitive underlying content. Recent advances in federated learning enable building complex machine-learned models that are trained in a distributed fashion. These techniques facilitate the calculation of research study endpoints such that private data never leaves a given device or healthcare system. We show—on a diverse set of single and multi-site health studies—that federated models can achieve similar accuracy, precision, and generalizability, and lead to the same interpretation as standard centralized statistical models while achieving considerably stronger privacy protections and without significantly raising computational costs. This work is the first to apply modern and general federated learning methods that explicitly incorporate differential privacy to clinical and epidemiological research—across a spectrum of units of federation, model architectures, complexity of learning tasks and diseases. As a result, it enables health research participants to remain in control of their data and still contribute to advancing science—aspects that used to be at odds with each other.
Highlights
Protecting privacy is crucial in designing, running, and interpreting health studies
Most health research to date uses data stored in a centralized database, where analysis and model fitting is done with full access to the sensitive underlying data
As there are growing concerns about the ability to maintain the privacy of research participant data as it becomes increasingly feasible to re-identify individuals through combining multiple sources of electronic health data[9,10] we show that new methods involving federated learning and differential privacy can provide very strong privacy protections with minimal reduction in utility
Summary
Protecting privacy is crucial in designing, running, and interpreting health studies. Most health research to date uses data stored in a centralized database (i.e., a database stored in a single site), where analysis and model fitting is done with full access to the sensitive underlying data. Recent advances in distributed machine learning (i.e., machine learning utilizing data stored across two or more sites) enable building complex machinelearned models without necessitating such centralized databases. Federated learning techniques enable calculation of research study endpoints in a privacy-preserving fashion such that private data never leaves a given device (e.g., a research participant’s smartphone, wearable or implanted device) or system (e.g., academic research center, clinical trial site or medical data repository). Focused model updates leave the clients[1], enabling the aggregation of learned patterns into a single global model without raw data disclosure. The communication between clients can be peer-to-peer but typically involves a central orchestrator that receives and aggregates clients’ updates
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
