The Wearipedia Project: a free and open-source resource for understanding and using wearables in decentralized clinical trials

Background: Finding the optimal wearable biomedical sensor (ref. wearable) for a clinical research study can be challenging. Many wearables are consumer electronics and are not designed for clinical research and their clinical variables vary widely. We aimed to build a resource for clinical researchers to select the best device for their research study, and programming tools to facilitate wearable research. Methods: For each wearable entry, we document the following; Open-source coding tools: we built data extraction, simulation, statistical testing, and educational materials; Clinical trial usage: trials using the device including ChatGPT-generated summaries; Privacy evaluation: low data risk, HIPAA compliance, de-identification, third-party data sharing, and third-party data sharing transparency; Security evaluation: wearable connectivity and API access protocols. Findings: The Wearipedia database consists of 19 wearables + 5 Apps; 7 smart watches, 4 fitness trackers, 2 chest straps, 2 CGM devices, 1 smart ring, 1 arm strap, 1 under the bed sleep tracker, 1 smart scale, 2 apps for diet tracking, 1 app for questionnaires, and 2 apps for data storage. For public coding tools, there where 891 pages of educational material across 22 wearables and apps. We support data extraction from 13 official APIs and 3 unofficial APIs under the20 Wearipedia pypi package. For clinical usage, there where 63 (std: 99) clinical trials per device. For security and privacy, a total of 87 citations and an average of 3.48 citations are referenced, mostly consisting of privacy policies, terms-of-service agreements, and wearable manuals. The Wearipedia database is conveniently accessible through a website at https://wearipedia.com. Interpretations: Wearables can accurately predict important physiological parameters, glucose, and sleep. However, access to high resolution data can be restrictive, characterizing data accuracy is difficult, and wearable data is often not protected from third party reselling, including government requests. Funding: This work was made possible by the support of the BV and Anu Jagadeesh Family Foundation.