REDCap, the free research data-management platform developed at Vanderbilt Health and now used by more than 8,400 institutions in 166 countries, has taken its first step into generative artificial intelligence.
Writing in JAMIA Open, a journal of the American Medical Informatics Association, Cathy Shyr, PhD, Assistant Professor of Biomedical Informatics, Paul Harris, PhD, Professor of Biomedical Informatics, and colleagues describe how the REDCap team designed, released and tracked early use of three AI-assisted features: a writing helper for text such as survey invitations; a summarization tool for open-ended responses; and an auto-translation tool for deploying forms and surveys in multiple languages.
Surveys and case report forms often include open-ended questions, and manually combing through the resulting free text to summarize responses is laborious. The summarization tool sits alongside these text fields in REDCap reports and produces a summary upon request.
The team deliberately started small, choosing low-risk “researcher helper” tasks in which the AI’s output is a suggestion that a person reviews before accepting, and in which the underlying research data are never altered. The AI features are off by default and can be activated only by each institution’s REDCap administrators, who are advised to connect REDCap to their own institutionally managed, secure AI service.
The features went out to the global REDCap Consortium in January 2025. Over the following seven months, 18 institutions activated the AI module, most of them outside the United States, at sites spanning Europe, Asia, Africa and South America. Eight institutions reported sustained use across 1,171 projects. At VUMC alone, 958 projects used at least one feature; the summarization tool was the most widely used.
The authors note that broader adoption is likely to hinge on institutions establishing their own approved AI infrastructure and governance. Plans for more AI tools in REDCap include extracting information from unstructured text, such as clinical notes, and support for additional AI model providers.
Other authors of the paper are Robert Taylor, MA; Vaishali Jagtap, MSc; and Alex Cheng, PhD. The study was supported by the National Institutes of Health under awards U24TR004432, U24TR004437 and UL1TR002243.
