The future of federally funded research at Harvard Medical School — supported by taxpayers and done in service to humanity — remains uncertain. Learn more.
The first large-scale study of how often clinicians use artificial intelligence tools in patient care, who those clinicians are, and what they are using it for finds that a significant number of medical practitioners are using AI in the clinic and that they are often using it to explore treatment options for their patients in areas outside their clinical specialty.
The study, led by researchers at Harvard Medical School and Brigham and Women’s Hospital, looked at how 950,000 physicians, physician assistants, and nurse practitioners in the United States used OpenEvidence, a popular AI-enabled clinical information platform, over the last two and a half years. Most queries were related to clinical decision making (51.9 percent), particularly treatment and management. Diagnosis accounted for 13.9 percent of queries.
The results of this federally funded study were published Oct. 15 in JAMA.
“Two key takeaways are that an enormous number of physicians are using AI for clinical care and that they are often using it in areas where they don’t have specific expertise,” said corresponding author Anupam Jena, the Joseph P. Newhouse Professor of Health Care Policy in the Blavatnik Institute at HMS and HMS professor of medicine at Massachusetts General Hospital.
“Everyone appreciates the huge potential for AI to change how medicine is practiced, but right now we have limited information on how many doctors are using it, what they are using it for, and how it’s impacting patients,” Jena said. The JAMA study tackles the first two of those questions.
Dramatic increase in use of AI
The researchers found that use has been dramatic, in terms of both the speed and scale of adoption, Jena said. That was in contrast to what research showed with electronic health records, where the pace of adoption was slow and “only really occurred after government regulation,” Jena said.
From January 2024 to June 2026, the number of monthly active OpenEvidence users increased from 355 to 703,265, the team found. Over that time, practitioners included in the study submitted 270 million queries to the platform.
Jena and colleagues used a large language model to classify the clinical purpose and topic areas of a random sample of 38.8 million queries.
What are doctors using AI for?
The data showed that queries were frequently for aiding in treatment decisions rather than diagnostic tasks and that questions were often about subjects that were outside of the practitioner’s specialty. For example, surgeons often ask questions that are not about surgery, the team found.
The level of use varied across clinician types, specialties, and career stages.
The researchers compared the number of queries made by a particular specialty with the percentage of the physician workforce that they represent. Some specialties used more than their share of queries. Oncology/hematology specialists accounted for 3.8 percent of queries, while they comprise just 1.9 percent of the workforce and internal medicine doctors accounted for 27.3 percent of physician queries, while they make up 14.5 percent of the physician workforce. Other specialties, including radiology, orthopedics, and ophthalmology, used less than their share.
More than half of the users in the sample (556,200) were physicians. They submitted an average of 37 queries per month of use. Early-career physicians and those in oncology, hematology, internal medicine, and family medicine submitted a higher-than-average share of queries.
The next step will be to examine whether use of these tools improves clinical care, the researchers said.
“AI tools have become integrated into routine in clinical practice faster than almost any technology we’ve seen, but until now we’ve had very little data on how clinicians actually use them,” said first author Vishal Patel, HMS clinical fellow in surgery at Brigham and Women’s. “This research gives us an early picture of where these tools fit into clinical decision-making and where we need to study their accuracy and effects on patient care next.”
Three of the co-authors are employees of OpenEvidence, which did not fund the study but provided the de-identified data that was analyzed.
Authorship, funding, disclosures
Additional authors include Samuel G. Finlayson, Michael Liu-Hass, Mondira Ray, Jesse Burk-Rafel, and Travis I. Zack.
This research was supported by the National Institutes of Health/National Institute of Biomedical Imaging and Bioengineering (grant F32EB040202).
Finlayson reported being a salaried employee of OpenEvidence during the conduct of the study. Liu-Hass reported receiving personal fees from Accompany Health for consulting outside the submitted work. Ray reported receiving personal fees from OpenEvidence as a full-time employee outside the submitted work. Zack reported being the chief medical officer of OpenEvidence during the conduct of the study. Jena reported receiving personal fees from Analysis Group; Freakonomics, MD; Doubleday Books; The New York Times; Los Angeles Times; The Wall Street Journal; and Harry Walker Agency outside the submitted work and serving as an unpaid board member for the United Network for Organ Sharing (UNOS). No other disclosures were reported.
