Joseph Ross

Delighted to share my 1st publication in @JAMAHealthForum, exploring the patient safety implications of @US_FDA-regulated, AI-enabled medical devices: jamanetwork.com/journals/jam… Our study finds AI devices, especially those developed by publicly traded companies, are significantly more likely to be recalled. Many of these tools were never tested on patients before entering clinical use. Publicly traded companies, the big fish in this still-small pond, developed just over half of the devices but were responsible for nearly all of the recalls. We were stunned to find that almost half of all AI device recalls occurred within the first year after approval. Even more concerning: 95% of these recalls involved tools with no reported clinical validation. If AI hasn’t been tested on people, then people become the test. Our team, spanning @HopkinsMedicine, @JHUCarey, @JohnsHopkinsSPH, @YaleMed, and @GUMedicine, examined nearly 1,000 @US_FDA-cleared AI devices. Grateful to Branden Lee, who brilliantly led our phenomenal group of @JohnsHopkins medical students—Patrick Kramer, Sara Sandri, Ritika Chanda, @CrystalFavorito, and Olivia Nasef—and to co-authors Drs. @jsross119 and @drJoshS.