In May, Bayesian Health announced FDA clearance for its sepsis-flagging device—the first continuous AI monitor permitted to warn clinicians about the risk of a medical emergency that contributes to more than one-third of all hospital deaths. Behind the tool is Suchi Saria, a Johns Hopkins computer scientist who has spent more than a decade turning hospitals’ electronic health records into early-warning systems.
The mission became personal in 2017, when Saria lost her nephew to sepsis. Today, Bayesian Health’s system analyzes clinical notes, lab results, clinical history, and vital signs for shifts that suggest a patient is at risk for an adverse medical event. “Pre-suspicion screening is what creates lead time,” she said in an interview with Johns Hopkins, “and lead time is what changes outcomes in sepsis.” At Cleveland Clinic, early data showed 46% more sepsis cases identified by Bayesian Health’s systems, 10 times fewer false alerts, and seven times as many warnings before antibiotics were administered.
The FDA approved the system to be used as an aid to clinicians, rather than a replacement for their judgment. Saria must now show that its reported gains survive expansion across hospitals with different patients, practices, and electronic-record systems.
