FDA Clears First Continuous AI Sepsis Monitor, Reducing Mortality by 18%
Tags AI · Enterprise
The FDA granted 510(k) clearance to Bayesian Health's Targeted Real-Time Early Warning System (TREWS), the first continuous AI sepsis monitor, on May 12, 2026. Developed by Johns Hopkins researchers led by Prof. Suchi Saria, the system detects sepsis 2-48 hours before clinical suspicion at 82% sensitivity and reduced in-hospital mortality by 18% when clinicians acted on alerts. Unlike other FDA-cleared sepsis tools that require a clinician to suspect sepsis first, TREWS runs continuously across all hospitalized patients. The system has been deployed at Cleveland Clinic, MemorialCare, and University of Rochester. FDA clearance positions Bayesian for Medicare/Medicaid NTAP reimbursement starting October 2026.
Technical significance
This is a landmark for clinical AI: the first time the FDA has cleared a continuous AI monitoring tool that operates before clinical suspicion, rather than confirming a doctor's existing hypothesis. The 18% mortality reduction, validated across 764,707 patient encounters, sets a new evidentiary bar for clinical AI. NTAP reimbursement eligibility starting October 2026 removes the financial barrier that has slowed hospital adoption of AI tools.