FDA Cleared an AI Sepsis Monitor in April; Prospective Data Show an 18.7% Relative Mortality Reduction

FDA Cleared an AI Sepsis Monitor in April; Prospective Data Show an 18.7% Relative Mortality Reduction — With Conditions

Bayesian Health's TREWS platform earned 510(k) clearance April 30, 2026 — the first AI system cleared for continuous sepsis flagging in adult hospital patients. The mortality benefit applied only when clinicians confirmed the alert within three hours.

The FDA on April 30, 2026 cleared the Bayesian Health Sepsis Flagging Device (510(k) number K250680) under the premarket notification pathway, judging it substantially equivalent to Prenosis’ earlier De Novo-authorized ImmunoScore. The software — the Targeted Real-time Early Warning System (TREWS), originally developed at Johns Hopkins University — continuously parses EHR data streams and issues Sepsis Deterioration Alerts for adult inpatients at rising risk. The cleared indication covers hospital-wide adults; the device is not limited to the ICU.

The pivotal evidence is a prospective, multi-site implementation study published in Nature Medicine in July 2022 (Adams et al.). Across five hospitals, 590,736 patients were monitored; 6,877 met sepsis criteria and received a TREWS alert before antibiotic initiation. This was a prospective pre-post observational design with concurrent unexposed comparators — not a randomized controlled trial.

The primary finding: among sepsis patients whose alert was confirmed by a clinician within three hours of firing, in-hospital mortality fell by 3.3 percentage points (adjusted absolute reduction; 95% CI 1.7–5.1), representing an 18.7% relative reduction (95% CI 9.4–27.0). For the subgroup additionally flagged as high-risk, the absolute mortality benefit was 4.5 points (95% CI 0.8–8.3). TREWS detected 82% of sepsis cases before antibiotic initiation.

The benefit is conditional. The mortality reduction applied specifically to patients whose alerts were acted on within three hours — not across all alerted patients. The prospective observational design does not establish causation, and a confirmatory randomized trial has not been completed. The clearance was effective April 30, 2026; the K-number and cleared indication cited here are from manufacturer and independent trade press sources — the FDA 510(k) database entry for K250680 was not directly reviewed for this brief.

Trace · every claim, sourced

Reported and written by Owen Tanaka, Digital Health & AI Desk, then each load-bearing claim was bound to the primary source it rests on and checked out-of-band against that source before publication. The full mapping is below — nothing here is taken on faith.

Bayesian Health Sepsis Flagging Device received FDA 510(k) clearance (K250680) on April 30, 2026; announced May 12, 2026.

source on file

FDA cleared it via 510(k) as substantially equivalent to Prenosis' De Novo ImmunoScore.

source on file

Prospective multi-site study: 590,736 patients monitored across five hospitals; 6,877 met sepsis criteria; study design is prospective observational, not an RCT.

source on file

In-hospital mortality: 3.3% absolute reduction (95% CI 1.7-5.1), 18.7% relative reduction (95% CI 9.4-27.0), among patients whose alert was confirmed within 3 hours.

source on file

Cleared indication covers adult inpatients hospital-wide, not ICU-restricted; device issues Sepsis Deterioration Alerts.

source on file

After publication, a separate AI panel re-verifies every edition against these same sources. The running claim-confirmation rate and every correction are public on the accuracy ledger.