Confirmed TREWS sepsis alerts are linked to 3.3% lower mortality

AI system

System: TREWS (Targeted Real-time Early Warning System)
Model: Not assessed
Release: Not assessed
Application developer: Johns Hopkins University (Suchi Saria's group)
Role: Not assessed
Configuration: Not assessed
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: EHR-integrated deployment with alerts presented to bedside providers. A provider confirmation workflow linked to sepsis treatment.
Human oversight: Not assessed
Relation to AI: The study is observational. It compares patients whose TREWS alert was confirmed within 3 hours with those confirmed later or not at all, adjusting for presentation and severity. It supports an association, not causation. A medRxiv critique titled “Randomized Clinical Trials or Convenient Controls: TREWS or FALSE?” (Nemati et al., 2022) questions the comparison design. The sources include only the paper's abstract, not its competing-interests statement.
AI contribution: The study is observational. It compares patients whose TREWS alert was confirmed within 3 hours with those confirmed later or not at all, adjusting for presentation and severity. It supports an association, not causation. A medRxiv critique titled “Randomized Clinical Trials or Convenient Controls: TREWS or FALSE?” (Nemati et al., 2022) questions the comparison design. The sources include only the paper's abstract, not its competing-interests statement.

Benefits

Outcome identifiers: Not assessed
Benefit: 3.3% absolute (18.7% relative) reduction in in-hospital mortality
Category: Improved Physical Health
Status: Not assessed
Beneficiary groups: People. Society.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Measurement

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: 3.3% absolute (18.7% relative) reduction in in-hospital mortality
Comparison: alerted sepsis patients whose alert was not confirmed by a provider within 3 h
Results: Reference: patients whose alert was not confirmed by a provider within 3 h. Result: reduced in-hospital mortality rate (3.3%, confidence interval (CI) 1.7, 5.1%, adjusted absolute reduction, and 18.7%, CI 9.4, 27.0%, adjusted relative reduction).
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Money

Kind of value: Not assessed
Amount: Not assessed
Currency and prices: Not assessed
Period and unit: Not assessed
Realization: Not assessed
Recipients and payers: Not assessed
Gross or net: Not assessed
Included costs: Not assessed
Comparison: Not assessed
Method: Not assessed
Return and payback: Not assessed

People and places

Participants: Developer: Johns Hopkins University (Suchi Saria's group). Operator: Johns Hopkins Medicine (five hospitals).
Jurisdictions: Benefit: Baltimore, United States. Developer: United States. Operator: United States.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Follow-up

Conditions for replication: Providers evaluate alerts promptly (the benefit was seen only for alerts confirmed within 3 hours). Comparable patient mix and alert thresholds.
Replication attempts: Not assessed
Later findings: No later development appears in the cited source.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Evidence

Source numbers: AIBD-0066-source-0001–0002
Citations: Numbered references below.
Language: Not assessed
Source relationships: Developer
Original evidence: Not assessed
Access and versions: Not assessed
Supported findings: Not assessed
Evidence links: Not assessed
Evidence strength: Not assessed
Missing evidence: Not assessed

Grades

Assessed benefits: People. Society.
Depth: Not assessed
Reasons: People: Cohort-level association (mortality, organ failure, length of stay) without an individually attributable event. Society: The study reports adjusted associations across 590,736 monitored patients, not a discrete population-level event.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: The makers' claim

Entry details

Kind

Not assessed

Related entries

Same system class and task, EHR-integrated machine-learning sepsis prediction: Epic's model showed high false-positive and false-negative rates on seriously ill patients, the sepsis-patient cohort TREWS's confirmed alerts benefited.

Field

Healthcare - sepsis / critical care. EHR-integrated early warning system.

Purpose

TREWS was built to identify patients developing sepsis early so that treatment can start sooner; the mortality association is its intended benefit.

Dates

Date basis unassessed: 2022 (precision unassessed). Updated: 22 October 2026.

Stage

Not assessed

Version

SHA-256: 185690fa4197…

References

Other identifiers: AI Incident Database: 123.
Category mappings: Not assessed
Brief: Not assessed
Revisions: 9 revisions. Latest: 22 October 2026.

Sources

Peer-reviewed study Developer

Adams R, Henry KE, Sridharan A, et al. (2022). Prospective, multi-site study of patient outcomes after implementation of the TREWS machine learning-based early warning system for sepsis. Nature Medicine 28(7):1455-1460. doi:10.1038/s41591-022-01894-0.