Creating an automated trigger for sepsis clinical decision support at emergency department triage using machine learning
Problem
Identifying patients with suspected infection early at emergency department triage is difficult, and the study tested whether free-text data adds predictive value beyond vitals and demographics.
Why it matters
Delayed sepsis recognition at triage increases mortality and cost in emergency departments.
What the AI did
Machine learning models (bag-of-words/topic-model text features with a support vector machine) were trained retrospectively to predict infection from ED free text, vitals, and demographics.
Results
not stated in source (abstract truncated before performance figures)
Impact
Retrospective single-centre demonstration only; no deployment or patient-outcome evidence stated in abstract.
Record
| Venue | PLoS ONE |
|---|---|
| Published | 2017 |
| Work type | article |
| Authors | Steven Horng, David A. Sontag, Yoni Halpern, Yacine Jernite, Nathan Ivan Shapiro, Larry A. Nathanson |
| Citations | 308 (OpenAlex, at import time) |
| References | 26 |
| Model or method | support vector machine (with bag-of-words and topic models) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1371/journal.pone.0174708 |
| Open access | gold |
| Keywords | sepsis, emergency department, triage, machine learning, clinical decision support, free text mining |
Importance rating
significant — thousands to millions affected
Early infection identification at ED triage matters for many patients; abstract quantifies 230,936 visits with ~14% diagnosed infections at one tertiary hospital, but no mortality/outcome impact is stated.