Performance effectiveness of vital parameter combinations for early warning of sepsis—an exhaustive study using machine learning
Problem
Detecting sepsis early using only noninvasive bedside vital signs so clinicians can intervene sooner.
Why it matters
Sepsis is life-threatening, and earlier detection from routine monitoring could enable timely treatment.
What the AI did
Gradient Boosted Decision Tree models (240) were built on 2630 patients using clinician-interpretable vital-sign features and validated on a geographically distinct dataset.
Results
HR and temperature was the minimal feature set with maximal predictability: AUC 0.94, sensitivity 0.85, specificity 0.90, externally validated.
Impact
Two routine vital signs predicted sepsis with AUC 0.94 on a geographically distinct validation dataset.
Record
| Venue | JAMIA Open |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Ekanath Srihari Rangan, Rahul Krishnan Pathinarupothi, Kanwaljeet J. S. Anand, Michael Snyder |
| Citations | 25 (OpenAlex, at import time) |
| References | 37 |
| Model or method | Gradient Boosted Decision Tree |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1093/jamiaopen/ooac080 |
| Open access | gold |
| Keywords | sepsis, early warning, gradient boosting, vital signs, machine learning |
Importance rating
significant — thousands to millions affected
Early sepsis detection from routine bedside vitals could enable timely intervention in a life-threatening condition; externally validated AUC 0.94, but no scale stated.