Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-0576
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-07-27T06:08:25.237225Z
status: published
lens: g_space
sector: health
headline: Beyond Thresholds: Can Machine Learning Improve Trauma Field Triage?
dek: BackgroundAccurate triage of trauma patients by Emergency Medical Services (EMS) is essential for optimal outcomes and resource allocation. The 2021 National Field Triage Guidelines (FTG) assist EMS in prehospital triage; however, its collective performance has never been evaluated using a national database. We aimed to evaluate an FTG surrogate and develop a predictive model to identify patients at risk for serious injury.MethodsThe Trauma Quality Improvement Program National Trauma Databank (2017-2020) was que…
gain_title: An XGBoost model trained on routinely collected prehospital vitals and demographics reduced under-triage and over-triage compared to a surrogate of the 2021 National Field Triage Guidelines in 1.2M trauma patients.
problem_title: (none)
trace_subject: (none)
gain_reading: An XGBoost model trained on routinely collected prehospital vitals and demographics reduced under-triage and over-triage compared to a surrogate of the 2021 National Field Triage Guidelines in 1.2M trauma patients.
gain_evidence: Machine learning outperformed a database-derived FTG surrogate on a national trauma database, reducing both under and over-triage rates.
problem_reading: (none)
problem_evidence: (none)
quick_read: Using a national trauma databank of over 1.2 million ambulance-transported adults from 2017-2020, researchers built an XGBoost model from routinely collected prehospital vitals, demographics, and field triage criteria to predict serious injury and compared it to a database-derived surrogate of the 2021 National Field Triage Guidelines.

The finding matters because accurate EMS triage directly affects survival and resource use, and a model that lowers both under- and over-triage using existing data could improve outcomes, but uncertainty remains about generalizability beyond retrospective data and performance in live EMS workflows without prospective validation.
limitation: 
tag: Evidence-backed gain
key_points: Study used Trauma Quality Improvement Program National Trauma Databank 2017-2020 for 1,267,039 adults transported via ambulance. | Serious injury defined as ISS 16, transfusion, hemorrhage control surgery, intubation, ICU admission, or death among other criteria. | Model used 60/20/20 train/validate/test split with SHapley Additive exPlanations for feature importance. | Most influential features were GCS Verbal, sex, prehospital blood pressure, pulse oximetry, and GCS Motor.
rundown: Researchers queried 1,267,039 adult trauma patients transported by ambulance in TQIP 2017-2020, of whom 32.7% met a composite serious injury definition including ISS 16, transfusion, angiography with intervention, hemorrhage control surgery, intubation, ICP monitoring, ICU admission, or death.

They engineered FTG criteria variables and trained an XGBoost model on prehospital vitals, demographics, and triage criteria, evaluating against an FTG surrogate, with SHAP analysis highlighting GCS Verbal, sex, blood pressure, pulse oximetry, and GCS Motor as top predictors.
sources:
- peer_reviewed | The American Surgeon™ | https://doi.org/10.1177/00031348261471469 | 2026-07-24
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

e0ee3dc4cd7b6bcc76ece916ba55c216292b09011687e6b5bcd18b43296d9c0c

previous

0000000000000000000000000000000000000000000000000000000000000000