Home Trauma Prehospital trauma scores poorly predict secondary transfer
Trauma
Prehospital trauma scores poorly predict secondary transfer
Prehospital Trauma Scoring Systems for Discriminating Recorded Secondary Transfer in Multiply Injured Patients: A Comparative Study with Machine Learning
Journal of Clinical Medicine · 2026-11-12 · Added 17 November 2026
A Turkish EMS record review of 1,194 multiply injured patients compared eight prehospital trauma scores for predicting secondary interhospital transfer, finding CRAMS performed best but only modestly (AUC 0.676), while a random forest model reached AUC 0.836.
- Study type
- Cohort
- Design
- Retrospective cohort with ROC comparison of eight trauma scores and machine learning
- Participants
- 1,194 multiply injured patients (47% secondarily transferred)
- Main finding
- CRAMS had the highest score discrimination (AUC 0.676); paediatric age (OR 4.46), elderly age (OR 2.44) and thoracic injury (OR 1.68) were independently associated with transfer; random forest AUC 0.836 versus 0.765 for logistic regression, with calibration compressed toward average risk
- Caveat
- Single-province EMS records, weak score discrimination, and a model needing recalibration and external validation
- Bottom line
- Hypothesis-generating
Why it matters in Australia (Medium relevance). Trauma destination and transfer decisions are made by crews everywhere, but the scores discriminated weakly and the better-performing machine-learning model is unvalidated.
Learning goal
Strengthen my judgement on when a multiply injured patient will need secondary interhospital transfer.
More
- Topic: Trauma
- Daily scan: 17 November 2026
Related papers on Trauma
-
Field amputation: broad capability, little standardisation
A survey of all English air ambulance services found every one could perform prehospital field amputation and four had completed six cases in five years, with a Gigli saw the standard bone instrument and little national standardisation.
-
rSIG cutoff flags trauma patients needing intervention
A Japanese single-centre retrospective cohort tested prehospital rSIG (reverse shock index multiplied by GCS) in 747 trauma patients, finding rSIG below 24.8 was associated with resuscitative interventions (adjusted OR 3.17) and 24-hour transfusion (aOR 4.20) in the validation cohort, but not with surgery, interventional radiology or in-hospital mortality.
-
GCS-based triage misses seriously injured trauma patients
A South Korean linked EMS–hospital cohort of 5,841 suspected severe trauma patients found weak agreement between prehospital GCS classification and diagnosis-based injury severity (κ 0.192), with one in ten patients (10.0%) severely injured despite GCS 9–15 and destination undertriage of 33.6% (GCS-based) and 47.3% (diagnosis-based), roughly doubled in rural areas.
-
Fibrinogen concentrate supply grows in Queensland
A ten-year Queensland Health supply analysis of 24,717 g of fibrinogen concentrate found 97% went to 54 hospitals and only 1% to ambulance services, while population-adjusted supply rose from 8.19 to 59.2 g per 100,000 between 2016 and 2025.