Prospective validation and comparison of clinical prediction models for early trauma care: A multicentre cohort study

ObjectiveWe aimed to prospectively validate and compare published prediction models and clinician-assigned triage categories for early trauma care. DesignProspective multicentre cohort study. SettingThree public hospitals in urban India: one secondary care hospital in Mumbai and one tertiary care teaching hospitals in Delhi and Kolkata each. ParticipantsAdult patients aged over 18 years presenting to the emergency department with a history of trauma between 2016 and 2022. A total of 13,041 patients were included in the final analysis. MethodsWe externally validated five published trauma prediction models (GAP, Gerdin, KTS, MGAP, and RTS) and clinician-assigned triage categories based on initial assessment. The primary outcome was 30-day all-cause mortality. Models were recalibrated using a separate updating sample prior to evaluation, and model performance was assessed in terms of discrimination (AUC), calibration (calibration slope and plots), and decision curve analysis. ResultsAll models and clinician gestalt-based triage demonstrated excellent discrimination (AUC range: 0.90-0.96) and good calibration after updating. The GAP model achieved the highest AUC (0.96, 95% 0.94-0.97), and RTS demonstrated the highest sensitivity (0.70). ConclusionSimple, physiology-based prediction models and clinician gestalt both demonstrated excellent performance in predicting 30-day mortality among adult trauma patients in Indian emergency departments. These findings provide a practical foundation for further development of trauma triage systems.