D. Silver1, L. Lu1, J. Beiriger1, S. Boland1, J. Sperry1, F. Guyette1, S. Wisniewski1, E. Moore2, M. Schreiber3, B. Joseph4, C. Wilson5, B. Cotton6, D. Ostermayer6, B. Harbrecht7, M. Patel8, J. B. Brown1 1University of Pittsburgh, Pittsburgh, PA, USA 2Denver Health, Denver, CO, USA 3Oregon Health And Science University, Portland, OR, USA 4University Of Arizona, Tucson, AZ, USA 5Baylor College Of Medicine, Houston, TX, USA 6University Of Texas Health Science Center At Houston, Houston, TX, USA 7University Of Louisville, Louisville, KY, USA 8Vanderbilt University Medical Center, Nashville, TN, USA
Introduction: The quality of prehospital care has a significant impact on patient outcomes. Our prior work demonstrated a relationship between the annual volume of trauma patients at the emergency medical services (EMS) agency level and early mortality. This study explores the relationship between EMS agency volume and unexpected early survival among patients with severe trauma.
Methods: We analyzed a prospective multicenter cohort of patients with injury severity score >8 transported by 20 participating EMS agencies. The cohort comprised eight trauma centers and 20 EMS air medical and metropolitan ground transport agencies. We used an XGBoost Classifier optimized with a gradient boosting framework and parallel tree boosting to predict 6-hour mortality as our primary outcome adjusted for demographics, comorbidities, injury severity/characteristics, vital signs, mechanism, transport mode, transfusion requirements, and need for hemorrhage control. We then identified the threshold mortality prediction that maximized discrimination measured by the F1 score. Patients above this threshold were classified as predicted 6-hour mortality. Any patient with predicted 6-hour mortality that survived was defined as an unexpected survivor. We characterized median agency annual volume, demographics, clinical features and interventions, and outcomes of unexpected survivors.
Results: Our model predicted mortality with excellent performance (ROC AUC 0.999). Within the cohort, 334 patients were predicted to die at 6 hours, and 31 were unexpected survivors. Compared with expected deaths, unexpected survivors had similar measures of injury severity, prehospital hemodynamics, and prehospital interventions compared to the expected mortality cases (p>0.05). Notably, there was no difference in prehospital crystalloid or blood product administration, and none of the patients underwent prehospital intubation. Median annual agency volume was significantly higher in the unexpected survivor group compared to the expected mortality group (1180 vs 660 trauma patients annually, p<0.001; Figure). This finding persisted for both the 24-hour mortality and in-hospital mortality. The 31 unexpected survivors came from two agencies, one hospital (16%) and one fire department affiliated (84%). Both agencies were ground-based and exhibited a median annual volume above the 75th percentile among all participating EMS agencies.
Conclusion: Annual EMS agency volume of moderate to severely injured patients may be associated with increased unexpected survivors. Further investigation of this volume-outcome relationship should focus on which injury patterns are most sensitive to volume to optimize regional benchmarking and educational targets.
