Little information is available concerning the ability of prehospital triage scores to predict endpoints other than mortality. We evaluated two cohorts (a national cohort of 1,360 patients during 2002 and a single center cohort of 1,003 patients in 2003-2005) of trauma patients receiving care from a prehospital mobile intensive care unit (ICU). We tested the ability of prehospital triage scores (MGAP, Revised Trauma Score [RTS], and triage RTS [T-RTS]) to predict a severe injury, the need for a prolonged ICU period, the occurrence of massive hemorrhage, and the need for emergency procedures, and compared them with a reference score (Trauma-Related Injury Severity Score [TRISS]). The areas under the receiver operating characteristic (AUC(ROC)) curves were measured. The MGAP, RTS, and T-RTS scores were able to predict an Injury Severity Score >15 (AUC(ROC): 0.75, 0.75, and 0.74, respectively), the need for a stay in ICU >2 days or death (AUC(ROC) of 0.85, 0.83, and 0.83, respectively), and the massive hemorrhage (AUC(ROC): 0.70, 0.72, and 0.73, respectively). In contrast, MGAP, RTS, T-RTS, and TRISS scores were not predictors of the need of an emergency procedure (AUC(ROC): 0.53, 0.51, and 0.52, respectively). Four independent predictors of emergency procedure were noted: penetrating trauma, intravenous colloid administration >750 mL, systolic arterial blood pressure <100 mm Hg, and heart rate >100 bpm. Prehospital triage scores were predictors of Injury Severity Score >15, prolonged ICU stay, and massive hemorrhage but not of emergency procedure requirement.