Predictive performance and statistical calibration of the TRISS method in patients with traumatic brain injury: a retrospective data analysis

Abstract

Introduction: Major trauma is a leading cause of death, particularly in young adults. The Trauma and Injury Severity Score (TRISS) is commonly used for mortality prediction, but its performance in patients with head trauma, especially in non-Western populations, remains unclear. The applicability of the TRISS, validated on North American cohorts, in different healthcare settings and injury patterns, such as those in North Africa, has not been thoroughly assessed. This study evaluated the predictive performance of the TRISS for hospital mortality in a cohort of Moroccan patients with traumatic brain injury. (TBI). Methods: A single-center retrospective study was conducted at Hassan II University Hospital in Fez, Morocco, over 24 months (January 2022 to December 2023), including 133 patients. The TRISS score was calculated using the Revised Trauma Score (RTS), Injury Severity Score (ISS), and age. The primary endpoint was all-cause mortality, and model performance was assessed through ROC curve analysis and calibration using graphical methods. Owing to the modest sample size, formal calibration tests were not emphasized. Results: The mortality rate was 27.8%. The TRISS demonstrated modest discriminative ability (AUC = 0.654). Although the specificity was high, the sensitivity was very low, resulting in a high false-negative rate (83.8%). Calibration analysis showed paradoxical mortality in the highest predicted survival group, indicating a significant miscalibration. Conclusions: The TRISS method displayed limited predictive performance in the TBI cohort, largely owing to its low sensitivity. This underscores the need for population-specific validation and development of more accurate predictive models to improve patient triage and outcomes.