Outcomes of traumatic brain injury: the prognostic accuracy of various scores and models

Introduction. Traumatic Brain Injury (TBI) is a worldwide health problem, and is a pathology that causes significant mortality and disability in Latin America. Different scores and prognostic models have been developed in order to predict the neurological outcomes of patients. We aimed to test the prognostic accuracy of the Marshall CT classification system, the Rotterdam CT scoring system, and the IMPACT and CRASH models, in predicting 6-month mortality and 6-month unfavourable outcomes in a cohort of trauma patients with TBI in a university hospital in Colombia. Methods. We analysed 309 patients with significant TBI who were treated in a regional trauma centre in Colombia over a two year period. Bivariate and multivariate analyses were undertaken. The discriminatory power of each model, as well as its accuracy and precision, were assessed by logistic regression and AUC. Shapiro Wilks, chi2 and Wilcoxon test were used to compare the actual outcomes in the cohort against the predicted outcomes. Results. The median age was 32 years, and 77.67% were male. All four prognostic models showed good accuracy in predicting outcomes. The IMPACT model had the greatest accuracy in predicting an unfavourable outcome (AUC 0.864; 95% CI 0.819 - 0.909) and in predicting mortality (AUC 0.902; 95% CI 0.862 - 0.943) in patients with TBI. Conclusion. All four prognostic models are applicable to eligible TBI patients in Colombia. The IMPACT model was shown to be more accurate than the other prognostic models, and had a higher sensitivity in predicting 6-month mortality and 6-month unfavourable outcomes in patients with TBI in a university hospital in Colombia