Predicted unfavorable neurological outcome is overestimated by the Marshall CT score, CRASH and IMPACT models in Severe TBI Patients management with early decompressive craniectomy

5 Introduction: Traumatic Brain Injury (TBI) is of public health in terest and produces significant mortality 6 and disability in Colombia. Calculators and prognos tic models have been developed in order to establis h 7 neurologic outcome. We tested prognostic models (th e Marshall CT score, IMPACT, and CRASH) for 8 14-day mortality, 6-month mortality, and 6-month ou tcome in TBI patients in a university Hospital in 9 Colombia. 10 Methods: A 127 patients’ cohort with TBI was treated in a re gional trauma center in Colombia over 2 11 years and Bivariate and Multivariate analyses were us d. Discriminatory power of the models, its 12 accuracy and precision was assessed by both logisti c regression and area under the receiver operating 13 characteristic curve (AUC). Shapiro Wilks, chi2 and Wilcoxon test were used to compare real outcomes 14 in the cohort against predicted outcomes 15 Results: Group’s median age was 33 years and 84, 25% was mal e. The ISS Median was 25 and median 16 GSC motor score was 3. Six-month mortality was 29.1 3%. Six-month unfavorable outcome was 37%. 17 Mortality prediction by Marshall CT score was 52.8% p=0.104 (AUC 0,585; 95% IC 0 0,489-0,681), the 18 Mortality prediction by CRASH prognosis calculator was 59.9% p<0.001 (AUC 0,706; 95% IC 0,59019 0,821), the unfavorable outcome prediction by IMPAC T was 77% p<0.048 (AUC 0,670; 95% IC 0,57520 0,763). 21 Conclusion: The Marshall CT score, IMPACT and CRASH models over estimate the adverse 22 neurological outcome in patients with severe head t r uma in a university hospital in Colombia. 23