TY - JOUR PY - 2020// TI - Comparison of the CRASH-predicted and real outcome of TBI in a retrospective analysis of 417 patients JO - World neurosurgery A1 - Dullaert, Matthias A1 - Oerlemans, Joyce A1 - Hallaert, Giorgio A1 - De Paepe, Peter A1 - Kalala Okito, Jean-Pierre SP - ePub EP - ePub VL - ePub IS - ePub N2 - BACKGROUND: The goal of this retrospective study is to externally validate the CRASH clinical calculator as a prognostic tool. Mortality at 14 days and an unfavourable outcome (i.e. a GOS score of less than 4) after six months were the primary endpoints. MATERIAL & METHODS: Adults patients admitted to the ER of Ghent University Hospital during 2010-2014 because of traumatic brain injury (TBI) were included in the study. The CRASH-score was calculated for every patient and then compared to real-life outcome at 14 days and 6 months. Researchers were blinded for each other's results and each observer either calculated the CRASH-score or obtained clinical outcome. Receiver operating characteristic (ROC)-analysis was used to validate the CRASH calculator. The prognostic value of other variables was tested using logistic regression (p<0.05 = significant).

RESULTS: A total of 417 patients were included. 94,7% of patients with TBI were still alive at 14 days while 65% had a GOS ≥4 at six months. ROC-analysis showed an area-under-the-curve (AUC) of 92,1% at 14 days and 90,7% at 6 months (p<0.05). The calculated cut-off value for the score at 14 days was 31,50% (sensitivity: 0.823; specificity: 0.895). At six months, the calculated cut-off value was 55,75% (sensitivity: 0.793; specificity: 0.830).

CONCLUSION: The CRASH calculator is a good predictor of outcome in TBI at 14 days and six months with a high sensitivity and specificity. However, it does not replace clinical judgement of the physician treating the patient in the ER, but constitutes a useful additional tool.

Copyright © 2020 Elsevier Inc. All rights reserved.

Language: en

LA - en SN - 1878-8750 UR - http://dx.doi.org/10.1016/j.wneu.2020.01.118 ID - ref1 ER -