Outcome prediction in moderate and severe traumatic brain injury: a focus on computed tomography variables

Abstract

Background: With this study we aimed to design validated outcome prediction models in moderate and severe traumatic brain injury (TBI) using demographic, clinical, and radiological parameters.

Methods: Seven hundred consecutive moderate or severe TBI patients were included in this observational prospective cohort study. After inclusion, clinical data were collected, initial head computed tomography (CT) scans were rated, and at 6 months outcome was determined using the extended Glasgow Outcome Scale. Multivariate binary logistic regression analysis was applied to evaluate the association between potential predictors and three different outcome endpoints. The prognostic models that resulted were externally validated in a national Dutch TBI cohort.

Results: In line with previous literature we identified age, pupil responses, Glasgow Coma Scale score and the occurrence of a hypotensive episode post-injury as predictors. Furthermore, several CT characteristics were associated with outcome; the aspect of the ambient cisterns being the most powerful. After external validation using Receiver Operating Characteristic (ROC) analysis our prediction models demonstrated adequate discriminative values, quantified by the area under the ROC curve, of 0.86 for death versus survival and 0.83 for unfavorable versus favorable outcome. Discriminative power was less for unfavorable outcome in survivors: 0.69.

Conclusions: Outcome prediction in moderate and severe TBI might be improved using the models that were designed in this study. However, conventional demographic, clinical and CT variables proved insufficient to predict disability in surviving patients. The information that can be derived from our prediction rules is important for the selection and stratification of patients recruited into clinical TBI trials.