Prognostic indicators and outcome prediction model for severe traumatic brain injury

Although some predictive models for patient outcomes after severe traumatic brain injury have been proposed, a mathematical model with high predictive value has not been established. The purpose of the present study was to analyze the most important indicators of prognosis and to develop the best outcome prediction model.

One hundred eleven consecutive patients with a Glasgow Coma Scale score of <9 were examined and 14 factors were evaluated. Intracranial pressure and cerebral perfusion pressure were recorded at admission to the intensive care unit. The absence of the basal cisterns, presence of extensive subarachnoid hemorrhage, and degree of midline shift were evaluated by means of computed tomography within 24 hours after injury. Multivariate logistic regression analysis was used to identify independent risk factors for a poor prognosis and to develop the best prediction model.

The best model included the following variables: age (p < 0.01), light reflex (p = 0.01), extensive subarachnoid hemorrhage (p = 0.01), intracranial pressure (p = 0.04), and midline shift (p = 0.12). Positive predictive value of the model was 97.3%, negative predictive value was 87.1%, and overall predictive value was 94.2%. The area under the receiver operating characteristic curve was 0.977, and the p value for the Hosmer-Lemeshow goodness-of-fit was 0.866.

Our predictive model based on age, absence of light reflex, presence of extensive subarachnoid hemorrhage, intracranial pressure, and midline shift was shown to have high predictive value and will be useful for decision making, review of treatment, and family counseling in case of traumatic brain injury.