Internal Validation of Two Models Developed for Prognostication of Patients with Isolated Traumatic Brain Injury.

Abstract

To evaluate the efficiency of two models for prognostication of patients with isolated traumatic brain injury. The models developed with the data of the patients who applied within ten years were subjected to internal validation with the data of the patients who applied within the following five years. The records of 204 patients with traumatic brain injury admitted into Neurosurgery Department and Intensive Care Units were reviewed. Models were applied to procure estimates of prognosis. The estimates were statistically compared with the actual clinical outcome of patients using discriminant analysis. For Model 1, the correct classification rate was calculated as 87.9%, the specificity as 66.7%, the sensitivity as 94.2%, the positive predictive value as 68.8%, and the negative predictive value as 93.6%. For Model 2 the correct classification rate was evaluated as 90.2%, the specificity as 57.6%, the sensitivity as 96.5%, the positive predictive value as 76%, and the negative predictive value as 92.2%. Both of the models had decent correct classification rates and may be efficient estimation tools for the prognostication of unfavourable outcome in patients with isolated traumatic brain injury. These models are good candidates to be used widely following the evaluation of their validity with national and international multicentric studies.