External Validation of the CRASH and IMPACT Prognostic Models in Severe Traumatic Brain Injury

• DOI: 10.1089/neu.2013.3003
• Publication Date: 2014-02-25T20:29:18Z
• Authors: Julian Han; Nicolas K.K. King; Sam J. Neilson; Mihir P. Gandhi; Ivan Ng
• Keywords: Adult; Male; Glasgow Outcome Scale; Middle Aged; Models, Theoretical; Prognosis; Severity of Illness Index; 3. Good health; Cohort Studies; Young Adult; 03 medical and health sciences; 0302 clinical medicine; Brain Injuries; Humans; Female; Aged

An accurate prognostic model is extremely important in severe traumatic brain injury (TBI) for both patient management and research. Clinical prediction models must be validated internally externally before they are considered widely applicable. Our aim to independently validate two models, one developed by the Corticosteroid Randomization After Significant Head (CRASH) trial investigators, other from International Mission Prognosis Analysis of Trials Traumatic Brain Injury (IMPACT) group. We used a cohort 300 patients with TBI (Glasgow Coma Score [GCS] ≤8) consecutively admitted National Neuroscience Institute (NNI), Singapore, between February 2006 December 2009. The CRASH (base CT) predict 14 day mortality 6 month unfavorable outcome. IMPACT (core, extended, laboratory) estimate Validation was based on measures discrimination calibration. Discrimination assessed using area under receiving operating characteristic curve (AUC), calibration Hosmer-Lemeshow (H-L) goodness-of-fit test Cox regression analysis. In NNI database, overall observed 47.7%, outcome 71.0%. base all three gave an underestimate values our when Using CT model, predicted 46.6% approximated outcome, whereas overestimate at 74.8%. Overall, showed good discrimination, AUCs ranging 0.80 0.89, conclude that satisfactorily TBI.