Citation

Abdullah JM, Idris Z, Ghani AR, Lim MS. J. Neurosurg. Sci. 2021; ePub(ePub): ePub.

Copyright

(Copyright © 2021, Minerva Medica)

DOI

10.23736/S0390-5616.21.05249-8

PMID

unavailable

Abstract

BACKGROUND: Traumatic brain injury (TBI) has recently become a major concern for public health care and a socioeconomic burden internationally. Prognostic models are mathematical models developed from specific populations which are used to predict the mortality and unfavourable outcomes especially in trauma centres. Hence, we formulate a study to perform an external validation of the IMPACT and CRASH prognostic models; the CRASH model to predict 14-day mortality and 6-month unfavourable outcome and the IMPACT model to estimate 6-month mortality and unfavourable outcome in a single centre cohort of TBI patients in Malaysia.

METHODS: All patients with traumatic brain injury (mild, moderate and severe) who were admitted to Hospital Queen Elizabeth from 1st November 2017 to 31st January 2019, were prospectively analysed through a data collection sheet. The discriminatory power of the models was assessed as area under the receiver operating characteristic curve and calibration was assessed using the Hosmer-Lemeshow (H-L) goodness-of-fit test and Cox calibration regression analysis.

RESULTS: We analyzed 281 patients with significant TBI treated in a single neurosurgical center in Malaysia over a 2-year period. The overall observed 14-day mortality was 9.6%, a 6-month unfavourable outcome of 23.5%, and a 6-month mortality of 13.2%. Overall, both the CRASH and IMPACT models showed good discrimination with AUCs ranging from 0.88 to 0.94 and both models calibrating satisfactorily H-L GoF p>0.05 and calibration slopes >1.0 although IMPACT seemed to be slightly more superior compared to the CRASH model.

CONCLUSIONS: The CRASH and IMPACT prognostic models displayed satisfactory overall performance in our cohort of TBI patients, but further investigations on factors contributing to TBI outcomes and continuous updating on both models remain crucial.

Language: en