Decompressive craniectomy in traumatic brain injury: the intensivist's point of view.

Abstract

To perform a score with early clinical and radiological findings after a TBI that identifies the patients who in their subsequent evolution are going to undergo DC. Observational study of a retrospective cohort of patients who, after a TBI, enter the Neurocritical Section of the Intensive Care Unit of our hospital for a period of 5 years (2014-2018). Detection of clinical and radiological criteria and generation of all possible models with significant, clinically relevant and easy to detect early variables. Selection of the one with the lowest Bayesian Information Criterion and Akaike Information Criterion values for the creation of the score. Calibration and internal validation of the score using the Hosmer-Lemeshow and a bootstrapping analysis with 1,000 re-samples respectively. 37 DC were performed in 153 patients who were admitted after a TBI. The resulting final model included Cerebral Midline Deviation, GCS and Ventricular Collapse with an Area under ROC Curve: 0.84 (95% IC 0.78-0.91) and Hosmer-Lemeshow p=0.71. The developed score detected well those patients who were going to need an early DC (first 24hours) after a TBI (2.5±0.5) but not those who would need it in a later stage of their disease (1.7±0.8). However, it seems to advice us about the patients who, although not requiring an early DC are likely to need it later in their evolution (DC after 24hours vs do not require DC, 1.7±0.8 vs 1±0.7; p=0.002). We have developed a prognostic score using early clinical-radiological criteria that, in our environment, detects with good sensitivity and specificity those patients who, after a TBI, will require a DC.