High-Frequency Physiological Measures Predict Post-Admission Surgical Intervention After Severe Traumatic Brain Injury

Abstract

After a traumatic brain injury, around 12% of patients require surgical interventions during their index hospitalization due to delayed or progressive intracranial hemorrhage or complications such as elevated intracranial pressure (ICP)¹. Compiling data from four harmonized studies with 288 patients that have high-frequency physiological measurements, including ICP, we aimed to determine factors associated with those surgeries and whether longitudinal physiological measurements could be used to predict the need for craniectomy or craniotomy at least 1 h before the surgery occurred. The outcome was the occurrence of the first cranial surgery 6–120 h post-injury with 2:1 matched controls for those without surgery. Covariates included baseline characteristics and dynamic physiological measurements. Univariate associations were assessed, and the area under the receiving operating characteristic curve (AUC) was used to compare various machine learning and multivariable statistical models for the prediction of surgery. It was found that means, medians, and transgressions of both ICP and mean arterial pressure, as well as the linear regression slope of ICP by time in the 6 h prior to surgery, were significantly and independently related to whether a patient had cranial surgery or not. The best-performing model was found using random forests supervised learning algorithm (AUC = 0.75, 95% confidence interval 0.61–0.88). This model may assist clinicians in predicting when they may need to perform an emergent neurosurgical procedure, thus preventing more damage from elevated ICPs.