Inpatient Complications Predict Tracheostomy Better than Admission Variables After Traumatic Brain Injury

Abstract

Data regarding who will require tracheostomy are lacking which may limit investigations into therapeutic effects of early tracheostomy. We performed an observational study of adult traumatic brain injury (TBI) patients requiring intensive care unit (ICU) admission for ≥ 72h and mechanical ventilation for ≥ 24h between January 2014 and December 2014 at a level 1 trauma center. Patients who had life-sustaining measures withdrawn were excluded. Multivariable logistic regression analyses were used to assess admission and inpatient factors associated with receiving a tracheostomy and to develop predictive models. Inpatient complications prior to day 7 were used to standardize data collection for patients with and without tracheostomy. Patients who received tracheostomy prior to day 7 were excluded from analysis. In total, 209 patients (78% men, mean 48years old, median Glasgow Coma Scale score (GCS) 8) met study criteria with tracheostomy performed in 94 (45%). Admission predictors of tracheostomy included GCS, chest tube, Injury Severity Score, and Marshall score. Inpatient factors associated with tracheostomy included the requirement for an external ventricular drain (EVD), number of operations, inpatient dialysis, aspiration, GCS on day 5, and reintubation. Multiple logistic regression analysis demonstrated that the number of operation room trips (adjusted odds ratio [AOR], 1.75; 95% CI, 1.04-2.97; P = 0.036), reintubation (AOR, 8.45; 95% CI, 1.91-37.44; P = .005), and placement of an EVD (AOR, 3.48; 95% CI, 1.27-9.58; P = .016) were independently associated with patients undergoing tracheostomy. Higher GCS on hospital day 5 (AOR, 0.52; 95% CI, 0.40-0.68; P < 0.001) was protective against tracheostomy. A model of inpatient variables only had a stronger association with tracheostomy than one with admission variables only (ROC AUC 0.93 vs 0.72, P < 0.001) and did not benefit from the addition of admission variables (ROC AUC 0.93 vs 0.92, P = 0.78). Potentially modifiable inpatient factors have a stronger association with tracheostomy than do admission characteristics. Multicenter studies are needed to validate the results.