Indicators for tracheostomy in patients with traumatic brain injury.

Our objective was to develop criteria to identify patients with traumatic brain injury (TBI) who require a tracheostomy (TR). From January 1994 to May 2000 all TBI patients requiring intubation on presentation and who survived >7 days were identified from our trauma registry. Demographics, Glasgow Coma Score (GCS), Injury Severity Score (ISS), and ventilator days, ICU days, hospital days, need for TR, and development of pneumonia were statistically analyzed. Of 246 patients with TBI 211 without TR and 35 with TR were identified (mean time to TR 13.3+/-7.0 days). Logistic regression analysis identified presenting GCS < or =8, ISS > or =25, and ventilator days >7 as significant predictors for TR. Applying these three predictors to our population identified 48 patients (21 with TR, 18 without TR, and nine who died on the ventilator without TR) with a sensitivity of 60 per cent, a specificity of 87 per cent, a positive predictive value of 44 per cent, and a negative predictive value of 93 per cent. Patients with TR had lower presenting GCS and higher ventilator, ICU, and hospital days (P < 0.05). Pneumonia rates were similar. Time to neurologic recovery (GCS > or =9) was longer for the TR patients as compared with the patients without TR. We conclude that patients with TBI presenting with a GCS < or =8, an ISS > or =25, and ventilator days >7 are more likely to require TR. Performing TR late did not reduce pneumonia rates or ventilator, ICU, or hospital days. By identifying the at-risk population early TR could be performed in an attempt to decrease morbidity and length of stay.