Prediction of tracheostomy in critically ill trauma patients: a systematic review.

Abstract

Tracheostomy is relatively common in mechanically ventilated patients in the intensive care unit (ICU). The prediction of which patients will receive a tracheostomy is crucial to both clinical decision making and the design of targeted interventional trials of its timing. We aimed to systematically review the literature to ascertain whether useful predictors of eventual tracheostomy can be identified, with a particular focus on trauma patients. We searched three electronic databases to identify all studies of any design evaluating potential predictors of tracheostomy in mechanically ventilated ICU patients. Bias was assessed using the Quality in Prognosis Studies tool. Of 140 potentially eligible studies, we identified 12 relevant observational studies recruiting a total of 119 945 mechanically ventilated patients, of whom 14 080 (11.7%) received a tracheostomy. Seven studies were performed in trauma populations and included 24 858 patients, of whom 6140 (24.7%) received a tracheostomy. Factors predictive of receiving a tracheostomy in the trauma population included patient factors (age and comorbidities), diagnostic factors (injury type and injury severity score), and intervention factors (craniotomy or laparotomy). Profound clinical and methodological heterogeneity prevented meaningful metaanalysis. Significantly, more predictors were present on the day of admission in trauma populations than in non-trauma patients with brain injury and in other populations (89.7% v 73.3% v 25.0%). There are a number of clinical factors associated with subsequent tracheostomy in mechanically ventilated patients, in particular trauma patients. Given the need to prevent patients from receiving an unnecessary tracheostomy, these findings indicate that better predictive models are needed before the conduct of interventional trials. PROSPERO registry no. CRD42018084987.