Variations in rates of tracheostomy in the critically ill trauma patient

OBJECTIVE: The utility of tracheostomy to expedite weaning and prevent complications in patients with acute respiratory failure is actively debated, with many physicians holding strong opinions regarding the value and timing of this intervention. We postulated that these opinions would be reflected in significant variation in tracheostomy rates across centers. Thus, we set out explore the extent and potential sources of this variation among injured patients cared for in trauma centers in the United States. DESIGN: This is a retrospective cohort study. We used stratification and hierarchical multivariate analysis to evaluate the effect of patient and institutional characteristics on tracheostomy rates and variance decomposition to determine the proportion of variance across institutions explained by patient characteristics. SETTING: Intensive care units within trauma centers participating in the National Trauma Databank. PATIENTS: Injured patients admitted over the years 2001-2003, age >/=16 yrs, with an Injury Severity Score >/=9 and a diagnosis of acute respiratory failure, excluding patients with burn injuries and those with a severe injury to the face or neck who might require tracheostomy for maintenance of an airway. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: There were 17,523 patients meeting inclusion criteria: 4,146 (24%) underwent tracheostomy. The mean tracheostomy rate across centers was 19.6 per 100 hospital admissions with a range of 0-59. This variation persisted after stratification by age, injury mechanism, and severity. Although several patient and injury characteristics were predictive of tracheostomy, there were no identifiable institutional characteristics associated with tracheostomy. Patient characteristics accounted for only 14% of the variance across centers. CONCLUSIONS: There is significant unexplained variation in the rates of tracheostomy in critically injured patients with acute respiratory failure. This variation might reflect preconceived notions of efficacy among physicians practicing in the absence of evidence to guide care. The variation provides evidence of equipoise and emphasizes the need for a well-conducted randomized controlled trial to evaluate the utility of this procedure.