Invasive Mechanical Ventilation for Acute Viral Bronchiolitis: Retrospective Multicenter Cohort Study.

OBJECTIVES:Bronchiolitis is a leading cause of PICU admission and a major contributor to resource utilization during the winter season. Management in mechanically ventilated patients with bronchiolitis is not standardized. We aimed to assess whether variations exist in management between the centers and then to assess if differences in PICU outcomes are found. DESIGN:Retrospective cohort study. SETTING:Three tertiary PICUs (Centers A, B, and C) in London, United Kingdom. PATIENTS:Patients under 1 year of age (n = 462) who received invasive mechanical ventilation for acute viral bronchiolitis from 2012-2016. INTERVENTIONS:None. DESIGN:Retrospective cohort study. MEASUREMENTS AND MAIN RESULTS:Data collected include all sedative agents administered, 48 hour cumulative fluid balance and location of endotracheal tube (oral or nasal). Primary outcome was duration of invasive mechanical ventilation. A generalized linear model was used to test for differences in duration of invasive mechanical ventilation between centers after adjustment for confounders: corrected gestational age, oxygen saturation index, bacterial coinfection, prematurity, respiratory syncytial virus status, risk of mortality score and comorbidity. Baseline characteristics were similar, other than a higher risk of mortality score at center A and higher admission oxygen saturation index at center C. Center A was associated with utilization of the most benzodiazepine and opiate sedation, the fewest nasal endotracheal tubes, and the highest mean cumulative fluid balance at 48 hours.Center A had an adjusted mean duration of invasive mechanical ventilation that was 44% longer than center C (95% CI, 25-66%; p < 0.001).The majority of confounders had an association with the duration of invasive mechanical ventilation; all were biologically plausible. Corrected gestational age was negatively associated with the duration of invasive mechanical ventilation for preterm infants less than 32 weeks, but not for term or 32-37 week infants (interaction effect). This meant that at a corrected age of 0 months, a less than 32-week infant had a mean duration that was 55% greater than a term infant: this effect had disappeared by 8 months old. CONCLUSIONS:Between-center variations exist in both practices and outcomes. The relationship between these two findings could be further tested through implementation science with "optimal care bundles."