Association Between Hospital-Level Volume of Extracorporeal Membrane Oxygenation and Complications: A Clustered, Mixed-Effect Analysis of the Extracorporeal Life Support Organization Registry

Objective: Extracorporeal membrane oxygenation (ECMO) has been used to provide pulmonary and circulatory support in patients with ongoing critical illness. Studies investigating whether an association exists between ECMO center-volume and complications are lacking. Our main aim was to investigate whether an association between ECMO center-volume and the rate of ECMO-related complications for adult patients exists. Design: Retrospective observational study. Setting: Patients from the Extracorporeal Life Support Organization Registry were collected from 2018 to 2021. Subjects: Adult patients (18 yr old or older) were grouped according to ECMO first mode and support modality: venovenous (VV), venoarterial (VA) or extracorporeal cardiopulmonary resuscitation (ECPR). Interventions: None. Measurements and Main Results: The association of support type-specific hospital center-volume on rates of complications was investigated using multivariate, clustered mixed-effects logistic regression. From 2018 to 2021, a total of 9,427 patients underwent VV ECMO across 281 centers, 10,794 patients underwent VA across 275 centers, and 3,595 patients underwent ECPR across 224 centers. Average annual center-volume was 74 (± 55), 80 (± 55), and 89 (± 58) for VV, VA, and ECPR, respectively. Complication rates varied widely across subgroups. Increased annual center-volume was associated with increased mechanical complications (VA: adjusted odds ratio (aOR) = 1.004 1.001–1.008), and decreased hemorrhagic complications (VV: aOR = 0.997 0.994–0.999; VA: aOR = 0.997 0.994–0.999) and pulmonary complications (VV: aOR = 0.997 0.994–1.000). Metabolic complications were not significantly associated with center volume. Conclusions: Higher annual center-volumes of ECMO were associated with decreased pulmonary complications for VV, decreased hemorrhagic complications for VA and VV cohorts and an increase in mechanical complications for VA cohort but not ECPR alone. These associations were small, with a 0.3% increase or decrease in odds per complication category, but are attributable entirely to the center volume and known at the moment of ECMO cannulation.