Keywords: surgery, volume-outcomes, regression splines, propensity score weighting
Studies of associations between surgeon or hospital operative volumes and postoperative mortality often select volume thresholds to ensure an equal number of patients across categories. Such studies also rarely define volume as time-varying and seldom evaluate the interaction between surgeon and hospital volumes. We demonstrate the combined use of regression splines to identify volume thresholds and multi-group propensity score weighting to estimate the effects of both surgeon and hospital volumes on mortality. We studied infants who underwent surgery for esophageal atresia/tracheoesophageal fistula at US children’s hospitals in 2005-2013. We defined surgeon and hospital volumes as the number of patients receiving this surgery from the provider in the previous 365 days. We identified no volume thresholds in the relationship between mortality and either surgeon or hospital volume. Furthermore, mortality did not differ across four groups defined by the intersection of surgeon and hospital volumes dichotomized at their medians.