Relationship between facility-level ARS utilization, diagnostic test use, and GERD prevalence. The size of the points corresponds to the number of infants discharged from each site. There was no independent or joint association between measures, as tested by ordinary least squares (P > .1 in all models).
Variation in adjusted and unadjusted ARS rates. Hollow and filled symbols show, respectively, actual and predicted probabilities of ARS from a mixed effects logistic regression model with random intercept for hospitals. Adjusted predictions are based on population-averaged values in which only the random effect (the hospital) varies, and fixed effects are constant.