Fig 3.

Variability in the use of high ventilatory frequencies over time and across individual providers. (a) Changes in the proportion of cases in each ventilatory frequency group (represented by the respective group median) and the predicted probability of postoperative respiratory complications (PRC) are plotted against time (year of surgery) over a 10-yr period. The fraction of cases ventilated with higher ventilatory frequencies increased over time. Despite this change in practice, the rate of postoperative respiratory complications remained unchanged over time. Predicted probabilities for the outcome of PRC were derived from the primary regression model; error bars indicate 95% confidence intervals. (b) The predicted preference for an individual anaesthesia provider to use high instead of low ventilatory frequency (cut-off at median) ranged from 22.2% to 88.0%, indicating a high variability across providers. Individual provider preference for using high ventilatory frequencies was derived from a mixed-effects logistic regression model with ventilatory frequency as the outcome, individual anaesthesia providers as the random effect and all covariates of the primary analysis as fixed effects. Data are expressed as mean (standard deviation, sd) of the adjusted preference of using high ventilatory frequencies across all cases performed by an individual provider.