Fig. 2

Graphical representation of the mixed effects regression analysis on the pooled dataset of the two impact studies. In both studies, a prediction model for postoperative nausea and vomiting (PONV) was implemented (red colors) and compared to care-as-usual (blue colors). In one study, the probabilities of the model were simply presented to physicians who implemented without a recommendation (i.e. an assistive format; less saturated colors). The other study also included an actionable recommendation (i.e. a directive format; more saturated colors). Both studies compared the impact on the physicians’ administration of antiemetic prophylaxis (a) and on the incidence of PONV (b). The bars and their 95% confidence intervals (CI) represent the fixed effects of the mixed effects regression analyses. The mixed effects models included fixed effects for the following variables: study, allocation group, predicted probability of PONV, and all interaction term between these variables. Because of the similarity of the results, the bars were calculated from the unadjusted analysis after multiple imputations. The 95% CIs were calculated from the covariance matrix for the variable study, allocation group, predicted probability, and their interaction terms. Further methodologic information and the numerical results of the regression models are available in Additional file 1