2 The effect of training and education at the outpatient clinic 33 model. Because patients within a physician might be more similar than patients from other physicians, (e.g. more satisfied, similar diseases) a random intercept for physician was added to the model. We hypothesized that the effect of the physicians training could be different for each physician due to differences in knowledge and experience, therefore we added a random slope for physicians training. Analyses were adjusted for patients’ age, gender, CCI, quality of life, and physicians’ gender and level (resident or specialist) based on previous literature (28–30). An interaction term between patient intervention and physician intervention was added to assess whether the effect of either intervention differed depending on the other intervention. The nonsignificant interaction term indicated that the effects of both interventions were independent. Therefore, the interaction term was subsequently removed from further analyses and, because the sample size calculated for the fixed hierarchical testing was not met, we additionally analysed the data as being a two-by-two factorial design. To evaluate preparedness of the physician, results of physicians’ questionnaires before and after training were compared and tested for statistical significance using Mann-Whitney U test. Patient appreciation of the conversation aid was described narratively. All analyses were performed using IBM SPSS Statistics 25.0.0.2 software. P values <0.05 were considered statistically significant. 2.6 Sample size calculation An a priori sample size calculation for the comparison of the main intervention group (physician trained, patient informed ) and reference group (physician not trained, patient uninformed) on the primary outcome was performed using the statistic program G*Power. In previous studies, patient satisfaction on an 11-point Likert scale (0 to 10) was found to be between 5 and 9, with standard deviations between 1.2 and 3.2 (18–21). Hence, we assumed the mean patient satisfaction score to be 7.0 (reference group) and 8.0 (intervention group) with a standard deviation of 2 (i.e. a Cohen’s effect size 0.5). To achieve a power of >80% with a (one-sided) alpha of 0.05, 51 patients per group were needed. To enable stratified analysis by gender and a loss to follow-up of 10%, the required sample size would be 232 patients.
RkJQdWJsaXNoZXIy MTk4NDMw