44 Chapter 3 including baseline scores of the corresponding outcome as covariate. Estimated marginal means and corresponding standard errors from the primary LMMs will be used to additionally calculate between-group effect sizes at each timepoint. Effect sizes will be expressed as Cohen’s d with 95% confidence intervals. In addition, a clinically significant change will be examined using the reliable change index (Jacobson and Truax, 1991). Moderation and mediation To get more insight in for whom the MBSR program works best, we will test moderating effects of school weight, age, past or present psychological problems and teachers’ years of experience by testing interactions with condition. In addition, we will test possible mediating variables between condition and the outcomes. The analyses regarding potential moderators and mediators are exploratory since the power for the trial is calculated to establish evidence for main effects on the primary outcome variables. For the mediation analysis, we will follow the recommendations of Preacher and Hayes for mediation analyses (Hayes and Rockwood, 2017; Preacher and Hayes, 2008). In the mediation models, X will be the categorical group variable (0=control, 1= intervention), Y will be the observed scores of primary or secondary continuous outcomes at posttest or follow-up and M will be the potential mediators, including self-compassion, mindfulness-skills and emotion regulation skills. Total, direct and (specific) indirect effects will be calculated for all models. To decide whether effects are significant, corresponding 95% bias-corrected and accelerated confidence intervals will be calculated based on 5000 bootstrap samples (Hayes and Rockwood, 2017; Hayes, 2017). Separate simple mediation analyses will be run for each putative mediator. In addition, multiple mediation analyses will be conducted with all potential mediators included. Data collection As soon as the participant is enrolled, across all assessments he or she will only be identifiable via a unique pseudocode identifier to anonymize all data. A separate protected database will link the unique pseudocode to the participants’ names. Anonymous and non-anonymous (e.g. informed consent forms) data will be stored in separate password protected folders. Questionnaire data will be collected and stored with the online electronic data capture software CASTOR EDC (Ciwit bv, 2016), which tracks and logs any manual changes made to raw data. When questionnaires are not completed within the expected period, teachers will be sent online reminders. If no response follows, they will be contacted per telephone to discuss the absence of response and to motivate them to complete the questionnaires. In case participation is discontinued, reasons will be noted. Also, the number of sessions attended by each participant will be noted and sessions will be recorded. An overview of all measurements is given in Table 1.
RkJQdWJsaXNoZXIy MTk4NDMw