Study protocol 77 Randomization will be stratified by team. To ensure the right distribution of clients per team in the two groups, envelopes containing five lots per team will be used each time. For this we will use a randomization plan from http://www.randomization.com. After completing the baseline interviews, the principal investigator will allocate participants to the experimental group or control group. Statistical analysis The analyses will be based on the intention-to-treat principle. Generalized linear mixed models will be used to investigate group difference between the experimental and control conditions (52, 53), with times of measurement nested within study participants. As well as the main effects of time and condition, we will investigate whether there is a time x condition interaction effect with respect to symptoms, health complaints, number of relapses and recovery. Because number of relapses is a count variable, the analysis will use a mixed model for the poisson or negative binomial distribution. We will also test whether the addition of a random effect for the time variable results in a better model fit in each of the analyses. For the recovery outcome, we will investigate the influence of illness management, symptoms, relapses and progress on personal goals, as well as their interaction with time. Akaike’s Information Criterion (AIC (54)) will be used for statistical model selection in these analyses. To explore associations between illness-management outcomes and recovery outcomes, we will use structural equation modelling. To investigate the associations between improvement of symptoms, recovery and fidelity (within IMR + CAU), linear mixed models will be used, with clients nested within IMR groups. As well as a fixed effect for the trainer’s fidelity score, a random intercept at the group level will be incorporated. Before testing the first and second primary hypotheses, we will test the fourth secondary hypothesis: ‘we expect that any improvement resulting from IMR + CAU will be associated with the fidelity with which IMR is implemented.’ If fidelity has an effect, it will be included as a covariate in the subsequent analyses. Clients will be classified as completers of the study if they have finished all three interviews. After the intention-to-treat analyses, we will perform secondary analyses to measure the effects in clients who have completed IMR. Ethical considerations The study protocol, information brochure and informed consent form were consistent with the declaration of Helsinki and approved by the Dutch Union of
RkJQdWJsaXNoZXIy MjY0ODMw