Chapter 7 174 independence (competence), independence (performance), and employment (10, 36, 37). Statistical analysis Following our research question, we hypothesized that the improvement of overall illness management had direct and indirect pathways via the improvement in clinical and functional recovery, to the improvement in personal recovery. Mediation analysis is a useful statistical method to examine inferentially such direct and indirect relations (38, 39). As recommended, our mediation analysis was based on theory, previous research, and logical argument (38). In this mediation analysis, pathways were described between the improvement in illness (self-) management (ΔIM) and the improvement in personal recovery (ΔPR), mediated by the improvement in clinical (ΔCR) and functional recovery (ΔFR) (Fig 1). Difference scores were constructed for all concepts by subtracting mean scores measured at T1 from mean scores measured at T3 (39). To avoid regression to the mean, we applied the baseline data of all concepts as covariates (39-41). This corrected the correlation between the difference score and the baseline measurement. The regression-equation for the ΔCR mediator consisted of an intercept, ΔIM, baseline IM, and baseline CR; for the ΔFR mediator the regressionequation consisted of an intercept, ΔIM, baseline IM, and baseline FR; for the ΔPR the regression-equation consisted of an intercept, ΔIM, ΔCR, ΔFR, baseline IM, baseline CR, baseline FR, and baseline PR. To facilitate comparability, standardized regression coefficients were used. The direct path consisted of a regression equation from ΔIM to ΔPR (c’). The two indirect paths consisted of regression equations from ΔIM to the mediators ΔCR (a1) and ΔFR (a2) respectively, and regression equations from those mediators to ΔPR (corrected for ΔIM to ΔPR) (b1 and b2) (Fig 1). As inferential tests to determine the significance of the indirect paths, confidence intervals were constructed by multiplying the regression weights of ΔIM in the first regression and of the mediators in the second. This was executed on 5000 bootstrap samples of our dataset, applying bias correction. For this purpose, the PROCESS macro (39) was used in SPSS 27. In this type of mediation analysis, using the PROCESS macro, there can be only one predictor and only one outcome, but multiple mediators can be specified (38, 39).
RkJQdWJsaXNoZXIy MjY0ODMw