57 Review of social network intervention studies 2 functioning as well as mental health treatment adherence. Larger effects on general functioning were found at follow-up (versus post-assessment), while effects on mental health treatment adherencewere found tobe smaller at followup (versuspost-assessment). Furthermore, only the outcome subcategory of abstinence moderated the effect size of substance use, indicating that larger effects of social network interventions were found on abstinence than on substance use (i.e., quantity and frequency of substance use). Publication bias Publication bias was examined through multiple methods generally providing no evidence for publication bias and selective outcome reporting bias on main outcomes, see Appendix A, Table 5. Only Egger’s test revealed an indication for publication bias of studies on social network outcomes (p = 0.004). In addition, trim-and-fill analyses showed that six effect sizes were imputed on the right side of the funnel plot (i.e., larger/positive effect sizes), indicating that mean effect sizes could be an underestimation of the effect on mental health treatment adherence. Noteworthy, for substance use, nine effect sizes were imputed on the left side (i.e., negative effect sizes) of the plot, indicating that mean effect sizes could be an overestimation of the effect. Finally, sunset funnel plots show that the median statistical power of the included effect sizes (k = 553) is low (medpower = 35%), and moderate true effect sizes (d33% = 0.3) are needed to reach this median power level (Appendix A, Figure 8.1 and 8.2) (Kossmeier, Tran, & Voracek, 2020).
RkJQdWJsaXNoZXIy MjY0ODMw