Identifying what works in mental health apps through meta-regression analyses of 169 trials

Abstract This meta-analysis aimed to code active cognitive behavioral elements in mental health apps and to examine the association between these elements and improvements in depression and anxiety. Trials evaluating mental health apps were coded based on 34 pre-registered elements. 169 trials with 1137 timepoints were included (N = 41,807; mean age = 34.3 years; 72.9% female). Psychoeducation, relaxation, mindfulness, and self-monitoring were used most frequently. Bivariate mixed-effect meta-regression models showed that many elements were weakly to moderately effective. Desensitization, stimulus control, and activity scheduling were most strongly and robustly associated with improvements in depression and exposure-based elements with improvements in anxiety. Ineffective elements included graded tasks and personal strengths, but in sum, there was considerable variation in the frequency and impact of active elements. Interventions incorporating a greater number of elements were more effective. This meta-analysis provides insight into how active elements in mental health apps are associated with therapeutic change, informing future interventions.