Digital mental health interventions for the treatment of depression: A multiverse meta-analysis.

Background: The varying sizes of effects in published meta-analyses on digital interventions for depression prompt questions about their efficacy.

Methods: A systematic search in Embase, PsycINFO, and PubMed identified 125 randomised controlled trials up to February 2023, comparing digital interventions for depression against inactive controls. The stability of results was evaluated with a multiverse meta-analysis, thousands of meta-analyses were conducted based on different combinations of analytical choices, like target populations, intervention characteristics, and study designs.

Results: A total of 3638 meta-analyses were performed based on 125 randomised controlled trials and 263 effect sizes, with a total of 32,733 participants. The average effect size was Hedges' g = 0.43, remaining positive at both the 10th (g = 0.16) and 90th percentiles (g = 0.74). Most meta-analyses indicated a statistically significant benefit of digital interventions. Larger effects were observed in meta-analyses focusing on adults, low- and middle-income countries, guided interventions, comparing interventions with waitlist controls, and patients with major depressive or unipolar mood disorders. Smaller effects appeared when adjusting for publication bias and in assessments after 24 weeks.

Limitations: While multiverse meta-analysis aims to exhaustively investigate various analytical decisions, some subjectivity remains due to the necessity of making choices that affect the methodology. Additionally, the quality of the included primary studies was low.

Conclusions: The analytical decisions made during performing pairwise meta-analyses result in vibrations from small to medium effect sizes. Our study provides robust evidence for the effectiveness of digital interventions for depression while highlighting important factors associated with treatment outcomes.