Abstract
Background: Digital mental health interventions (DMHIs) are widely used in workplaces to address common mental health conditions such as depression, anxiety, and stress. Their content and design vary widely, and meta-analyses repeatedly found substantial heterogeneity in their efficacy, but there is no systematic evaluation of the comparative effect of content on efficacy.
Objective: We aimed to identify whether intervention design, therapeutic approaches, and content features are associated with greater efficacy in randomized controlled trials (RCTs) of DMHIs in workplaces.
Methods: We conducted a systematic review of RCTs evaluating DMHIs using standardized measures of depression, anxiety, or stress in employed adults in 4 electronic databases (MEDLINE, PsycINFO, CENTRAL, and Embase) from 2004 to April 2024. A Bayesian multilevel meta-regression was used to estimate the pooled effect size (θ) and compare intervention characteristics (therapeutic approach, design, and content features) with sample demographics in explaining heterogeneity for each outcome. The influence of intervention characteristics on DMHI effectiveness was evaluated using posterior probabilities (pp) and evidence ratios (ERs), with ER>3 indicating likely therapeutic benefit. Risk of bias was assessed with the Cochrane Risk of Bias tool (version 2).
Results: We included 81 RCTs evaluating 98 different DMHIs in 25,500 participants. Small but significant pooled effect sizes were identified for all 3 outcomes: depression (θ=0.167, 95% credible interval [CI] –0.31 to –0.03; ER 35.4; pp=0.972), anxiety (θ=–0.211, 95% CI –0.36 to –0.07; ER 113; pp=0.991), and stress (θ=–0.165, 95% CI –0.28 to –0.05; ER 199; pp=0.995), with intervention characteristics explaining more outcome heterogeneity than sample characteristics. In total, 71 (82.70%) trials had high risk bias, 4 (4.60%) had some concerns, and 11 (12.70%) had low risk bias, largely due to waitlist controls and inability to blind participants. Interventions using mindfulness and stress management approaches were more effective than cognitive behavioral therapy for anxiety and stress outcomes. Intervention designs incorporating person support (ERs 3.9-10.6) and expert design (stress: ER 25.7) enhanced therapeutic benefit for all outcomes. There was greater therapeutic benefit in interventions with video content (ER 3.69-5.71; pp=0.79-0.85) for all outcomes and those providing feedback scores (ER 6.55; pp=0.87), and reminder texts (ER 96.56; pp=0.99) showed moderate to strong benefit for stress outcomes.
Conclusions: The effectiveness of workplace DMHIs in RCTs is influenced by their therapeutic approach, design, and content features. These influences can vary with the outcome targeted. Mindfulness and stress management approaches, expert involvement in design, and interventions blended with person support showed evidence of greater therapeutic benefit, as did some types of content. These findings inform a more deliberate, evidence-informed approach to addressing the lack of progress in improving DMHI efficacy in workplaces and highlight the limitations of assumptions of benefit from all features or participatory design processes without evaluation.
DOI: 10.2196/71253