Understanding how digital mental health interventions can be optimised to improve longer term sustainability: Findings from a causal mediation analysis of the CONEMO trials

Two CONEMO trials in Lima, Peru and São Paulo, Brazil evaluated a digital mental health intervention (DMHI) based on behavioural activation (BA), that demonstrated improved symptoms of depression at three, but not six-months. To understand how we can optimize features of the DMHI to improve outcomes for longer-term effectiveness and eventually scale-up, we investigated mediators through which the intervention improved symptoms of depression, separately for the two trials and then using a pooled dataset. Trial data included adults with depression (Patient Health Questionnaire – 9 (PHQ-9) score ≥10) and comorbid hypertension and/or diabetes. Interventional effects, a robust causal inference based approach to mediation, was used to decompose the total effect of DMHI on improved symptoms of depression (PHQ-9 scores improving by at least 50% between baseline and six months (yes/no)) into indirect effects via: understanding session content without difficulty; number of activities completed to improve levels of activation; and levels of BA captured using the Behavioural Activation for Depression Short-Form. Understanding the content of the sessions without difficulty mediated a 10% [0.10: Bias corrected 95% CI: 0.03 to 0.15] improvement in depression symptoms; completing activities mediated a 12% improvement [0.12: 0.01 to 0.23]; BA mediated a 2% [0.02: 0.01, 0.05] improvement. Our findings suggest that DMHI based on BA should consider developing strategies to help participants complete activities they find enjoyable to improve levels of activation and maintain the effects of the intervention in the longer-term. To improve longer term outcomes, DMHI should also ensure that sessions are tailored to the characteristics of the intended population, so it is simple for patients to understand and therefore they are able to complete activities.