Variable-dose support in an online mental health intervention: A randomized, controlled exploratory study

Abstract

<p>Digital health interventions are widely considered a highly promising solution to issues with access to evidence-based care. However, digital health interventions are limited by lower rates of engagement than are typically seen in traditional, face-to-face interventions. Despite the importance of engagement for intervention efficacy, engagement is rarely a primary outcome in intervention studies, and few studies have empirically tested intervention changes to improve engagement. Hazel was a two-arm, randomized controlled exploratory study designed to investigate the impact of intervention design on engagement as a primary outcome. We conducted a trial of a mental health intervention, based on the Unified Protocol, delivered entirely online. Participants were randomized to complete the 12-week intervention self-guided (the unsupported arm) or with 4 weeks of therapist support (the supported arm). We sought to measure engagement as comprehensively as possible and therefore collected numerous self-report, behavioral, and objective measures of multiple facets of engagement, including how participants felt about the intervention, how they used the intervention technology, and how their behavior changed over the course of the intervention. We collected engagement outcomes at baseline, weekly during the intervention, and immediately following completion of the intervention. We hypothesized that a low dose of therapist support would improve engagement and subsequent mental health outcomes while being more scalable and feasible to implement than offering therapist support during all intervention weeks. The aims of this small, exploratory study were to learn more about patterns of engagement with the intervention and identify issues with the current intervention and trial design. We used descriptive statistics and visualizations to understand and describe trends in the data rather than p-value significance testing, in line with best practices for pilot studies. We recruited 23 North Carolina-based adults with clinically elevated depression and/or anxiety symptoms. Overall, the intervention and trial design appeared acceptable to participants. Participants in both arms had relatively positive attitudes toward the intervention at all time points, and there was no evidence that the response burden of our outcome measures was too high. However, several of our measures showed little variation between participants or over time. Our findings provides guidance for future studies to select alternative measures and/or to administer some engagement measures less frequently. This represents a meaningful step forward for the field of engagement research, as prior to this study, there was almost no empirical guidance on how to select measures of engagement or how frequently to administer them. We observed a substantial drop in engagement for participants in the supported arm when video sessions began, contrary to our expectations that video sessions would improve engagement. Notably, video sessions began partway through the intervention and participants were blinded to condition until the onset of these sessions; changes to blinding and/or timing of video sessions might improve engagement. At the same time, those participants in the supported arm who did engage in video sessions (n=4) completed more of each weekly lesson than participants in the unsupported arm, indicating that there is evidence that therapist support can increase engagement. Secondary outcomes were sample diversity and depression and anxiety symptom change. Our sample was comparably diverse to other digital health studies in terms of race and socioeconomic status, and we overrecruited LGBTQ+-identified participants. Despite finding little evidence of changes in behavior or the theoretical mechanisms underlying the intervention, the majority of participants showed improvement in their symptoms.</p>