Key points are not available for this paper at this time.
Background: Treatment-as-usual (TAU) conditions are intended to reflect the support typically received in routine treatment settings. For digital mental health interventions (DMHIs) delivered online, TAU conditions should reflect the usual patterns of online help-seeking. The lack of ecologically valid TAU control conditions has been a gap in effectiveness trials of online DMHIs. In this study, mental health-related popular online content (eg, advice TikToks, lived experience vlogs, and self-care infographics) was examined as a valuable TAU control condition. Objective: This study examined the feasibility of popular online content as a TAU control condition in DMHI trials. Methods: This study was a secondary analysis of two randomized controlled trials. Both trials recruited participants online, primarily via an online study recruitment platform. In study 1 (N=916), US adults with elevated depression or anxiety were randomized to either (1) complete a single-session DHMI for depression and anxiety (n=291), (2) search the web for popular online content relevant to their struggles (n=312), or (3) search a curated library of mental health-related popular online content (n=313). In study 2 (N=431), US adults with elevated loneliness were randomized to (1) complete a single-session DHMI for loneliness (n=136), (2) search a curated library of popular online content related to loneliness (n=145), or (3) complete an attention-matched control condition (n=150). All 6 programs took approximately 10 to 20 minutes to complete and were entirely self-guided. Participants rated each program's credibility and expected benefit, as well as their feelings of distress (study 1) and loneliness (study 2). The studies did not involve interaction between participants and the research team. Results: In study 1, dropout during the treatment was 4.8% (14/291) for the single-session intervention, 25.9% (81/312) for online help-seeking, and 9.6% (30/313) for the curated library. The curated library's credibility and expected benefit score did not differ from that of the single-session intervention (Cohen d=0.08; P=.88) and was higher than that of unguided help-seeking (Cohen d=0.23; P=.01). In study 2, dropout was higher in the curated library condition (7/145, 4.8%) than in the single-session intervention and the attention-matched control condition (0/136, 0.0% and 0/150, 0.0%). The mean credibility and expected benefit score for the curated library was comparable to that of the attention-matched control condition (Cohen d=0.00; P>.99) but lower than that of the single-session intervention (Cohen d=0.32; P=.02). Changes in distress and loneliness from baseline to 8-week follow-up did not differ across the conditions in study 1. All effect sizes were small in study 1 (Cohen d.06). Similarly, in study 2, all effect sizes were small (Cohen d.25). Conclusions: Curated libraries of popular online content are a feasible, ecologically valid TAU benchmark for effectiveness trials of online DMHIs. Future research on TAU conditions in online help-seeking contexts should better align with observed DMHI attrition rates and account for the increasingly central role of conversational artificial intelligence in online mental health support.
Kaveladze et al. (Fri,) studied this question.