(PS7-A14) Comparative Efficacy of Digital Just-in-time Interventions for Emotion Regulation

Background: Mental health conditions are highly prevalent in the United States. However, many who need mental health treatments do not receive care. There is an immense need for scalable mental health interventions which address barriers to the provision and receipt of evidence-based mental healthcare. Digital mental health interventions provided through mobile applications are a promising avenue for mental healthcare, although the efficacy of such interventions has not been comprehensively explored. The present study examines engagement with, and efficacy of, cognitive-behavioral just-in-time interventions (JITIs) provided through a mental health app, Youper.

Method: The present study used data from users (N = 8,323) of Youper collected from July to October of 2020. The following JITIs were examined: acceptance, cognitive restructuring, gratitude, mindfulness, problem-solving, and self-compassion. A logistic regression with cluster-robust standard errors assessed the probability of user engagement (yes or no) with JITIs. A multilevel model, accounting for nesting of interventions within conversations within users, examined helpfulness (1 = not helpful, 2 = sort of helpful, 3 = definitely helpful) of JITIs.

Results: Compared to acceptance-based interventions, odds of engaging with cognitive restructuring (OR = 2.41, SE = .08, p < .001), problem-solving (OR = 1.76, SE = .06, p < .001), self-compassion (OR = 1.60, SE = .05, p < .001), and gratitude (OR = 1.54, SE = .05, p < .001) were significantly higher, but odds of engaging with mindfulness were significantly lower (OR = .59, SE = .02, p < .001). As far as helpfulness, cognitive restructuring (M = 2.45; b = .25, SE = .01, p < .001), gratitude (M = 2.29; b = .10, SE = .01, p < .001), mindfulness (M = 2.31; b = .13, SE = .01, p < .001), and problem-solving (M = 2.32; b = .13, SE = .01, p < .001) were rated as more helpful than acceptance. There was not a significant difference between the efficacy of self-compassion (M = 2.21; b = .02, SE = .01, p = .11) and acceptance (M = 2.23). Pairwise comparisons also demonstrated that problem-solving was more effective than gratitude (p < .01) and self-compassion (p < .001), that gratitude and mindfulness were more effective than self-compassion (all ps < .001), and that cognitive restructuring was more effective than all other interventions (all ps < .001).

Discussion: Results demonstrate that overall, cognitive-behavioral JITIs provided through a mental health app are perceived as helpful by users. Differences in both engagement and perceived helpfulness between specific interventions emerged, broadly suggesting that users are more likely to engage with, and find helpful, skills overtly targeting thoughts or problems than they are skills that foreground acceptance and mindfulness techniques. These findings have implications both for the utility of app-based mental health interventions and for the personalization of digital interventions.