(PS8-70) Examining clinical moderators of treatment outcomes of a smartphone-delivered interpretation bias intervention in a community sample.

Abstract Text:

Background: Interpretation bias, or the tendency to interpret ambiguous situations negatively, is implicated in many emotional disorders, and is an important intervention target. Digital interpretation bias modification (IBM) interventions directly target this mechanism and have potential to increase access to care. HabitWorks (HW) is a personalized IBM smartphone app that has shown promise in small, pilot trials. However, it is not clear for whom it benefits the most. The goal of this project is to conduct a secondary analysis of the HW RCT to examine whether baseline anxiety severity, depression severity, and interpretation bias moderate the effectiveness of HW vs Symptom Tracking (ST), a comparison arm.

Methods: 337 participants (Mage = 33 years; 54.9% White, 15.4% Multiracial, 13.1% Black, 10.4% Asian, 0.6% American Indian or Alaskan Native; 63.8% women) who endorsed at least mild anxiety or depression symptoms were randomized to HW or ST. During the four-week intervention period, HW participants (n=163) were asked to complete IBM exercises at least three times a week, while ST participants (n=174) completed mood surveys three times a week. Both conditions completed measures of interpretation bias, depression symptom severity, and anxiety symptom severity pre- and post-intervention. Multiple linear regression analyses were conducted to test the moderator X condition interaction effect on pre-post change in anxiety and depression symptoms using six separate models. When interaction effects were non-significant, we examined the simple effects of the baseline moderator on clinical outcomes.

Results: Neither baseline depression nor baseline anxiety symptom severity moderated the effect of condition on changes in symptom outcomes. Across conditions, pre-intervention anxiety and depression severity significantly predicted decreases in pre-post changes in anxiety (b = 0.33, p < 0.01) and depression severity (b = 0.20, p < 0.01). Pre-intervention depression severity also significantly predicted decreases in pre-post changes in anxiety (b = 0.23, p < 0.01) and depression severity (b = 0.35, p < 0.01). Baseline interpretation bias did not moderate the effect of condition on pre-post change in anxiety severity. There was also no simple effect of baseline interpretation bias on pre-post changes in anxiety. However, baseline interpretation bias moderated the effect of condition on pre-post change in depression symptoms (b = 0.12, p < 0.05), such that those in the ST condition with more positive interpretation biases experienced a greater pre-post increase in depression symptoms from pre- to post-intervention (b = -0.07, p < 0.05). However, HW participants did not differ in their pre-post changes in depression outcomes based on baseline interpretation bias, p > .05.

Conclusion: Participants with greater depression and anxiety symptom severity in both conditions experienced greater pre-post improvements in their clinical symptoms. This suggests that digital tools, including IBM and symptom monitoring apps, may be appropriate for those with more severe anxiety and depression symptoms. Future secondary analyses of this RCT will examine other potential moderators of outcomes, such as IBM dosage and demographic characteristics.