4 - (SYM 54) Participant-level Predictors of Engagement with a Smartphone-delivered Interpretation Bias Intervention

Background: Cognitive bias modification of interpretations (CBM-I) has been shown to effectively reduce interpretation bias and thereby decrease anxiety and depression symptoms, but there are high attrition rates with these digital interventions. Although it is well established that improving platform features and usability can help encourage engagement with digital mental health interventions (DMHIs), the current literature is mixed on which participant-level factors predict engagement with CBM-I. The present study collected data from a randomized control trial comparing the use of a smartphone-based CBM-I intervention, “HabitWorks”, and online Symptom Tracking to examine participant-level predictors of different aspects of engagement with DMHIs.

Method: Currently, 299 adults (M = 32.8 years; 63.2% women; 54.5% White) with mild to moderate anxiety were administered measures pre and post the four-week intervention period. Bivariate associations were examined between participant-level predictors (demographic variables, symptom severity, and beliefs about DMHIs) and various aspects of engagement (adherence, attrition, level-of-use, and perceived engagement). Significant associations at an alpha level <.20 will be included in multivariate regression models. Final regression models will demonstrate which participant characteristics predict each aspect of engagement.  

Results: Preliminary analyses suggest that, using the predetermined alpha of 0.20, the following variables will be retained for multivariate models. Age (r = .32, t(140) = 4.03, p < .001, depression severity (r = .13, t(140) = 1.67, p = .09),anxiety severity (r = .15,  t(140) = 1.85, p = .07), and expectancy (r = .12, t(137) = 1.41, p = .16) were retained as predictors of level-of-use of HabitWorks. Age (χ²(1) = 5.75, p = 0.02), condition (χ²(1) = 18.07, p < .001), Hispanic ethnicity (χ²(1) = 3.29, p = 0.07), depression severity (χ²(1) = 2.39, p = 0.12), and expectancy (χ²(1) = 1.86, p = 0.17) were retained as predictors of attrition. Age (χ²(1) = 13.06, p < 0.001), condition (χ²(1) = 34.96, p < 0.001), and race (χ²(3) = 4.93, p = 0.18) were retained as predictors of adherence. Age (r = .12, t(257) = 1.91, p = .06), condition (t(257) = -1.95, p = .05), race (F(3, 205) = 2.83, p = .04), credibility (r = .49, t(256) = 8.92, p < .001), and expectancy (r = .52, t(253) = 9.58, p < .001) were retained as predictors of perceived engagement. Final regression models with the full dataset will be presented and implications of participant-level predictors of engagement with DMHIs will be discussed.