Smartphone CBT engagement and depressive symptoms: secondary analysis of the RESiLIENT trial using a time-varying exposure approach

Abstract Background Smartphone-based cognitive behavioral therapy (CBT) programs offer accessible interventions for subthreshold depression, yet engagement needed for meaningful benefit remains unclear. We examined how lesson and worksheet engagement relate to depressive symptom improvements in a behavioral activation (BA) intervention, accounting for time-varying confounders. Methods This secondary analysis included 298 adults assigned to the BA arm of the RESiLIENT trial, a randomized controlled trial in Japan. Lesson and worksheet completion were treated as time-varying exposures, each yielding four engagement patterns: minimal (Few-Few) , early (Many-Few) , late (Few-Many) , and consistently high (Many-Many). Outcomes were depressive symptom changes measured by the Patient Health Questionnaire-9 (PHQ-9) at weeks 6 and 26. We applied the parametric g-formula to estimate counterfactual PHQ-9 changes under each pattern, adjusting for baseline and time-varying confounders. Results Early lesson engagement during weeks 0–3 was associated with larger PHQ-9 reductions at both weeks 6 and 26, even when later engagement declined ( Many-Few vs. Few-Few : week 6: −1.47 [95% CI −2.52 to −0.53]; week 26: −1.27 [−2.53 to −0.17]). In contrast, higher worksheet engagement was linked to improved PHQ-9 at week 6, with maximal benefit among consistently high engagers ( Many-Many vs. Few-Few : −1.25 [−2.17 to −0.44]) and late engagers ( Few-Many vs. Few-Few : −1.18 [−2.20 to −0.08]), but not persist to week 26. Conclusions Greater engagement with smartphone-delivered BA is associated with larger symptom reductions. Early lesson engagement drives sustained benefit, whereas worksheet engagement did not persist. These findings may guide digital CBT design by emphasizing early lesson completion alongside concurrent skill practice.