Characterizing Behavioral Engagement in Smartphone Cognitive Behavioral Therapy for Obsessive Compulsive Disorder: Secondary Analysis of a Randomized Controlled Trial

Abstract

Background: Digital mental health interventions (DMHIs) are increasingly popular for their ability to disseminate evidence-based treatments in a scalable, accessible way. Engagement with digital treatments is critical for effectiveness but remains poorly defined and measured. Most studies rely on subjective self-reports or 1 or 2 objective metrics, overlooking the multidimensional nature of how users interact with DMHIs. Objective: This exploratory secondary analysis applied a road map for comprehensive behavioral engagement measurement, developed for a prior smartphone cognitive behavioral therapy program for body dysmorphic disorder, to a 12-week randomized controlled trial of smartphone cognitive behavioral therapy for obsessive compulsive disorder (OCD; N=60). Aims were to (1) comprehensively characterize behavioral engagement using 21 objective app analytics-based variables, (2) identify distinct behavioral engagement profiles via cluster analysis, (3) assess whether profiles were differentially associated with OCD symptom change, and (4) evaluate associations between profiles and pretreatment credibility, expectancy, posttreatment satisfaction, and perceived app quality. Methods: Behavioral engagement was measured via 21 objective variables derived from app analytics, spanning use volume, frequency, feature usage, session duration, and temporal patterns. Hierarchical agglomerative cluster analysis identified engagement profiles. Associations between cluster membership and OCD severity change (Yale-Brown Obsessive Compulsive Scale) were examined using hierarchical generalized linear mixed models. Objective measures were complemented with subjective self-reports: namely, cluster differences in treatment credibility, expectancy, satisfaction, and perceived app quality were examined using ANOVAs and Wilcoxon rank-sum tests. Results: Cluster analysis indicated a 3-cluster solution representing high (cluster 1: 21/60, 35%), moderate (cluster 2: 31/60, 52%), and low (cluster 3: 8/60, 13%) app users. All 3 clusters experienced significant reductions in OCD severity, with no significant differences in pre- to posttreatment Yale-Brown Obsessive Compulsive Scale change across clusters. Descriptively, cluster 1 had a higher rate of treatment responders than cluster 3. While clusters 1 and 2 had similar perceptions of treatment credibility and expectancy at baseline and satisfaction at posttreatment, cluster 3 perceived the intervention as significantly less credible, had lower expectations, and reported significantly lower satisfaction than cluster 2. All clusters rated app quality as similarly high. Conclusions: Comprehensive behavioral engagement measurement is feasible and informative across different DMHIs and clinical populations. The finding that all engagement profiles showed comparable symptom improvement challenges simple dose-response assumptions and may suggest that multiple paths of engagement can be therapeutically beneficial; nonsignificant differences in improvement across clusters may also reflect limited power to detect differences. Pretreatment credibility and expectancy are associated with engagement patterns, identifying a potential modifiable intervention target.

The authors' abstract, as published at the source. JMIR Formative Research, 2026 · DOI ↗