Mining User Reviews for Key Design Features in Cognitive Behavioral Therapy-Based Mobile Mental Health Apps.

Abstract

Background: Cognitive behavioral therapy (CBT)-based mobile apps have been shown to improve CBT-based interventions effectiveness. Despite the proliferation of these apps, user-centered guidelines pertaining to their design remain limited. The study aims to identify design features of CBT-based apps using online app reviews. Methods: We used 4- and 5-star reviews, preprocessed the reviews, and represented the reviews using word-level bigrams. Then, we leveraged latent Dirichlet allocation (LDA) and visualization techniques using python library for interactive topic model visualization to analyze the review and identify design features that contribute to the success and effectiveness of the app. Results: A total of 24,902 reviews were analyzed. LDA optimization resulted in 86 topics that were labeled by two independent researchers, with an interrater Cohen's kappa value of 0.86. The labeling and grouping process resulted in a total of six main design features for effective CBT-based mobile apps, namely, mental health management and support, credibility support, self-understanding and personality insights, therapeutic approaches and tools, beneficial rescue sessions, and personal growth and development. Conclusions: The high-level design features identified in this study could evidently serve as the backbone of successful CBT-based mobile apps for mental health.