earticle

영어

Digital therapeutics deliver clinically validated treatments autonomously, yet behavioral mechanisms sustaining long-term change remain underexplored. This study examines how user interaction patterns with a mobile insomnia app influence treatment adherence and sleep outcomes. Analyzing two randomized controlled trials using fixed-effects regression, generalized random forest, and Hidden Markov Models, we find significant improvements in sleep efficiency, total sleep time, and sleep onset latency. Early-stage gains were driven by baseline depressive and anxiety symptoms, while later improvements depended on behavioral engagement patterns like bedtime regularity and routine adherence. HMM analysis identified three behavioral profiles—normative-timing, highly routinized, and disorganized—with users transitioning toward structured routines. Critically, disorganized participants showed strongest recovery following external intervention prompts, highlighting the importance of human-mediated support. These findings reframe digital therapeutics as dynamic behavioral infrastructures scaffolding self-regulation, offering design implications for adaptive, user-centered interventions.