Behavioral Mechanisms of a Digital Health Intervention for Self-Management in Type 2 Diabetes Mellitus: Prospective Longitudinal Cohort Study.

BACKGROUND: Sustaining self-management is critical for optimizing clinical outcomes in individuals with type 2 diabetes mellitus (T2DM). Although digital health interventions (DHIs) have shown benefits for glycemic control and self-care, much of this evidence has focused on efficacy, and the behavioral mechanisms through which DHIs produce sustained effects remain insufficiently understood. Clarifying these mechanisms could inform the development of theory-driven interventions. OBJECTIVE: This study examined the longitudinal behavioral pathways through which Artificial Intelligence-based Health Education Accurately Linking System, a WeChat (Tencent)-based digital health program, influences T2DM self-management, using the Extended Multi-Theory Model (MTM) of health behavior change. METHODS: An explanatory sequential mixed methods prospective longitudinal cohort study was conducted among adults with T2DM (aged ≥18 y and proficient in WeChat use), recruited from 45 primary health care institutions in Beijing, China, between July 2023 and July 2024. Self-management behavior was assessed as the primary outcome using the Summary of Diabetes Self-Care Activities, and psychosocial determinants using the Extended MTM Scale and the Diabetes-related Skills Scale. Exploratory and confirmatory factor analyses assessed the construct validity of the Extended MTM Scale. Structural equation modeling examined longitudinal pathways among Artificial Intelligence-based Health Education Accurately Linking System users across baseline and 3, 6, and 12 months. For the qualitative phase, a purposive subsample was selected through maximum variation sampling based on baseline glycated hemoglobin; interviews were analyzed thematically until thematic saturation, and integrated with quantitative findings using a joint display. RESULTS: Of the 406 enrolled participants, 391 completed baseline assessments. The Extended MTM Scale demonstrated a 6-factor, 22-item structure with excellent internal consistency (Cronbach α=0.928) and satisfactory construct validity. The structural equation modeling showed satisfactory fit (CFI=0.984, RMSEA=0.036). Changes in the social environment (β=0.23, 95% CI 0.07-0.38; P=.003) and physical environment (β=0.25, 95% CI 0.10-0.40; P=.001) at baseline, and diabetes-related skills at month 6 (β=0.16, 95% CI 0.03-0.29; P=.01), were directly associated with self-management behavior at month 12, whereas behavioral confidence and emotional transformation showed no significant direct effects. Social environment changes were indirectly associated with behavioral confidence through participatory dialogue at month 3 (β=0.20, 95% CI 0.04-0.36; P=.01; β=0.66, 95% CI 0.57-0.74; P<.001). Thematic analysis of 17 interviews identified 3 domains: environmental context, cognitive processes, and attitudes and skills. Environmental factors converged across both data strands, while qualitative data expanded on the cognitive and attitudinal processes underlying sustained self-management. CONCLUSIONS: This study is among the first to apply the Extended MTM framework to DHI-supported T2DM self-management, with environmental factors emerging as key drivers alongside selected cognitive, attitudinal, and skills-related processes. Complementing efficacy-focused research, it illuminates the psychosocial pathways underlying sustained self-management and refines the Extended MTM in digital health contexts. These insights can inform the design of theory-driven DHIs in primary care.