A Study on the Continued Use of AI-Enabled mHealth Services Among Older Adults from a User Experience Perspective: The Moderating Effect of AITS

INTRODUCTION: Against the backdrop of an accelerating aging population and a high prevalence of chronic diseases among the elderly, the imbalance between supply and demand for traditional healthcare resources has become increasingly prominent. While AI-enabled mobile health (mHealth) emerges as a vital solution to the challenges of elderly health management, critical issues such as the low Continued Usage Intention (CUI) among elderly users and the limitations of outdated research frameworks remain inadequately addressed. OBJECTIVES: This study investigates the mechanisms underlying older adults' continued use of AI-enabled mHealth services, constructing and validating an adapted age-friendly technology acceptance framework that integrates the Expectation-Confirmation Model (ECM-ISC) with Quality of Experience (QoE), while examining the moderating role of AI Technology Services (AITS). METHODS: Grounded in the ECM-ISC, this research introduces QoE as a core construct. User experience was decomposed into five dimensions: Sensory Experience (SE), Interaction Experience (IE), Functional Experience (FE), Emotional Experience (EE), and Trust Experience (TE). AITS was incorporated as a moderating variable to build an "Experience Dimensions–QoE–User Satisfaction (US)–CUI" framework. A questionnaire survey was conducted in East China in 2025, yielding 536 valid responses. Data were analyzed using Structural Equation Modeling (SEM) combined with moderation effect tests. RESULTS: Empirical results demonstrate that: (1) All five experience dimensions significantly and positively influence QoE. IE, FE, and TE had the highest point estimates (β = 0.24), although pairwise tests showed the differences among the five dimensions (β = 0.18–0.24) were not statistically significant; EE had the lowest mean score (M = 4.71, SD = 0.98), indicating a current supply gap despite significant latent demand. (2) QoE acts as the core driver of CUI, with a path coefficient to US of 0.79 and a direct effect on CUI of 0.70—substantially higher than the direct effect of US itself (β = 0.20). This suggests elderly users rely more on rational judgments of overall service reliability than on transient emotional satisfaction when deciding to continue usage. (3) AITS significantly and positively moderates the relationships between SE, IE, FE, EE, and QoE; however, its moderating effect on the TE–QoE relationship was not significant (B = 0.041, 95% CI −0.004 to 0.086, p = 0.081), so this study provides no evidence that AI features directly alleviate trust-related concerns. CONCLUSION: This study contributes theoretically by adapting the traditional IS continuance model for an aging context, proposing an "adapted age-friendly technology acceptance framework," and delineating the experiential empowerment boundaries of AI technology. Practically, it proposes a tiered, differentiated support design strategy prioritizing "functionality and trust, optimizing interaction, and compensating emotion." These findings provide guidance for developing age-friendly digital health services that foster long-term engagement among the elderly, although longitudinal validation of actual usage behavior remains needed.