Exploring Heterogeneity in Smartphone Addiction Risk Among Community-Dwelling Older Adults in Guangzhou: A Machine Learning Analysis

Smartphone addiction risk is an underexamined concern in later life, and prior research often treats older adults as a homogeneous group. Using cross-sectional survey data from 2,275 community-dwelling adults aged 60 years or older in Guangzhou, this exploratory study applied XGBoost with SHAP analysis to identify predictive correlates, followed by K-means clustering to examine subgroup heterogeneity among participants classified as high risk. Social participation and depressive symptoms were associated with the risk with nonlinear patterns. Two exploratory profiles were identified: the Socially Active profile (40.7%), characterized by high participation and more regular non-kin contacts, and the Emotionally Withdrawn profile (59.3%), characterized by elevated depressive symptoms, fewer regular non-kin contacts, and entertainment-oriented use. The findings lend credence to the paradigms of social overload and emotional compensation and suggest that digital inclusion among older adults must be coupled with support for digital self-regulation, psychological well-being, and social and family engagements.