Population aging poses growing challenges to the maintenance of older adults’ independent mobility, which is a cornerstone of healthy and active aging. Existing research often relies on small-scale, cross-sectional data and narrowly emphasizes observed travel behavior, overlooking the underlying ability for mobility and the dynamic, nonlinear mechanisms shaping mobility changes over time. This study addresses these gaps in a rapidly aging, non-Western context. This study analyzed nationally representative longitudinal data from over 15,000 older adults in the Chinese Longitudinal Healthy Longevity Survey (CLHLS) to predict mobility changes across three-year periods. A data-driven approach combining Random Forest with SHapley Additive exPlanations (SHAP) is adopted to identify key predictors and reveal their nonlinear relationships, interaction effects, marginal effects, threshold effects, and individual heterogeneity. Mobility decline is most pronounced among individuals aged 75 and above, particularly women. Attentive family caregiving, sound cognition, moderate leisure engagement, and psychological resilience are core predictors for maintaining mobility. Cognitive function interacts synergistically with “family–community–society” support systems to enhance mobility. Older adults living in warm, humid regions with longer frost-free periods demonstrate better mobility outcomes. A simplified model with the top 10 predictors achieved comparable accuracy, enhancing interpretability and practical relevance. By integrating functional mobility measures with explainable machine learning models, this study provides new empirical insights for developing personalized aging interventions, age-sensitive transport policies, and inclusive urban environments that support mobility in later life. • Longitudinal survey of 15,000 Chinese older adults identifies mobility predictors • Family caregiving, cognition, age, psychological resilience, and leisure engagement sustain older adults' mobility • Cognitive function interacts with family, community, and social support networks to enhance mobility • Climate factors (precipitation and frost-free periods) and early-life urban–rural disparities are key environmental determinants • Random forest with SHAP reveals nonlinear, interaction, marginal, and threshold effects
Jiang et al. (Mon,) studied this question.