Site-specific pain dynamics: associations between accelerometer-measured physical activity patterns and pain in older adults.

BACKGROUND: Physical activity (PA) has emerged as a promising non-pharmacological intervention for pain management, the relationship between objectively measured PA patterns and multi-site pain remains poorly understood. This exploratory study investigated associations between accelerometer-derived PA patterns and pain across various anatomical sites in older adults, and evaluated the predictive utility of machine learning model for pain outcomes. METHODS: This study utilized data from the National Health and Aging Trends Study in 2021-2022. Wrist-worn accelerometers measured PA, derived total activity, sedentary time, and activity/sedentary fragmentation) and time-frequency domain features. Cross-sectional and longitudinal analyses examined associations between PA patterns and site-specific pain using multivariable logistic regression with false discovery rate correction, while restricted cubic splines explored non-linear dose-response relationships. Random forest models with recursive feature elimination were developed to predict current pain status and pain relief. RESULTS: Cross-sectional analysis indicated that moderate sedentary time was associated with back pain (OR = 2.24, 95%CI: 1.12-4.47) and neck pain (OR = 2.12, 95%CI: 1.07-4.20), while moderate-to-vigorous activity fragmentation was associated with lower leg pain prevalence (OR = 0.34, 95%CI: 0.15-0.78), and moderate sedentary fragmentation with foot pain (OR = 1.89, 95%CI: 1.07-3.33). Longitudinal analysis revealed that moderate-to-vigorous activity fragmentation was associated with pain persistence in head (OR = 0.21, 95%CI: 0.05-0.90), though associations did not survive FDR correction. Machine learning prediction models achieved performance for pain status (AUC: back = 0.56, wrist = 0.54) and pain relief prediction (AUC: wrist = 0.85, back = 0.72). CONCLUSION: Site-specific tailoring of PA intensity and fragmentation is warranted for effective chronic pain management in older adults based on associations between PA and anatomical pain distribution.