Changes in physical activity and transitions between functional states among adults aged 50 years or older: a longitudinal study of five aging cohorts

Physical activity and functional status both change with age, yet little is known about how becoming active or inactive relates to subsequent functional decline and recovery. We examined these associations in adults aged 50 years or older. We analyzed longitudinal data from five population-based aging cohorts. Change in physical activity from T1 to T2 was classified as stable inactive, became active, became inactive, or stable active. Median T1–T2 intervals were 2 years in four cohorts and 3 years (IQR, 2–9) in the Mexican cohort. Functional status comprised Independent, instrumental activities of daily living (IADL) disability only, and basic activities of daily living (BADL) disability. Death was not modeled as an outcome state; observation ended at the last valid functional-state assessment. Cohort-specific estimates from continuous-time three-state Markov models were combined using random-effects meta-analysis. Complete-case analysis was primary; multiple imputation was a sensitivity analysis. Of 266,204 participants meeting the age criterion, 115,140 contributed eligible paired physical-activity and functional-state data and 112,855 entered the primary complete-case analysis. In the primary analysis, becoming active and remaining active were generally associated with less deterioration and greater recovery than stable inactivity; stable activity showed the most favorable overall profile. Within baseline-activity strata, activity initiation versus stable inactivity was associated with a lower transition intensity from Independent to IADL disability only (HR, 0.58; 95% CI, 0.46–0.72) and a higher recovery intensity from BADL disability to Independent (HR, 1.73; 95% CI, 1.40–2.13). Activity cessation versus continued activity was associated with the opposite pattern (HRs, 1.28 [95% CI, 1.03–1.58] and 0.66 [95% CI, 0.61–0.72], respectively). I² ranged from 26.1% to 98.5%, and 12 of 18 prediction intervals included 1. Multiple-imputation results were directionally concordant. Becoming active after a period of inactivity and remaining active were associated with less functional deterioration and greater recovery, whereas becoming inactive was associated with less favorable subsequent transitions than continued activity. The substantial variation between cohorts and the observational design warrant cautious interpretation.