Feed

Physical activity is essential for preventing cognitive decline, stroke and dementia in older adults. A new cardiovascular diagnosis offers a critical window for positive lifestyle changes. However, sustaining physical activity behavior change remains challenging and the underlying mechanisms are poorly understood. To identify the neural, behavioral, and contextual determinants of long-term physical activity change after a new cardiovascular diagnosis, we applied support vector machine learning to predict 4-year trajectories of both self-reported and accelerometer-derived moderate-to-vigorous physical activity in 295 cognitively unimpaired older adults from the UK Biobank, testing three models that incorporated baseline: (i) demographic, cognitive, and contextual factors, (ii) baseline resting-state functional connectivity alone, and (iii) combined multimodal features across all predictors. The combined multimodal model had the highest predictive power (r = 0.28, P = 0.001). Key predictors included greenspace access, social support, executive function and between-network functional connectivity within the default mode, and frontoparietal control networks. These findings underscore the importance of behavioral factors and social determinants of health and uncover neural mechanisms that may support lifestyle modifications. In addition to furthering our understanding of the mechanisms underlying successful physical activity behavior change, these findings help to guide the design of interventions and health policy with the ultimate goal of preventing cardiovascular disease burden and late-life cognitive decline.