[Cluster predictors of trajectories of leisure-time physical activity intensity in men and women from ELSA-Brasil].

Abstract

The maintenance of physical activity over time is a challenge for public health. Predictors of different physical activity intensities have not been sufficiently analyzed. This study aimed to identify clusters of trajectories of physical activity intensity in leisure time, their predictors and the profile of the participants in the clusters. Baseline data and two follow-up visits of 11,262 participants from the Brazilian Longitudinal Study of Adult Health (ELSA-Brazil) were included. physical activity was assessed at three moments using the International Physical Activity Questionnaire (IPAQ). Clusters of physical activity trajectories according to intensity (weak, moderate and strong) were identified via longitudinal K-means. The number of clusters was based on the within-clusters sum-of-squares (WCSS) measure and the classification was based on scientific recommendations. Machine learning was used to verify predictors importance. Five clusters were identified for men and four for women. Men in the adequate cluster with a strong increase in physical activity had higher income, schooling level, and daily consumption of fruits and vegetables; they were younger; had never smoked and had a normal nutritional status. On the other hand, women in the adequate cluster with moderate physical activity increase had higher income and schooling level; had never smoked and had a normal nutritional status. In both sexes, age and schooling level were the most important predictors for classification in clusters. Actions to promote physical activity should be implemented over time, and be adapted to sociodemographic and behavioral factors.