Abstract:
|
Latent class analysis (LCA), which has become more widely used in physical behaviors research, has identified unique patterns like weekend warriors, who engage in most of their physical activity (PA) one or two days a week. However, these analyses identify patterns for each intensity of physical behaviors in isolation, even though this does not reflect real world patterns of behavior. We extend LCA methods to incorporate data from multiple physical behaviors, including sedentary behavior in combination with (a) average steps (b) average VM/min, and (c) percent of light low, light high, and moderate-to-vigorous activity. This allows researchers to differentiate groups of individuals with similar patterns of one activity behavior by the patterns of another. For example, groups with similar sedentary behaviors can have different patterns of average steps. We illustrate these methods using data from 17,045 women aged 62-89 from the Women’s Health Study, with 7 days of accelerometer data. By identifying heterogeneous patterns capturing the compositional nature of PA behaviors, our method can lead to valuable insight into the associations between activity behaviors and health outcomes.
|