Accelerometry data is a promising avenue to provide individual-specific data that can be beneficial in precision health frameworks. Other studies analyze accelerometry data by utilizing summary statistics and applying regression-based cutoff thresholds to determine Physical Activity Levels. However, these cutoffs are often not generalizable across populations, devices, or studies. Thus, a more functional-data approach would be useful in analyzing accelerometer data. We consider at subject’s probability activity curves to holistically represent their activity profile; these “probability activity curves” measure what percentage of an individual’s time is spent at or below a continuum of activity levels. The shape of these curves can provide useful information on a subject's activity profile. Using an adjusted fused lasso approach, we use these functional activity curves as covariates to determine what patterns of activity affect health outcomes, such as BMI and other anthropometric body measurements. We also consider these curves on an hour-by-hour basis in order to maintain the important time-specific information.