Identifying Days of Valid Wear for Studies Using Consumer-Grade Accelerometer-Based Activity Trackers: ActiWearCheck, a Python Library

Obtaining reliable data from activity trackers is crucial for physical activity studies. Except for higher end models that offer heart rate (HR) monitoring, consumer-grade devices typically do not provide a method for estimating wearing time. This manuscript introduces a new Python software (“ActiWearCheck”) developed to identify days of valid wear of Fitbit wrist-worn activity trackers, without relying on HR data. The software utilizes minute-per-minute time series for energy expenditure (EE) estimates extracted from accelerometer data and freely provided by the official Fitbit Web API. A case analysis was conducted on a data set of 3,339 days collected from 72 individuals who used the Alta HR or Inspire 2 devices to test the feasibility of estimating days of valid wear using ActiWearCheck with two different configurations, EE estimates ≥ 600 min over the basal metabolic rate and ≥10 hr with at least 1 min of EE over the basal metabolic rate. The valid wear classifications under both configurations showed accuracies of 72% and 86%, respectively, against the one performed using HR data (also available). The average daily estimates of the number of steps and EE were calculated for the 72 participants after identifying the days of valid wear, performed using ActiWearCheck. This procedure allowed exclusion of up to 25% of potentially invalid days, leading to 12% and 6% increases in the step count and EE estimates, respectively. These results demonstrate the feasibility and necessity of obtaining valid wear estimates for physical activity studies employing consumer-grade devices.