Harmonization of three different accelerometers to classify the 24 h activity cycle

Abstract

Increasing interest in measuring key components of the 24-Hour Activity Cycle (24-HAC) [Sleep, Sedentary Behavior (SED), Light Physical Activity (LPA), and Moderate to Vigorous Physical Activity (MVPA)] has led to a need for better methods. Single wrist-worn accelerometers and different self-report instruments can assess the 24-HAC but may not accurately classify time spent in the different components or be subject to recall errors. Objective: To overcome these limitations, the current study harmonized output from multiple complimentary research grade accelerometers and assessed the feasibility and logistical challenges of this approach. Approach: Participants (n=108) wore an: (a) ActiGraph GT9X on the wrist, (b) activPAL3 on the thigh, and (c) ActiGraph GT3X+ on the hip for 7-10 days to capture the 24-HAC. Participant compliance with the measurement protocol was compared across devices and an algorithm was developed to harmonize data from the accelerometers. The resulting 24-HAC estimates were described within and across days. Main Results: Useable data for each device was obtained from 94.3-96.7% of participants and 89.4% provided useable data from all three devices. Compliance with wear instructions ranged from 70.7% of days for the GT3X+ to 93.2% of days for the activPAL3. Harmonized estimates indicated that, on average, university students spent 34% of the 24-hour day sleeping, 41% sedentary, 21% in LPA, and 4% in MVPA. These behaviors varied substantially by time of day and day of the week. Significance: It is feasible to use three accelerometers in combination to derive a harmonized estimate the 24-HAC. The use of multiple accelerometers can minimize gaps in 24-HAC data however, factors such as additional research costs, and higher participant and investigator burden, should also be considered. 
Key Words: Sleep, Sedentary Behavior, Light Physical Activity, Moderate to Vigorous Physical Activity, Accelerometer, Algorithm