Comparing Five Generations of ActiGraph Devices using an Orbital Shaker.

Abstract

INTRODUCTION ActiGraph accelerometers are used extensively to objectively assess physical activity, sedentary behavior, and sleep. Here, we present an objective validation of five generations of ActiGraph sensors to characterize potential differences in output arising from changes to hardware or firmware. METHODS An orbital shaker generated accelerations from 0 to 3700 milli-g in a randomized order to test the wGT3X-BT, GT9X, CentrePoint Insight Watch (CPIW) 1.0 and 2.0, and the GT3X+ devices used in the 2011-2014 National Health and Nutrition Examination Survey (NHANES). Absolute (±50 milli-g) and relative (±5%) raw vector magnitude (VM) agreement to a NIST-calibrated accelerometer (3DM) was the primary outcome for static offset (delta from 3DM VM at 0 g), drift (delta VM start - minus end), and dynamic response. RESULTS We observed inter-generation differences [(Mean ± SD) -27.5 ± 8.7 mg (GT9X) to 4.8 ± 32.0 mg (GT3X+)] and intra-generational variability in static offset that carried over to lower shaker accelerations (< ~1300 milli-g), but these variations were mitigated with post-hoc calibration or integrated units (e.g., counts, MIMS). All generations fell within the ±5% of the 3DM across the tested acceleration range and were within ±50 mg of the 3DM below 3000 milli-g's. Drift was small [0.91 ± 0.27 mg (GT3X+), 0.07 ± 0.31 mg (GT9X)] relative to ≥1000 milli-g VM signal. CONCLUSIONS Using an orbital shaker and criterion accelerometer we found small differences in the raw output of five ActiGraph generations across a range of accelerations typical of free-living data. However, the devices largely fell within established limits of agreement to the NHANES GT3X+ devices and variability was minimized with auto-calibration or when data was expressed in MIMS or counts.