Impact of ActiGraph sampling rate on free-living physical activity measurement in youth

Abstract

ActiGraph sampling frequencies of more than 30 Hz may result in overestimation of activity counts in both children and adults, but research on free-living individuals has not included the range of sampling frequencies used by researchers. Objective: We compared count- and raw-acceleration-based metrics from free-living children and adolescents across a range of sampling frequencies. Approach: Participants (n=445; 10-15 y) wore an ActiGraph accelerometer for at least one 10-h day. Vector Magnitude counts, Mean Amplitude Deviation, Monitor-Independent Movement Summary units, and activity intensity classified using six methods (four cut-points, two-regression model, and artificial neural network) were compared between 30 Hz and 60, 80, 90, and 100 Hz sampling frequencies using mean absolute differences, correlations, and equivalence testing. Main results: All outcomes were statistically equivalent, and correlation coefficients were ≥0.970. Absolute differences were largest for the 30 vs. 80 and 30 vs. 100 Hz count comparisons. For comparisons of 30 with 60, 80, 90, or 100 Hz, mean (and maximum) absolute differences in minutes of moderate-to-vigorous physical activity per day ranged from 0.1 to 0.3 (0.4 to 1.5), 0.3 to 1.3 (1.6 to 8.6), 0.1 to 0.3 (1.1 to 2.5), and 0.3 to 2.5 (1.6 to 14.3) across the six classification methods. Significance: Acceleration-based outcomes are comparable across the full range of sampling rates and therefore recommended for future research. If using counts, we recommend a multiple of 30 Hz because using a 100 Hz sampling rate resulted in large maximum individual differences and epoch-level differences, and increasing differences with activity level.

Feasibility of Using a GENEActiv Accelerometer with Triaxial Acceleration and Temperature Sensors to Monitor Adherence to Shoulder Sling Wear Following Surgery

“…In a true patient scenario, setting the measurement frequency to 10-20 Hz would extend storage and battery life to at least 1-2 months making it more clinically practical without the need to download logged data before that time. Although the magnitude of acceleration may be reduced at lower sampling frequencies, there appears to be minimal impact on the variability of accelerations (suggesting this algorithm would perform similarly on data collected at lower sampling frequencies [28]. However, this should be confirmed in further research.…”

Association between school environment with sedentary behavior and physical activity intensity in children

“…Third point is that the questionnaires of school environment answered by the principals and the sociodemographic questions answered by parents may suffer from memory bias. Fourth point is that the accelerometry data was downloaded at a sampling rate of 80 Hz that could overestimate physical activity in children 44 .…”

The Intrinsic Properties of ActiGraph Counts and Alternatives

“…The proprietary nature of the AG counts method is not the only aspect of the summary measure which has been criticized. An intensity leveling off with running above 12 km/hr has been demonstrated with hip-worn devices John et al, 2012), which is not observed with wrist-and ankle-worn devices (LaMunion et al, 2017), changes to firmware (John et al, 2014), and sampling frequency issues (Brond & Arvidsson, 2016;Clevenger et al, 2019Clevenger et al, , 2022Garnotel et al, 2021) has further challenged the use of AG counts, even though this is not directly related to the algorithm per se. Moreover, the units of AG counts have been deemed arbitrary (Chen & Bassett, 2005;Straker & Campbell, 2012), as this is not a unit proposed by the International System of Units (SI).…”