TY - JOUR
T1 - Effect of brief intermittent walking bouts on step count accuracy of wearable devices
AU - Toth, Lindsay
AU - Park, S.
AU - Pittman, W. L.
AU - Sarisaltik, D.
AU - Hibbing, P. R.
AU - Morton, A. L.
AU - Springer, C. M.
AU - Crouter, S. E.
AU - Bassett, D. R.
PY - 2019/1/1
Y1 - 2019/1/1
N2 - Purpose : To examine the effect of brief, intermittent stepping bouts on step counts from 10 physical activity monitors (PAMs). Methods : Adults ( N = 21; M ± SD , 26 ± 9.0 yr) wore four PAMs on the wrist (Garmin Vivofit 2, Fitbit Charge, Withings Pulse Ox, and ActiGraph wGT3X-BT [AG]), four on the hip (Yamax Digi-Walker SW-200 [YX], Fitbit Zip, Omron HJ-322U, and AG), and two on the ankle (StepWatch [SW] with default and modified settings). AG data were processed with and without the low frequency extension (AGL) and with the Moving Average Vector Magnitude algorithm. In Part 1 (five trials), walking bouts were varied (4–12 steps) and rest intervals were held constant (10 s). In Part 2 (six trials), walking bouts were held constant (4 steps) and rest intervals were varied (1–10 s). Percent of hand-counted steps and mean absolute percentage error were calculated. One sample t -test was used to compare percent of hand-counted steps to 100%. Results : In Parts 1 and 2, the SWdefault, SWmodified, YX, and AGLhip captured within 10% of hand-counted steps across nearly all conditions. In Part 1, estimates of most methods improved as the number of steps per bout increased. In Part 2, estimates of most methods decreased as the rest duration increased. Conclusion : Most methods required stepping bouts of >6–10 consecutive steps to record steps. Rest intervals of 1–2 seconds were sufficient to break up walking bouts in many methods. The requirement for several consecutive steps in some methods causes an underestimation of steps in brief, intermittent bouts.
AB - Purpose : To examine the effect of brief, intermittent stepping bouts on step counts from 10 physical activity monitors (PAMs). Methods : Adults ( N = 21; M ± SD , 26 ± 9.0 yr) wore four PAMs on the wrist (Garmin Vivofit 2, Fitbit Charge, Withings Pulse Ox, and ActiGraph wGT3X-BT [AG]), four on the hip (Yamax Digi-Walker SW-200 [YX], Fitbit Zip, Omron HJ-322U, and AG), and two on the ankle (StepWatch [SW] with default and modified settings). AG data were processed with and without the low frequency extension (AGL) and with the Moving Average Vector Magnitude algorithm. In Part 1 (five trials), walking bouts were varied (4–12 steps) and rest intervals were held constant (10 s). In Part 2 (six trials), walking bouts were held constant (4 steps) and rest intervals were varied (1–10 s). Percent of hand-counted steps and mean absolute percentage error were calculated. One sample t -test was used to compare percent of hand-counted steps to 100%. Results : In Parts 1 and 2, the SWdefault, SWmodified, YX, and AGLhip captured within 10% of hand-counted steps across nearly all conditions. In Part 1, estimates of most methods improved as the number of steps per bout increased. In Part 2, estimates of most methods decreased as the rest duration increased. Conclusion : Most methods required stepping bouts of >6–10 consecutive steps to record steps. Rest intervals of 1–2 seconds were sufficient to break up walking bouts in many methods. The requirement for several consecutive steps in some methods causes an underestimation of steps in brief, intermittent bouts.
KW - accelerometer
KW - activity monitor
KW - pedometer
KW - step counter
U2 - 10.1123/jmpb.2018-0050
DO - 10.1123/jmpb.2018-0050
M3 - Article
SN - 2575-6605
VL - 2
SP - 13
EP - 21
JO - Journal for the Measurement of Physical Behaviour
JF - Journal for the Measurement of Physical Behaviour
IS - 1
ER -