Accuracy of Wrist-Worn Activity Monitors during Wheelchair Use and Arm Ergometry

Abstract

Recently, the use of wrist worn activity monitoring devices using tri-axial accelerometers to estimate step counts has become widespread in the general population. PURPOSE: To evaluate whether common, commercially available fitness monitoring devices can accurately estimate wheelchair stroke counts and arm ergometer revolutions. METHODS: 30 able-bodied wore three commercially available activity trackers using tri-axial accelerometers on the right wrist in random order. Participants propelled a wheelchair on a roller treadmill at three separate frequencies (30, 45 and 60 strokes per minute), 3 minutes each. They also propelled the wheelchair on the rollers at pre-determined varied frequencies, ranging from 30 bpm to 80 bpm for 2 minutes. Each frequency was performed 3 times in random order. The participants were shown a video recording of a metronome at each of the frequencies to maintain the prescribed frequencies for each task. Participants also freely wheeled through an obstacle course twice. Two experimenters used tally counters to count actual wheelchair strokes. A second group of subjects performed an arm ergometry task at three different frequencies (40, 60 and 80 cycles per minute) for three minutes each in randomized order. RESULTS: Mean (SD) percentage error for 30, 45 and random stroke frequencies were ≥20 (11)% for all monitors. At 60 rpm, mean percent error was 12 (13), 12 (11), 5 (5)% for monitors A, B and C, respectively. For obstacle course strokes, intraclass correlation coefficients (95% CI) were 0.888 (.767-.946), 0.853 (.693-.930), 0.917 (.828-.961) and Lin’s concordance coefficient (95% CI) were 0.792 (0.655-0.879), 0.740 (0.531-0.864), 0.846 (0.715-0.919) for monitors A, B and C, respectively. For arm ergometry, standard error of the measurement varied widely for each monitor across all frequencies from 20, 33 and 261 strokes for monitors A, B and C, respectively at 40 rpm; 49, 190 and 8 strokes at 60 rpm; to 268, 267 and 107 strokes at 80 rpm. CONCLUSION: At low wheel chair stroke frequencies, and at low and high ergometer cycling frequencies, commercially available wrist worn fitness monitors perform rather poorly in estimating counts. Activity monitors specific to activities commonly performed by persons with lower limb paralysis are necessary to accurately estimate their activity.