Abstract P121: Consumer-based Physical Activity Monitors and Research-grade Accelerometers Underestimate Physical Activity in a Semi-structured Setting
Introduction: Consumer-based physical activity (PA) monitors have become popular tools to track PA (steps and energy expenditure [EE]), but little is known about their accuracy and comparability to research-grade PA monitors.
Hypothesis: We hypothesized that consumer-based PA monitors would have lower PA measurement accuracy than a research-grade PA monitor.
Methods: Healthy adults (n=30 [n=15 male], aged 18-79 years) were fitted with three Fitbit PA monitors (One and Zip, left hip; Flex, non-dominant wrist), one Jawbone PA monitor (UP24, non-dominant wrist), and one research-grade, accelerometer-based PA monitor (ActiGraph Link; right hip). Subjects performed ≥12 activities from a list of 21 sedentary, household, and ambulatory activities during an 80-minute, semi-structured protocol while wearing all five monitors and a portable metabolic analyzer. Steps and EE were recorded from PA monitors at the beginning and end of the protocol. A previously validated EE prediction equation (Freedson) was used to estimate EE from PA (PAEE) using ActiGraph data. A previous lab visit by subjects was used to estimate resting EE; resting EE was added to PAEE to estimate total EE from the ActiGraph. Criterion measures included a trained research assistant counting total steps (with a hand tally counter) and the metabolic analyzer for measurement of total EE and PAEE (total EE - resting EE). Repeated measures analysis of variance was used to assess step, total EE, and PAEE measures among the consumer-based PA monitors, ActiGraph, and criterion methods.
Results: Participants averaged 3,231±104 steps, 294±10 kcals for total EE, and 235±9 kcals for PAEE according to criterion measures. Consumer-based PA monitors and the ActiGraph significantly underestimated steps (mean difference [MD]: 714-1,047 steps [22.1-32.4%], p<0.01), total EE (MD: 37-109 kcals [12.6-37.1%], p<0.05), and PAEE (MD: 35-107 kcals [14.9-45.5%], p<0.05) compared to criterion measures. The Fitbit One, Zip, and Flex monitors estimated significantly higher steps than the ActiGraph (MD: 199-302 steps [9.1-13.8%], p<0.05), but there was no significant difference in steps estimated by the UP24 compared to the ActiGraph (MD: 144 steps [6.6%], p=0.21). All consumer-based monitors estimated significantly higher total EE and PAEE than the ActiGraph (MD: 23-60 kcals [12.6-32.7%] and 22-53 kcals [17.3.0-41.2%], respectively; p<0.05).
Conclusions: All PA monitors underestimated steps, total EE, and PAEE compared to the criterion measures. Large variability existed in the step and EE estimates from the PA monitors tested. Contrary to our hypothesis, the research-grade ActiGraph monitor (using Freedson EE prediction equation) had more severe PA underestimates than the consumer-based PA monitors. In conclusion, estimates of PA between studies using consumer-based vs. research-grade PA monitors should be interpreted with caution.
Author Disclosures: A.H. Montoye: None. M.S. Tuttle: None. M.B. Nelson: None. J.A. Steeves: None. L.A. Kaminsky: None.
- © 2016 by American Heart Association, Inc.