24-hour Measurement Of Sleep, Sedentary, And Physical Activity Behaviors With Wearable Monitors

Abstract

Combining sleep, sedentary behavior (SED), light intensity physical activity (LIPA) and moderate-to-vigorous physical activity (MVPA) research in a meaningful way has become possible due to the availability of small wearable monitors that measure all domains in the 24-hour activity cycle. PURPOSE: This study compares 9 monitors for accuracy in 24-hour activity measurement. METHODS: Healthy adults (N=40, 47% male) wore: Actigraph (AG), ActivPAL (AP), Fitbit One (FB), GENEActiv (GA), Jawbone Up (JB), Lumoback (LB), Nike Fuelband (N), Omron pedometer (O), and Z-Machine (Z) for 24-Hours. Data were summarized for each participant into a waking day of activity and a full night of rest. Comparisons were made on total sleep, time spent in SED, LIPA, MVPA, and steps. Statistical analysis included mean absolute percent error (MAPE), equivalency testing, and Bland-Altman plots. RESULTS: For sleep, MAPE values were 8.1% (AG), 8.6% (JB), 9.5% (FB), and 16.9% (GA) relative to Z. For SED, MAPE values were 9.5% (LB), 13.2% (FB), 15.8% (AG), and 65.8% (GA) relative to AP. For LIPA, MAPE values were 19.7% (GA) and 28.0% (FB) relative to AG. For MVPA, MAPE values were 51.8% (JB), 62.1% (FB), 75.7 (GA), and 92% (N) relative to AG. For steps, MAPE values were 14.1% (AG), 19.1% (JB), 20.3% (FB), 21.7% (LB), 22.5% (AP), and 29.9% (N) relative to O. Equivalence testing indicated only AG for steps and LB for sedentary behaviors were equivalent to the criterion. Bland-Altman plots indicated the AG had the narrowest 95% CI for sleep, with similar results from JB and FB, but not GA. For SED, the LB was closest to AP due to posture measurement. LIPA was not measured accurately by either the FB or the GA, but AG cutpoints have similar error with more precise standards. JB had the closest agreement on MVPA followed by the FB, with large error for N and GA. AG had the best step measurement, but FB produced similar results. CONCLUSIONS: Currently, no single consumer-based wearable monitor can validly measure the full 24h spectrum of behaviors. Sleep measurement was most consistent across monitors; SED, LIPA and MVPA had larger errors but may still be acceptable as feedback tools within intervention studies. GA with cutpoint analysis on the wrist is not recommended, nor is using the N for any measurement.