Comparison of Caloric Expenditure in a Smart Watch and Portable Metabolic Cart

Abstract

PURPOSE: To validate a Smart Watch (SW) for wheelchair users by comparing caloric energy expenditure (EE) against expired gas analysis. Valid activity tracking will be helpful for clinicians, patients and consumers and may help increase physical activity levels among wheelchair users. METHODS: Five wheelchair users (age=50.0(5.6)) and three able-bodied (age=25.3(3.2)) participants completed a series of exercises including wheelchair treadmill propulsion at 30, 45, and 60 strokes per minute (spm) and arm cycle ergometry at 45, 60, and 80rpm. They were equipped with a SW on their dominant hand, heart rate monitor, and a portable metabolic tracking cart. The bundled workout app was used for each task. Caloric expenditure data was extracted from both devices and compared by Bland-Altman analysis. RESULTS: For treadmill tasks, the SW reported the average EE at 30, 45, and 60spm frequencies were 7(2.0), 8(2.3) and 9(2.0) kcals, respectively. At the same frequencies, the metabolic cart expenditures read 10(3.9), 12(5.3), and 15(6.7). Bland-Altman analysis showed relatively poor agreement between the cart and watch at 30spm (mean difference 3 with limits of agreement (LoA) -4–9). Mean absolute percent error (MAPE) was 21.56%. Agreement worsened at higher stroke frequencies, 45spm (4(- 4–12)) and 60spm (6(-4–10)). MAPE was 29.11% and 35.88%, respectively. For arm ergometry, the average EE reported by the watch at 45, 60, and 80rpm were 7(1.0), 9(1.5), and 11(1.4). Metabolic cart expenditures were 6(3.0), 7(2.4), and 8(2.8) at the same frequencies. Bland-Altman analysis showed good agreement at 45rpm (-0.4(-6–5)) with a MAPE of 32.69%. Agreement worsened at higher frequencies, 60rpm (-3(-6–1)) and 80rpm (-2(-7–2)). MAPE was 58.57% and 48.54%, respectively. CONCLUSION: While performing a treadmill task, the SW underestimated caloric expenditure, but overestimated for arm ergometry. The activity tracker records EE with good validity only at lower frequency tasks.