Secondary Logo

Accuracy of Smartwatches in Assessing College Students’ Energy Expenditure in Exercise with Different Intensities: 1660 Board #335 June 1 800 AM - 930 AM

Gao, Zan FACSM1; Li, Xianxiong2; Zeng, Nan1; Pope, Zachary1; Yang, Huimin2; Liu, Wenfeng2; Xiong, Hui2; Chen, Yanting2; Li, Jiao2; He, Wei2

Medicine & Science in Sports & Exercise: May 2017 - Volume 49 - Issue 5S - p 474
doi: 10.1249/01.mss.0000518190.73055.5a
C-49 Free Communication/Poster - Physical Activity Assessment in Youth Thursday, June 1, 2017, 7:30 AM - 12:30 PM Room: Hall F

1University of Minnesota, Minneapolis, MN. 2Hunan Normal University, Changsha, Hunan, China.


(No relationships reported)

PURPOSE: Despite the recent prevalence of health wearable devices (e.g., smartwatches), the validity and reliability of smartwatches’ physiological outcomes remain largely unexplored. This study was designed to examine the accuracy of the Microsoft Band (MS), Fitbit Surge HR (FB), TomTom Cardio Watch (TT) and Apple Watch (AW) in assessing energy expenditure (EE) during three exercise sessions with different intensities.

METHODS: Participants were 25 college students (12 males; M age= 23.52, SD = ± 1.04) from Southcenter region in China. They completed three separate 10-minute exercise sessions at light physical activity (LPA, walking at 3.0 km/h), moderate PA (MPA, running at 5.0 km/h), and vigorous PA (VPA, running at 7.0 km/h) conditions on the Hpcosmos treadmill at a highly controlled laboratory. In this study, the AW and TT were placed on the right wrist while the FB and MS were worn on the left wrist. All demographic information (i.e., height, weight, age, gender) was loaded onto each smartwatch prior to testing. EEs of smartwatches were then validated against EE data from a Actigraph wGT3X accelerometer worn on the right side of waist.

RESULTS: ANOVAs with repeated measures revealed significant differences between smartwatches for EE, F(9, 16)=45.73, p<0.01, η2=0.98. When validated against accelerometer, post-hoc comparisons suggested significantly different EE assessments for the FB at LPA and MPA conditions (p<0.01), and for MS and FB at the VPA (p<0.01). Interclass correlations between the accelerometer and smartwatches revealed reliability for EE for the TT at LPA (r=0.65, p<0.01), for the FB, TT and AW at MPA (r=0.53-0.59, p<0.01), and for MS, FB and TT at VPA (r=0.59-0.65, p<0.01). However, only the TT demonstrated excellent agreement with the accelerometer at LPA with coefficient of variation of 2.0%. Additionally, Bland-Altman plots yielded satisfactory precision/no bias for FB and TT measurements against the accelerometers at all conditions (p<0.05), for MS at LPA and VPA (p<0.05), and for AW at VPA (p < 0.05).

CONCLUSIONS: Findings suggested EE measurements of smartwatches are far from ideal in terms of validity and reliability. In this study the most valid and reliable measurements were from the TT. Future study may investigate the validity and reliability of smartwatch measurements in everyday life.

© 2017 American College of Sports Medicine