Accuracy of Apple Watch fitness tracker for wheelchair use varies according to movement frequency and task

Evan Glasheen; Antoinette Domingo; Jochen Kressler

doi:10.1016/j.rehab.2020.03.007

Accuracy of Apple Watch fitness tracker for wheelchair use varies according to movement frequency and task

Ann Phys Rehabil Med. 2021 Jan;64(1):101382. doi: 10.1016/j.rehab.2020.03.007. Epub 2020 May 4.

Authors

Evan Glasheen¹, Antoinette Domingo², Jochen Kressler³

Affiliations

¹ Exercise Physiology, School of Exercise and Nutritional Sciences, San Diego State University, ENS Building 351, 5500, Campanile Drive, 92182-7251 San Diego, CA, USA. Electronic address: glasheen.4@osu.edu.
² School of Exercise and Nutritional Sciences, San Diego State University, ENS Building 351, 5500, Campanile Drive, 92182-7251 San Diego, CA, USA.
³ Exercise Physiology, School of Exercise and Nutritional Sciences, San Diego State University, ENS Building 351, 5500, Campanile Drive, 92182-7251 San Diego, CA, USA.

PMID: 32335302
DOI: 10.1016/j.rehab.2020.03.007

Abstract

Objectives: Individuals with disabilities have high prevalence of sedentary lifestyle, obesity, and cardiometabolic disease. Physical activity monitors (i.e., step counters) are ill-suited for tracking wheelchair pushes. The study purpose was to investigate the validity of a consumer-level fitness tracker (Apple Watch) designed for wheelchair users.

Methods: Validation study. A total of 15 wheelchair users with disabilities and 15 able-bodied individuals completed 3-min bouts of wheelchair propulsion on a treadmill and arm ergometry at pre-determined cadences as well as overground obstacle and Figure 8 courses. Tracker stroke counts were compared against direct observation.

Results: We found no interaction of tracker counts and ability status across all tasks (P≥0.550), so results are presented for the combined sample. For treadmill tasks, Bland-Altman analysis (bias±limits of agreement) showed good agreement for only higher-rate fixed-frequency tasks (-15±48, -1±14, 0±5, and 0±27 for low, moderate, high, and variable cadence, respectively). Mean absolute percentage error (MAPE) was 22%, 3%, 1%, and 6%, respectively. Intraclass correlation coefficients (ICCs) (95% confidence intervals) were -0.18 (-0.51-0.20), 0.47 (0.13-0.71), 0.98 (0.96-0.99), and 0.22 (-0.16-0.54). We found significant overestimation by the tracker at low frequency (P<0.01). Arm ergometry showed good agreement across all cadences (0±5, -1±3, 0±8, 6±6). MAPE was 1%, 1%, 1%, and 4%. ICCs were 0.88 (0.77-0.94), 0.95 (0.89-0.97), 0.88 (0.76-0.94), and 0.97 (0.87-0.97). We found minimal (2rpm) but significant differences at variable cadence (P<0.01). Overground tasks showed poor agreement for casual-pace and fast-pace obstacle course and Figure 8 task (-5±18, 0±23, and -18±32, respectively). MAPE was 15%, 18%, 21% and ICCs were 0.90 (0.79-0.95), 0.79 (0.59-0.90), and 0.82 (0.64-0.91). Significant differences were found for propulsion at casual pace (P<0.01) and the Figure 8 task (P<0.01).

Conclusions: Apple Watch is suitable for tracking high-frequency standardized (i.e., treadmill) pushing and arm ergometry but not low-frequency pushing or overground tasks.

Keywords: Activity trackers; Disabled persons; Physical activity; Wheelchairs.

Publication types

Validation Study

MeSH terms

Exercise
Exercise Test
Fitness Trackers* / standards
Humans
Movement
Wheelchairs*