JMIR Publications

ABSTRACT

Background:

Smart watches (SWs) can serve as a tool to counteract obesity and to promote physical activity in wheelchair users (WCU). For estimated energy expenditure (EE) in WCU, the wheelchair-specific Apple watch (AW) series 1 had an error of ~30%, and the corresponding error for HR provided by the Fitbit Charge 2 was ~10-20%. While this has not yet been investigated, one would expect companies to further improve the accuracy of their EE and HR estimation algorithms.

Objective:

To assess the accuracy of the AW Series 4 (wheelchair-specific setting), and the Fitbit Versa (treadmill running mode) for estimating EE and HR during wheelchair propulsion at different intensities.

Methods:

Data of 20 manual WCU (11 males, 9 females, body mass: 75 ± 19 kg) and 20 people without a disability (PWOD) (11 males, 9 females, body mass: 75 ± 11 kg) were included in the study. Criterion EE data (kcal·min-1) were calculated from gas exchange data of an ergospirometer and criterion HR data (beats·min-1) were measured with a Polar H10 sensor. Treadmill wheelchair propulsion was performed on three separate test days (0.5%, 2.5% or 5% incline). Each day consisted of three 4-min stages at increasing speed. The mean absolute percentage error (MAPE) was used to indicate the absolute agreement between criterion device and the SWs for EE and HR. Additionally, linear mixed model analyses were used to assess if the SW error in EE and HR increased with exercise intensity, and any effects from group and sex. Interclass correlation coefficients (ICCs) were used to assess relative agreement between criterion devices and SWs for EE and HR.

Results:

The AW underestimated EE (MAPE±SD: 29.2±22% for WCU, 30.0±12% for PWOD), while the Fitbit overestimated EE (73.9±57% for WCU, 44.7±38% for PWOD). Both SWs underestimated HR (MAPE±SD for AW: 8.5±10% for WCU, 8.1±14% for PWOD; for Fitbit: 17.4±12% for WCU, 14.3±11% for POWD). The device error for EE and HR increased with intensity for both SWs (all comparisons: P<.001), and the only significant difference between groups was found on HR from the AW (-5.27bpm for WCU, P=.02). There was a significant effect of sex for EE, with the AW having worse accuracy for females (-0.69 kcal·min-1, P<.001) and the Fitbit better accuracy for females (-2.08, P<.001). For HR, sex-differences were found only for the AW with a smaller error for females (5.23 bpm (P=.02)). ICCs showed poor to moderate relative agreement for both SWs apart from two stage-incline combinations, with the AW ranging from 0.12-0.57 for EE and 0.11-0.86 for HR. The Fitbit had corresponding ranges of 0.06-0.85 for EE and 0.03-0.29 for HR.

Conclusions:

Neither the AW nor Fitbit were sufficiently accurate for estimating EE or HR during wheelchair propulsion.