Methods for Activity Monitor Validation Studies: An Example With the Fitbit Charge

Abstract

Purpose: To advance research practices with consumer monitors, standard validation methods are needed. This study provides an example of best practices through systematically evaluating the validity of the Fitbit Charge (FBC) under free-living conditions using a strong reference measure and robust measurement agreement methods. Methods: 94 healthy participants (Mage 41.8 ±9.3 yrs) wore a FBC and two research grade accelerometers (Actigraph GT3X and activPAL) as they went about normal activities for a week. Estimated daily minutes of moderate to vigorous physical activity (MVPA) from the FBC were compared against reference estimates obtained from the Sojourns Including Posture (SIP) methodology, while daily step counts were compared against the activPAL. Results: Correlations with reference indicators were high for average daily MVPA (r = 0.8; p < .0001) and steps (r = 0.76; p < .0001), but the FBC overestimated time spent in MVPA by 56% and steps by 15%. The mean absolute percent errors of MVPA and steps estimated by FBC were 71.5% and 30.0%, respectively. Neither of the MVPA and step estimates from the FBC fell into the ±10% equivalence zone set by the criterion. The Kappa statistics of the classification agreement between the two MVPA assessment methods was 0.32 with a low sensitivity of 30.1% but a high specificity of 96.7%. Conclusion: The FBC overestimated minutes of MVPA and steps when compared to both reference assessments in free-living conditions. Standardized reporting in future studies will facilitate comparisons with other monitors and with future versions of the FBC.