D-47 Free Communication/Slide - Free-Living Validation Studies Thursday, June 1, 2017, 3:15 PM - 5:00 PM Room: 103
PURPOSE: Research is accumulating regarding the accuracy of wrist-worn consumer activity monitors in controlled lab settings. However, there is a lack of evidence of validity under free living conditions. The study investigated the accuracy of Fitbit Charge (FBC) for estimating minutes of moderate and vigorous physical activity (MVPA) compared to research grade accelerometers. A secondary purpose was to evaluate the accuracy of steps/day.
METHODS: Ninety-four healthy men and women (mean age 41 ± 9 years) wore a FBC as part of a 12-week intervention. The participants were also asked to wear research grade accelerometers concurrently as the criterion physical activity measure during the last week of intervention. A minimal of 10 hours/day wear-time was applied to both methods. Estimated daily MVPA and steps taken from the FBC were compared against the criterion using indicators of Pearson correlation, mean absolute percent errors, mean percent errors, and equivalence testing. The activity classification agreement of MVPA at the minute level was evaluated with Kappa, sensitivity and specificity.
RESULTS: Complete data were collected on 61 individuals with an average of 5.5 days of wear-time. The average daily MVPA measured by accelerometer was 76.3 minutes compared to a higher value, 118.7 minutes, estimated by the FBC. The correlation between the two methods was 0.8 (p<0.0001). The mean absolute percent errors and mean percent errors were 68.2% and -64.2% indicating a consistent overestimating MVPA by the FBC. The average daily steps were 8,897 and 7,716 measured by FBC and accelerometer, respectively, with a correlation of 0.76 (p<0.0001). The mean absolute percent errors and mean percent errors of steps estimated by FBC were 30.0% and -20.1%. Neither of the MVPA and steps measured by FBC fell into the ±10% equivalence zone set up by the accelerometer. The Kappa statistics of the classification agreement between the two methods was 0.32 with a low the sensitivity of 30.1% but a high specificity of 96.7%.
CONCLUSIONS: This FBC estimated substantially higher minutes of MVPA in free living conditions among healthy adults and significantly higher steps compared to research grade accelerometer. The researchers who use FBC to track physical activity need to interpret their results with caution.