Validation Of Visual Inspection For Determining Accelerometer Wear Days

Abstract

Visual inspection of accelerometer output may provide an alternative to needing participant-reported wear logs for ascertaining wear days when monitors are sent and returned by mail, for participants not completing logs. However, there are no data on the reliability and validity of visual inspection compared to wear logs. PURPOSE: To compare estimates of wear days and physical activity using visual inspection, with participant-reported wear logs. METHODS: Participants (N = 197, mean age = 71.0 years) were from the Women’s Health Study. In 2011-2014, women were mailed an accelerometer (ActiGraph GT3X+) to be worn for 7 days and then returned in the mail. They also indicated on a log the days they wore the monitor. Accelerometer days were classified as “worn” or “not worn” based on participant-reported wear logs. For each participant, we created a series of line charts of accelerometer counts over the course of the day (1 chart per day of accelerometer data). Two raters independently inspected these graphs to identify each day as ‘worn’ or ‘not worn’. We calculated the sensitivity and specificity for each rater, compared against log determined wear days, as well as the percent agreement between raters. We tested for differences in median wear time and physical activity estimates between visual inspection and wear logs using Wilcoxon Signed-Rank tests. RESULTS: The median (interquartile range) number of valid wear days for both visual inspection and log was 7 (7, 7). For rater 1, the sensitivity was 99.7% (95% confidence interval (CI): 99.2%, 99.9%); specificity, 97.2% (95.2%, 98.6%). For rater 2, the sensitivity was 99.7% (99.2%, 99.9%); specificity, 97.0% (95% CI: 94.9%, 98.4%). The inter-rater percent agreement was 99.5%. There was no significant difference in the median number of minutes of moderate-to-vigorous physical activity per day between the two methods (visual inspection: 20.7 (8.9, 37.4); wear logs: 20.0 (8.9, 37.4); P = 0.95). CONCLUSIONS: Visual inspection of accelerometer data is a valid and reliable alternative to participant-reported wear logs to determine accelerometer wear days, for those not providing logs, in direct mail study designs.