How many hours are enough? Accelerometer wear time may provide bias in daily activity estimates

Background: There is little consensus on how many hours of accelerometer wear time is needed to reflect a usual day. This study identifies the bias in daily physical activity (PA) estimates caused by accelerometer wear time. Methods: 124 adults (age = 41 ± 11 years; BMI = 27 ± 7 kg·m^-2) contributed approximately 1,200 days accelerometer wear time. Five 40 day samples were randomly selected with 10, 11, 12, 13, and 14 h·d^{-1 of} wear time. Four semisimulation data sets (10, 11, 12, 13 h·d^-1) were created from the reference 14 h·d^-1 data set to assess Absolute Percent Error (APE). Repeated-measures ANOVAs compared min·d^-1 between 10, 11, 12, 13 h·d^-1 and the reference 14 h·d^-1 for inactivity (<100 cts·min^-1), light (100-1951 cts·min^-1), moderate (1952-5724 cts·min^-1), and vigorous (≥5725 cts·min^-1) PA. Results: APE ranged from 5.6%-41.6% (10 h·d^-1 = 28.2%-41.6%; 11 h·d ^-1 = 20.3%-36.0%; 12 h·d^-1 = 13.5%-14.3%; 13 h·d^-1 = 5.6%-7.8%). Min·d^-1 differences were observed for inactivity, light, and moderate PA between 10, 11, 12, and 13 h·d^-1 and the reference (P < .05). Conclusions: This suggests a minimum accelerometer wear time of 13 h·d^-1 is needed to provide a valid measure of daily PA when 14 h·d^-1 is used as a reference.