Background: A limitation of accelerometer measures of moderate to vigorous physical activity (MVPA) is nonwear time. Nonwear-time data is typically deleted prior to estimating MVPA. In this study, we used an approach that used sociodemographic, health, and time data to guide the imputation of nonwear-time data. We determined whether imputing nonwear-time data influences estimates of MVPA and the association between MVPA, body mass index, and blood pressure.
Methods: Seven days of accelerometer data were collected on 332 children aged 10-13 years. MVPA was estimated in a "nonimputed dataset," wherein nonwear-time data were deleted prior to estimating MVPA, and in an "imputed dataset," wherein nonwear-time data were imputed using sociodemographic and health characteristics of participants and time characteristics of the nonwear period prior to estimating MVPA.
Results: Nonwear time represented 7% of waking hours. Average MVPA estimates did not differ in the nonimputed and imputed datasets (56.8 vs 58.4 min/d). The strength of the relationship between MVPA and the 2 health outcomes did not differ in the nonimputed and imputed datasets.
Conclusions: Studies achieving high accelerometer wear-time compliance can obtain MVPA estimates without substantial bias if they use the traditional approach of deleting nonwear-time data.
Keywords: biostatistics; child; motor activity.