Associations between Body Mass Index (BMI) and Accelerometer Time Series Data: National Health and Nutritional Examination Survey (NHANES) 2005-2006

Higher rates of impact physical activity (PA) are thought to promote weight loss and health. However, little cross-sectional epidemiological research exists relating average activity and weight, obesity or body mass index (BMI). Wearable technology provides a reliable measure of the frequency, duration, intensity, and timing of PA. The objective of this study was to examine the combined association between body mass index (BMI) and PA in a large cross-sectional study. The National Health and Nutrition Examination Survey (NHANES) has collected PA and demographic data of American participants in 2005. Linear regression was used to explore the association between BMI and accelerometry data with multiple covariates including age, race, and sex. In addition, a random forest model to predict BMI was implemented and the variable importance of PA was assessed relative to other predictors. Our results showed that BMI was associated with PA, in addition to race and age. Random forest variable importance metrics suggested age as having the greatest impact on predicting BMI followed by five principal component directions of variation of daily PA. In summary, the results show a consistent and strong relationship between physical activity and body mass index