Multilevel functional methods for modeling actigraphy data and its application to predicting mortality in the US population

We will present the results of analysis of physical activity data collected with accelerometers on 10000+ subjects in National Health and Nutrition Examination Survey (NHANES), a nationally representative sample of the US population. The data have been analyzed with recently developed multilevel functional data analysis approaches that focus on modeling variability in 24-hour diurnal patterns by i) separating and quantifying the systematic and random circadian patterns of physical activity, and ii) modeling those patterns as functions of age, gender, and dominant comorbidities. We will demonstrate that these patterns are accurate predictors of the follow-up mortality in NHANES 2003-2006 cohorts.