Clustering physical activity phenotypes using the ATLAS index on accelerometric data from an epidemiologic cohort study.

While many studies have identified associations of physical activity with both mortality and morbidity, the effects of different dose-patterns are still unclear. Employing the recently introduced ATLAS index, the aim of the research work for this paper was to investigate whether physical activity phenotypes can actually be extracted from a large-scale epidemiological accelerometer data set, and if these match the ones proposed. The ATLAS parameters were computed for 6386 data sets from the NHANES 2005-2006 cohort study, and x-Means clustering was performed on the results. Thus, four distinct clusters were identified, named: 'insufficiently active', 'irregularly active', 'busy bee' and 'physical worker/weekend warrior'. In conclusion, it is possible to identify different activity phenotypes using the ATLAS index. More research is necessary with regard to the regularity component and the relation of activity patterns and medical parameters in the groups identified.