Comparison of 3 different analytic approaches for determining risk-related active and sedentary behavioral patterns in adolescents.

BACKGROUND Much of the research conducted to date implies overweight youth exhibit uniform active and sedentary behavioral patterns. This approach negates the possibility that multiple co-occurring, and seemingly contrasting, behaviors may manifest within the same individual. We present a substantive dialogue on alternative analytical approaches to identifying risk-related active/sedentary behavioral patterns associated with overweight in adolescents. METHODS Comparisons were made among latent profile analysis (LPA), cluster analysis (CA), and multinomial logistic regression (MLR). A cross sectional sample of youth (N = 6603; 12-18 yrs) completed a questionnaire assessing: physical activity (PA); competing activities (COMP); and sedentary activities (SED). Demographics associated with PA (age, sex, BMI) were used as covariates/predictors. RESULTS Comparisons among methods revealed that LPA and CA detected subgroupings of behavioral patterns associated with overweight, each unique in regards to behaviors and demographic characteristics, whereas MLR results followed established associations of low PA and high SED without subgroup separation. CONCLUSIONS Use of LPA and CA provides a rich understanding of behavioral patterns and the related demographic characteristics. Decisions guiding the selection of analytical techniques are discussed.