Multilevel models assessing the impact of contextual and individual characteristics on obesity epidemic in Argentina

POU SA; VELÁZQUEZ GA; DIAZ, MARIA DEL PILAR; ABALLAY LR; BECARIA COQUET J

Barcelona

Conferencia; IBS XXIXth International Biometric Conference; 2018

International Biometric Society

Introduction: Previous epidemiological research has clearly established associations between obesity and individual behaviour linked to lifestyles. Nowaday it is also highlighted the environmental aetiology of the obesity epidemic. Research studying contextual effects on health suggest that multilevel regression analysis contributes to our understanding of the distribution and determinants of geographical, social, and individual disparities in health status.Objective: To assess the impact of contextual and individual characteristics on obesity in Argentina (2013) using a multilevel modelling approach.Methods: From the Third National Survey of Risk Factors for Chronic Diseases dataset (n= 31,471 adults), we analysed associations between obesity status outcome and some individual and contextual selected variables, taking into account its spatial clustering (province as cluster variable, n=24). We performed the analysis in three steps: 1) a variance component model (Model I) and then, a two-level random-intercept model with individual variables (Model II); 2) contextual variables (urbanization and poverty indicators), including an interaction terms as fixed effects (Model III), and 3) random-intercept and random-coefficients (Model IV).Results: Model I allowed to identify a significant contextual component (sd=0.16, LR test vs. logistic model: 57.27, p