Race/ethnicity as a confounder in the association between adult obesity and selected rural populations in Texas

Background: Rural-urban differences in obesity in the United States have been documented. Methods: We investigated the effects of socio-demographic factors on rural-urban differences in obesity (body mass index ≥30 kg/m2), as part of the evaluation of a Centers for Disease Control and Prevention (CDC)-funded community transformation grant. 2,330 adults in three urban and three rural Texas' counties, selected using probability-based dual-frame random digit dialing, were surveyed via computer-assisted telephone interviews between September and November 2012. A series of logistic regression models between obesity and county (with an urban county as the reference), were run in a forced forward step-wise manner, controlling for demographic variables (race/ethnicity, age, gender, income, education), fruit/vegetable consumption, and physical activity. Survey weights correcting for the sample design, non-response, and post-stratification, were applied in obtaining the population-based odds ratio (OR) estimates and confidence intervals (CIs). Results: In the unadjusted model, adults in two rural counties were more likely to be obese than those in the reference urban county, with ORs and 95% CIs of 2.07 (1.35-3.19) and 2.06 (1.34-3.19). This association was eliminated in the model with race/ethnicity, but persisted in concurrent models with the other control variables. In the full model with all control variables, Hispanics as well as Others (inclusive of non-Hispanic blacks) had higher odds of being obese than non-Hispanic whites, with ORs and 95% CIs of 2.80 (1.67-4.66) and 1.88 (1.02-3.46), respectively. Conclusions: Socio-demographic differences, such as race/ethnicity, may drive the differences in obesity, and should be targeted in obesity control campaigns.