The association between the physical environment and the risk of being overweight or obese are difficult to estimate because residents are not randomly distributed by neighborhood. When neighborhood selection bias exists, estimates of the association between neighborhoods and BMI that do not account for its effects will misstate the strength of the causal relationship.
Our aim i s to characterize the contributions of causal and selection explanations regarding the association between neighborhoods and BMI, using non-experimental data, by testing three hypotheses: 1. Walkable neighborhoods, as measured by population density, pedestrian-friendly design, and land-use diversity are associated with lower levels of individual BMI. 2. Individual characteristics, as measured by age, sex, race, family history of obesity, influence individual BMI. 3. The size of the causal effects identified in hypotheses 1 and 2 will be attenuated after adjusting for the effects of non-random selection into neighborhoods. To test these hypotheses, we will implement three methods for assessing the influence of causal effects in the presence of non-random residential selection. These methods are (1) structural equations models with instrumental variables; 2) counterfactual approaches using propensity scores; and 3) longitudinal mover-stayer models. Results of these analyses will contribute crucial information to understanding the relationship between neighborhoods and the individual risk of overweight or obesity. This study relies on an unparalleled population-based data source, the Utah Population Database (UPDB). The vast genealogical records in the UPDB are linked to state-wide administrative and vital records (driver licenses, birth certificates) that contain longitudinal data on height and weight (used to construct BMI) and residential location. The UPDB is also linked to U.S. Census, state, and county information on neighborhood characteristics using Geographic Information Systems (GIS) databases. The study design is responsive to several goals described in (PA-06-151) """"""""Secondary Analyses in Obesity, Diabetes, Digestive & Kidney Diseases."""""""" Using the UPDB for this project is cost-effective because it already links longitudinal data on demographic and socio-cultural variables, BMI, residential history, family measures, and vital records.
This study seeks to determine how neighborhood characteristics affect the risk of being overweight or obese. The objective of this study is to determine if this association is causal or whether it is because certain individuals choose to move to neighborhoods so that the association is observed. Achieving this objective will help us to better understand the ways in which policy-makers can alter neighborhood environments in order to reduce overweight or obesity risks among Americans. ? ? ? ?
