While body weight has multiple determinants, from biological to environmental, the development of an obesity epidemic over the last several decades implicates changes in environmental conditions above other factors. An environmental factor that has received considerable attention is the role of the food environment in obesity and, specifically, the role of fast-food establishments. Evidence regarding the association between an individual's food environment and his or her BMI remains inconclusive. Prior studies are limited by cross- sectional design, by incomplete or poor ascertainment of proximity to food establishments, or by a narrow focus on the residential food environment rather than a more comprehensive assessment that accounts for exposures at home and work. Using a sample that combines subjects from the Framingham Heart Study (FHS) Offspring and Omni Cohorts, we will build on prior work by 1) examining nearly 40-year longitudinal data on both BMI and the exposure to food establishments; 2) using multiple sources of data on food establishments; 3) including residential and workplace neighborhoods as areas of exposure with longitudinal data to account for residential and workplace mobility; 4) precisely measuring distance between residential and workplace addresses and food establishments over time (accounting for the opening and closing of establishments across time); and 5) identifying food establishments along common routes to and from work. Data on food establishments will come from multiple sources, including local Boards of Health, historical Yellow and White Pages, and a commercial database. For the primary analyses, the sample will include subjects who were living in a 4 town, geographically-concentrated area with 17,446 observations from 3,160 FHS Offspring Cohort subjects and additional subjects and observations from the FHS Omni Cohort. We will examine associations of BMI with proximity to 6 types of food establishments, including fast-food restaurants, full-service restaurants, bakeries/coffee shops, convenience stores, grocery stores, and chain supermarkets, from 1971 to 2008. We also will determine the proportion of the variation in BMI over time that arises from clustering at the level of residential or workplace neighborhoods (in this case, census tracts) for both this geographically-concentrated sample as well as for a sample that combines the entire FHS Offspring and Omni Cohorts. To facilitate these analyses, we will develop new statistical software capable of handling multilevel data where repeated observations are made on individuals whose group membership (in our case, residential and workplace neighborhoods) may change over time. This software will accommodate a family of cross- classified multilevel random effect models using Bayesian methods for model estimation. Overall, the work from this grant will provide important answers to questions regarding the association of the food environment with BMI. The results will help determine whether creating healthier food environments is a strategy that has the potential to improve health. We will disseminate the methods and software for use in studies with similarly complex data.
Among subjects in the Framingham Heart Study Offspring and Omni Cohorts; we propose to study the association of residential and workplace proximity to food establishments with body mass index; from 1971 to 2008. We will develop new data on food establishments and employers and will create new statistical software to facilitate analyses of complex longitudinal and geographic data. This study will provide important information to help policymakers determine whether creating healthy food environments can improve health.
