Abstract
BACKGROUND: Studies evaluating the impact of the neighbourhood food environment on obesity have summarised the density or proximity of individual food outlets. Though informative, there is a need to consider the role of the entire food environment; however, few measures of whole system attributes have been developed. New variables measuring the food environment were derived and used to study the association with body mass index (BMI). METHODS: Individual data on BMI and sociodemographic characteristics were collected from 48 482 respondents of the 2002-2006 community health survey in New York City and linked to residential zip code-level characteristics. The food environment of each zip code was described in terms of the diversity of outlets (number of types of outlets present in a zip code), the density of outlets (outlets/km(2)) and the proportion of outlets classified as BMI-unhealthy (eg, fast food, bodegas). RESULTS: Results of the cross-sectional, multilevel analyses revealed an inverse association between BMI and food outlet density (-0.32 BMI units across the IQR, 95% CI -0.45 to -0.20), a positive association between BMI and the proportion of BMI-unhealthy food outlets (0.26 BMI units per IQR, 95% CI 0.09 to 0.43) and no association with outlet diversity. The association between BMI and the proportion of BMI-unhealthy food outlets was stronger in lower (