Methodological Approaches to Account for Residential Self-Selection and Time-Varying Confounding in the Association Between the Neighborhood Environment and Cardiovascular Disease.

Abstract

Despite the growing body of research investigating the relationship between the neighborhood environment and cardiovascular disease-related outcomes, many studies have methodological limitations. A major challenge in the study of neighborhood exposures and health-related outcomes arises due to the possibility that individuals may select where to live based on individual behaviors or preferences (i.e. residential self-selection). Another challenge is the need to account for time-varying confounding, which may result when neighborhood exposures and individual-level factors vary over time, possibly acting as both confounders and mediators. This dissertation used longitudinal data from the Multi-Ethnic Study of Atherosclerosis (MESA) to examine: 1) whether individual physical activity and diet quality were related to selection of neighborhoods based on availability of physical activity and favorable food resources; 2) whether changes in the availability of physical activity resources was related to changes in individual physical activity; and 3) whether neighborhood socio-economic composition was related to incident cardiovascular disease after accounting for time-varying confounding. The findings show that among individuals who moved, those who were more physically active moved to neighborhoods with greater availability of physical activity resources. The results also show that changes in physical activity resources over time were related to changes in physical activity over time, with a stronger association observed among older adults. Overall, these results suggest that individuals select their neighborhoods based on physical activity behaviors, but at the same time changes in physical activity resources over time are associated with simultaneous changes in the physical activity of residents. The estimation of the relationship between neighborhood socio-economic composition and incident cardiovascular disease revealed that increasing neighborhood deprivation was associated with a greater risk of incident cardiovascular disease. This association was similar after using methods to account for time-variation in both the exposure and covariates. Using longitudinal methods to account for residential selection and time-varying confounding in the context of neighborhoods and health provides more robust evidence that can inform policies aimed at improving residential environments to help promote healthy living.