Do Active Communities Support Activity, or Support Active People?: Residential Self-selection in the Estimation of Built Environment Influences on Physical Activity

In a growing body of research, the built environment, composed of neighborhoods, roads, and buildings in which people live, work, and play, has been shown to be related to physical activity. While promising for physical activity promotion, substantial limitations must be addressed before built environment research can adequately inform policy recommendations. To this end, we focused on three methodological challenges of particular concern for built environment research: (1) quantitative characterization of a complex environment, (2) confounding by other inter-related environment characteristics, and (3) residential self-selection bias resulting from systematic sociodemographic and behavioral differences among individuals selecting different types of neighborhoods. We used data from the National Longitudinal Study of Adolescent Health, a nationally representative cohort of over 20,000 adolescents followed over seven years into young adulthood. A large scale geographic information system (GIS) linked community-level built (e.g., recreation facilities, land cover, street connectivity) and socioeconomic (e.g., median household income, crime rate) environment characteristics to individual-level sociodemographic and behavioral data in space and time. Using factor analysis, we identified several built and socioeconomic environment constructs. We found that the socioeconomic environment is a potentially important confounder of built environment and physical activity relationships, but is often omitted in existing studies. We also show that commonly used built environment measures may act as proxies for complex environment constructs; for example, intersection density, typically used to indicate street connectivity, may represent general density of development. Lastly, in longitudinal analysis, we observed an increase in physical activity among males with a greater number of physical activity-related pay facilities and a decrease in physical activity among males and females with higher neighborhood crime rate. Other built environment characteristics were unrelated to physical activity. Additional analysis suggests that residential self-selection can attenuate, as opposed to magnify, environment-physical activity associations. This research revealed complexities in the environment that have implications for analysis and interpretation of this and related research. Findings suggest that built environment characteristics may influence physical activity, yet raises additional questions to be answered by an evolving field