The spatial dimensions of neighborhood effects

In the decade or so of renewed interest in neighborhood contexts and health, significant progress has been made conceptualizing the relationships between the urban environment and public health. Applied research on the link between the environment and health remains limited by the way spatial concepts, such as "the neighborhood" or "the built environment" are operationalized. In this paper we argue that representations of these spatial concepts in statistical models should be based upon the individuals, the place, and the problem under study. Through a series of simulation experiments we describe the sensitivity of estimates of the association between neighborhoods and health to the operationalization of spatial concepts. We explore the practice of conducting the same analysis at multiple scales and find that using model fit to "discover" the spatial dimension is problematic. In sum, there is a gap between our understanding of how the environment influences health and spatial statistical modeling techniques. For quantitative spatial inquiry into the relationship between the neighborhood environment and health to be effective this gap must be closed.