Invited Commentary: Advancing Theory and Methods for Multilevel Models of Residential Neighborhoods and Health

Abstract

Received for publication July 29, 2002; accepted for publication October 11, 2002. In less than 10 years, the field of epidemiology has been transformed. During this time, multilevel modeling has gone from a little-known and perhaps even unwelcome method of analysis to a household name. Multilevel models appeared much earlier in the fields of demography, sociology of education, and criminology (refer, for example, to Raudenbush et al. (1), Mason et al. (2), and Nuttal et al. (3)), but public health and epidemiology have lagged behind in conceptualizing and measuring how contexts affect individual-level health risks and outcomes. The growth in the publication of multilevel studies has paralleled and contributed greatly to the reemergence of social epidemiologic research (4–11). Multilevel models assessing the effects of neighborhood residential environments on health outcomes have been the most common type of contextual study to date. Research has examined associations between neighborhood characteristics, frequently socioeconomic position, and a variety of health outcomes, including perinatal health (12–15), mortality (16, 17), health behaviors (18, 19), women’s health (20, 21), heart disease (22–24), disability (25), and child health (26) among others. Furthermore, many reviews have been published concerning the statistical methods and rationale for, as well as the quality of, the existing multilevel studies on health (27–34). Yet, it is perhaps still premature to celebrate the exponential growth of multilevel epidemiologic studies on neighborhood residential effects on health. On the one hand, the growth in publication of these studies reflects increased interest in and support for social epidemiologic approaches. This increased support is encouraging in light of recent controversies about the legitimacy of social epidemiology as a subfield of epidemiology (35–40). On the other hand, these studies have emerged in public health and epidemiology without the appropriate theoretical and methodological foundations to guide their implementation. Hence, despite dozens of multilevel studies of neighborhoods and health, we still lack a clear picture of the intervention and policy implications of this body of work. In this issue of the Journal, Buka et al. (41) examine whether levels of neighborhood social support affect the birth weights of African-American and White mothers residing in Chicago, Illinois. Their paper contributes to the small, but growing literature on multilevel analyses of neighborhood residential effects on perinatal outcomes (12–15, 42). It offers several methodological advancements over previous multilevel studies on neighborhood residence and health. Moreover, it highlights additional theoretical and methodological challenges that multilevel analysis, given its recent appearance in public health, must overcome if significant gains are to be made in understanding disease etiology and designing appropriate interventions and policies to prevent adverse health. These theoretical and methodological issues are the topic of this commentary. Perhaps the most pressing issue standing in the way of progress in multilevel research is lack of theory (i.e., system of hypotheses) on the mechanisms by which neighborhood environments affect health risks, protective factors, and outcomes. This is where researchers in other fields such as sociology of crime (43–45) and community psychology (46–48) have advanced well beyond public health scientists. While epidemiologists can draw from theories on mechanisms from these other fields, we epidemiologists will have to develop and identify our own theoretically informed mechanisms of how neighborhood environmental processes affect health. Moreover, given the lack of strong public health hypotheses regarding neighborhood effects, qualitative research should be undertaken and used to inform and identify mechanisms by which neighborhoods impact health risks and outcomes (49–51). In their focus on neighborhood social support, Buka et al. (41) rely on the literature concerning individual-level social support and birth weight to inform their hypotheses regarding how high levels of neighborhood social support are related to higher birth weights. While this step is a good start, we are still left with uncertainty regarding the conceptual and mechanistic links between neighborhood support—defined as social cohesion, trust, and reciprocated ex-change—and increased birth weight. One limitation is that the literature used to inform these proposed mechanisms on social support at the individual level is equivocal (52, 53); some studies have reported no association, including several large randomized trials of social support and birth weight (54). As noted by Buka et al., increased fetal growth and birth weight were associated with intimate partner or family support only and not other types of support (e.g., instrumental support) (53). Given the inconsistent evidence among studies o