Invited commentary: places, people, and health.

A growing body of research in epidemiology and public health has examined how characteristics of the places where people live are related to a variety of health outcomes, including health-related behaviors (1–4), prevalence and incidence of disease (5–7), and mortality (3, 8–13). The paper by Balfour and Kaplan (14) in this issue of the Journal examines how neighborhood characteristics are related to functional loss in the elderly. The fact that health varies across geographic locations is well-established (15). The assumed explanation for these geographic differences (particularly in the modern era of epidemiology, with its emphasis on individual-level risk factors) has usually been that areas differ because of the characteristics of the people who live in them. In recent years, however, there has been renewed interest among social scientists in the ways in which neighborhood contexts may affect individual-level outcomes (16, 17). For example, neighborhood characteristics have been related to employment and single parenthood (18), violence (19), and child development (20). In public health, it has been argued that neighborhoods may also be relevant to health (21–25). Contextual and multilevel analyses have been used to investigate area effects on health after accounting for individual-level factors (26–28). The persistence of an independent area effect would suggest that things about the area itself are important to the health of its residents. Research on neighborhood effects on health has been part of resurgent interest in the social determinants of health and in moving beyond causal explanations that focus exclusively on the characteristics of people. However, the investigation of neighborhood effects is not only of academic interest: The demonstration of a causal link between neighborhoods and health would have implications for disease prevention and health policy. Although existing research has, for the most part, documented an independent effect of neighborhood socioeconomic characteristics on health after controlling for individual-level socioeconomic indicators (24, 28), many unresolved issues remain. The magnitude of neighborhood effects and their relative importance have varied (24, 28). More fundamentally , questions remain regarding whether these associations truly reflect causal processes. A key criticism, enunciated many years ago by Hauser (29) in his critique of contextual analysis, has to do with misspecification of the model at the individual level or residual confounding by individual-level variables. Existing work has attempted to address this critique through stratification and tight control for individual level variables (mainly individual-level socioeconomic indicators, which are perceived to be the key confounders …