Invited Commentary: The Enduring Role of “Place” in Health—A Historic Perspective

Abstract

In public health, it has long been observed that “place”—specifically, where one lives—affects individual health, with the main research question distinguishing between the effects of “context” (defined as area characteristics) and “composition” (the characteristics of inhabitants) on health outcomes. There have been many studies in which the spatial patterning of disease has been explored, but they were often ecological in design, used broad census geographic levels, lacked individual-level data, or when available, did not simultaneously analyze community- and individual-level risk factors using appropriate modeling techniques. The paper by Diez-Roux et al. (Am J Epidemiol. 1997;146(1):48–63) represents an important expansion of the literature in terms of analytic methods used and level of geography studied. The authors demonstrated that both neighborhood- and individual-level measures of socioeconomic status work together to play an important role in shaping disease risk. Analyses incorporating both levels of data have the potential to provide epidemiologists with a deeper understanding of the divergent pathways via which neighborhood affects health.

In public health, it has long been observed that “place” (specifically, where one lives) affects individual health. In the 1970s, Peter Rossi described the “local community” as the setting for major life events over the life course:

. . . [it] supplies to its individual citizens the medical facilities in which he is born, the schools in which he is taught, the housing in which he lives, the social milieu in which he finds his mate and sets up his household, the factories and businesses in which he finds employment and finally the cemetery in which he is buried (1, p. 89).

Centuries earlier in his book On Airs, Waters, and Places (2), Hippocrates stressed critical appraisal of the environment as part of the medical examination. Specifically, he wrote:

Whoever wishes to investigate medicine properly. . . when one comes into a city to which he is a stranger, he ought to consider the rising of the sun. . . And concerning the waters which the inhabitants use. . . And the ground, whether it be naked and deficient in water. From these things he must proceed to investigate everything else (2).

Later, August Hirsch, a 19th century German physician, medical historian, and major contributor to the field of medical geography, emphasized the need to investigate whether the symptoms attributed to a disease occurred in the same forms “at all times and in all places” (3, p. 104) and what effect the external environment had on the disease form. All 3 writers essentially conveyed the same underlying sentiment, which continues to motivate research studies today: Are the roles played by community attributes simply a backdrop to the lives of residents or do they significantly contribute to the health and wellbeing of residents, above and beyond individual characteristics?

In 1997, Diez-Roux et al. (4) took an important step toward answering this question and distinguishing between the roles of “context” (defined as area characteristics) and “composition” (the characteristics of inhabitants) in health outcomes (5). In this commemorative edition of the American Journal of Epidemiology, we reflect upon the evolution of the study of neighborhoods and health and on the important contribution of Diez-Roux et al. to the field.

There have been many studies in which investigators have explored the patterning of disease according to place. Among them was the 1916 investigation of pellagra incidence in South Carolina by Goldberger et al. (6). Pellagra, a disease caused by a vitamin deficiency, had long been associated with low socioeconomic status (SES), but Goldberger et al. went further and compared 2 villages with comparable SES yet differential incidences of disease. This descriptive study illuminated the complex role of community food availability—the only difference between the 2 villages—as an independent macro-level influence on risk of disease incidence that transcended individual-level income. Specifically, those who, in theory, could afford to buy food had higher rates of disease if they resided where food was unavailable. More recently, in a series of ecological analyses, Wing et al. (7) explored rates of ischemic heart disease mortality according to geography. Using data from the National Center for Health Statistics (1968–1978) combined with US Census population estimates, county-level death and population counts were combined to create state economic areas, which were further divided into either metropolitan or nonmetropolitan counties. Wing et al. found that among white males, declines in ischemic heart disease mortality rates occurred earlier in metropolitan economic areas than in nonmetropolitan economic areas, and the Northeast and Pacific states showed earlier declines than did the South and Midwestern sections of the country (7). Hypothesizing that the social and economic characteristics of communities may have played a role in early or late decline in rates of mortality from ischemic heart disease, the authors examined the influence of 3 census-level indicators of socioeconomic development and community resources: income, educational level, and occupation. Late onset of decline was observed among participants in the lowest category of each variable. Specifically, persons in state economic areas with low levels of each factor were 2–10 times more likely to experience later onset of disease decline than were those in areas with high levels of those attributes (8). These and the previous findings were also observed among white women (9). Logue and Jarjoura (10) examined heart disease mortality among adults older than 34 years of age who resided in 8 Ohio counties by regressing age- and sex-adjusted census tract–specific (1980) heart disease rates on tract-level social class characteristics defined according to occupational class. They found that compared with persons in the upper middle/middle class, lower middle class residents experienced a heart disease mortality rate that was 1.9 times the higher, whereas the working poor experienced a heart disease rate that was 4.4 higher. Haan et al. (11) compared 9-year mortality rates (1965–1974) according to poverty area residence among residents in Alameda County, California, and found a 50% higher risk of mortality among those living in poverty areas compared with those residing in nonpoverty areas. Unlike in the previous studies cited above, Haan et al. were able to perform multivariate adjustment for a number of baseline individual-level variables, such as age, sex, race, high blood pressure, and diabetes. The observed higher mortality risk persisted after multivariate adjustment.

The paper by Diez-Roux (4) represents an important expansion of this literature on several fronts. The first involves the analytic methodology used. Despite the previous findings in support of a relationship between geography and cardiovascular health, there had been no analyses exploring both the independent and combined effects of community- and individual-level risk factors. The simultaneous inclusion of both area- and individual-level variables in regression equations allows for exploration of area associations after controlling for individual-level covariates, as well as the examination of individual-level variables as modifiers of the area effect (12). Used widely in the fields of sociology and demography, such analyses, also referred to as multilevel or hierarchal modeling, involve individuals nested within areal units (13). Using this analytic method, Diez-Roux et al. (4) explored whether neighborhood characteristics (income, educational level, occupation, and housing value) were related to the prevalence rates of coronary heart disease and associated risk factors (e.g., smoking, elevated blood pressure, and elevated cholesterol level) after adjustment for individual-level measures of SES (family income, completed education, and occupation). They were also able to explore the data by study site, sex, and race, with further stratification by sex within groups delineated by both study site and race. The data utilized were from the Atherosclerosis Risk in Communities Study (14), which had a geographically and racially diverse cohort comprising 15,792 adults aged 45–64 years who resided in 4 US communities (Forsyth County, North Carolina; Jackson, Mississippi; Minneapolis, Minnesota; and Washington County, Maryland).

The second contribution involves the level of geography. The geographic areas investigated previously involved county-level (7–9) and census tract–level (10, 11) data. The choice of these areal units is often practical when large quantitative analyses require reliance on easily linked standard administrative data. Nevertheless, the level of available data may not represent the appropriate geographical context (12), and identification of geographically relevant areas remains a challenge for studies of neighborhoods and health (15). For example, counties may provide important geographical context if outcomes are potentially related to policies or economic structure (e.g., school districts, municipal services). In other circumstances, the environment immediately surrounding the residence (e.g., city blocks) where social encounters or services are obtained may be most important (4, 12, 16). In their analysis, Diez-Roux et al. (4) explored area characteristics at the level of the census block-group, an administrative unit that averaged approximately 1,000 individuals, as a proxy for neighborhood (17). Compared with census tracts, block groups are more homogenous and are believed to estimate the immediate socioeconomic environment of residents. Additionally, the block group can potentially identify pockets of affluence or deprivation that may otherwise be masked in the more heterogeneous census tracts (18). Indeed, the study findings of Diez-Roux et al. reflect such detail and represent the third contribution to the literature.

The overall study results among whites differed from those among blacks. In analyses by study site, neighborhood affluence was inversely associated with coronary heart disease, smoking, systolic pressure, and cholesterol in Forsyth County, Minneapolis, and Washington County (all white), results that persisted after adjustment for individual-level social class variables. Although this inverse association largely held true for the Jackson site (all black), blacks showed variability by sex and neighborhood SES in ways not observed among whites. For example, there was a curvilinear relationship between neighborhood SES and cholesterol level for both women and men in Jackson, where cholesterol was highest in the middle category of neighborhood SES. Jackson men also had a lower overall prevalence of coronary heart disease than did women, but once again, prevalence was highest in the intermediate neighborhood SES category. As possible explanations for the observed cholesterol results, the authors hypothesized that increases in income can lead to increased psychosocial stress due to status incongruity (19, 20) or that increased neighborhood poverty could lead to decreased survival of individuals with coronary heart disease (21), factors that continue to contribute to health disparities.

Since the initial publication of the article by Diez-Roux et al. in 1997, there have been innumerable studies in which researchers explored multiple aspects of neighborhoods in relation to health outcomes. Hierarchical modeling remains a constant but has been joined by geographic information systems and spatial analysis techniques (22). The types and sources of community-level data have also evolved to include measures of the physical and social environments, such as air pollution, land use, transportation, crime statistics, and family structure. Finally, experimental approaches to the study of neighborhood and health have been conducted, most notably the Moving to Opportunity Study (23), in which differential results among black youth according to sex have been highlighted.

In conclusion, Diez-Roux et al. (4) demonstrated that working together, both neighborhood- and individual-level measures of SES play important roles in shaping disease risk. Such analyses have the potential to provide epidemiologists with a deeper understanding of the divergent pathways by which neighborhood affects health.