Just as risk factor profiles have been developed based upon our knowledge of how individual-level determinants affect health outcomes—it may also be important to develop profiles using information about the contextual characteristics of communities. In collaboration with the Centers for Disease Control and Prevention, we have undertaken a project to address the conceptualization and assessment of community contextual characteristics that are believed to impact population health and disparities in population health. Our goals include (1) providing a conceptualization of which contextual characteristics could plausibly affect patterns of population health; and (2) developing an extensive data library of geocoded datasets containing data to measure community contextual characteristics.

The main purpose of this article is to share ideas about how to conceptualize the contextual characteristics of communities that may affect health, and to present this work as a resource for public health research and advocacy. We describe the consultative process we followed to identify 12 overarching dimensions of a community contextual health profile with specific subcomponents and indicators within each dimension. We then present the rationale supporting each dimension and provide examples of the data sources assembled in our data library. We conclude with a discussion of three important research approaches for addressing the impact of community contextual characteristics on population health and health disparities.

In recent years, there has been growing research interest in examining associations between characteristics of places and a wide range of health outcomes including all-cause mortality (Lynch et al. 1998; Waitzman and Smith 1998a, 1998b; Bosma et al. 2001; Ross et al. 2000), cardiovascular disease and mortality (Diez-Roux et al. 2001; Casper et al. 1999; Armstrong et al. 1998; LeClere et al. 1998), tuberculosis (Barr et al. 2001; Acevedo-Garcia 2001), injury (Cubbin, LeClere, and Smith 2000; Reading et al. 1999), suicide (Kaplan and Geling 1999; Whitley et al. 1999), AIDS (Zierler et al. 2000), self-rated health (Subramanian, Kawachi, and Kennedy 2001; Malmstrom, Sundquist, and Johansson 1999), mental illness (Ross 2000; Aneshensel and Sucoff 1996), and adverse birth outcomes (Gorman 1999; Matteson, Burr, and Marshall 1998; Collins and David 1997). There has also been increasing awareness that health risks traditionally seen as arising from individual lifestyle choices are strongly influenced by the characteristics of the environments in which individuals grow and live out their daily lives (Lynch and Kaplan 2000; Lynch, Kaplan, and Salonen 1997; Macintyre, Ellaway, and Cummins 2002). For example, unhealthy dietary patterns and obesity (Diez-Roux et al. 1999; Ellaway, Anderson, and Macintyre 1997), smoking (Duncan, Jones, and Moon 1999; Kleinschmidt, Hills, and Elliot 1995), reduced physical activity levels (Yen and Kaplan 1998; Ellaway and Macintyre 1996), and greater alcohol consumption (Diehr et al. 1993; Hart, Ecob, and Smith 1997) have each been shown to be related to aspects of the communities in which people live, over and above the characteristics of the individuals living there.

Although there are now a large number of empirical studies documenting relationships between particular aspects of residential areas and selected health outcomes, critics have noted a relative paucity of attention to conceptualizing and testing the particular pathways underlying the observed associations (Pickett and Pearl 2001; Macintyre, Ellaway, and Cummins 2002; Diez-Roux 1998; Yen and Syme 1999). In the absence of conceptual frameworks, selection of contextual variables for study has been somewhat ad hoc and constrained within the range of those commonly used or readily available through routinely collected data (Mitchell et al. 2000). To some extent contextual effects research on health remains mired in a “poverty paradigm” (Rowley et al. 1993), focusing mostly on the association between census-based indicators of community socioeconomic position and individual health outcomes, with a heavy emphasis on the deleterious effects of concentrated poverty and other forms of disadvantage. The main thrust of such studies has been to show that poorer places are associated with worse health outcomes, above and beyond the characteristics of the individuals who live there (Robert 1999). This emphasis on disadvantage, to the exclusion of other facets of contextual environments, is partly the result of a lack of appropriate data, but it also reflects a paucity of alternative models to inform questions about how community contextual characteristics may plausibly affect specific types of health outcomes.

For instance, studies have found smoking to be significantly associated with neighborhood deprivation level after accounting for individual characteristics (Duncan, Jones, and Moon 1999; Kleinschmidt, Hills, and Elliot 1995). The area-level variables generally included in these analyses, however, consist largely of census-based indicators of economic status such as mean household income, unemployment rate, proportion of the population living in overcrowded housing, proportion without access to a car, and proportion of household heads with lower social class occupations, considered either separately or in an aggregate index. While living in deprived communities is found to be associated with higher smoking levels in the analyses, it may be instructive to examine a wider set of characteristics of poor areas that are theoretically relevant to smoking rates, including the cost and availability of cigarettes, targeted advertising, the quality and quantity of preventive and smoking cessation services available, and smoking policies in schools and other public areas.

In the sections that follow we describe the process of developing a set of community characteristics that can affect health and we provide details of the specific factors within each dimension.

Drawing upon relevant theoretical and empirical literature on community contextual characteristics and their relationship to health status, a preliminary set of contextual dimensions was developed. This framework served as a basis for a structured workshop for invited consultants, which included prominent investigators from the United States and around the world engaged in cutting-edge research on community contextual characteristics and their relation to health, in a wide range of disciplines including epidemiology, sociology, geography, medicine, demography, economics, developmental psychology, education, and toxicology. Others with interests and expertise in the conceptual design and practical implementation of the indices were also invited, including government experts on data sources and geographic information systems, public health practitioners, and experts on community consultation and processes. The names and affiliations of those attending the workshop are available from the authors.

The initial sessions were full group discussions, in which participants addressed fundamental issues including:

• Is it worthwhile to create indices measuring community contextual health characteristics?
• Would such an effort be feasible?
• For which audiences would the indices be useful?
• Is the preliminary set of dimensions reasonable, and what others should be included?

In the course of the discussion general consensus was reached on a core set of 12 contextual dimensions plausibly linked to health status. These dimensions are diagrammed in Figure 1. There was widespread agreement among the participants that the construction of these dimensions and the identification of traditional and nontraditional data sources to measure these dimensions would potentially be of significant benefit to both the research community and to public health practitioners and advocates.

Figure 1 Twelve Dimensions of the Proposed Community Contextual Health Profile

In the next phase of the workshop, participants formed smaller working groups to brainstorm ideas for sets of components within each dimension that would most optimally quantify that dimension of contextual health. Each working group generated detailed lists of these components, along with suggestions for possible data sources and specific variables that might be used to measure the components of each dimension. This content was then presented to the full group for further input and revision. The workshop concluded with a discussion of issues related to measurement design and index construction, with participants sharing their own experiences and recommendations concerning assessment of contextual factors and their relationship to health outcomes.

After the workshop, we evaluated and further refined the lists of components and data indicators for each dimension, taking into consideration both conceptual relevance and availability of appropriate data at the local level. Table 1 presents the resulting set of contextual dimensions, components, and indicator variables. In the next section we discuss the rationale for inclusion of the components within each dimension, based on input from meeting participants as well as reviews of relevant literature.

Table 1 Contextual Dimensions, Components, and Indicators Potentially Related to Health

It is important to note that although all the proposed indicators offer further plausible conceptual differentiation of the various dimensions of context, it remains to be seen whether their effects on health can be empirically distinguished from more traditional indicators such as poverty (Krieger et al. 2002; 2003). Thus we make no claims about their empirical veracity; rather our goal is to take the first step in providing a broader conceptualization of the specific contextual factors that may affect health.

The fact that data for most of these economic indicators are readily available for small areas in census data is undoubtedly an important factor accounting for their widespread use (Mitchell et al. 2000). Our consultants encouraged a broadened perspective to more fully assess the economic status of communities. On one hand, this involved identifying a more diversified set of indicators for commonly studied components, such as considering various types of income (earnings, investments, and transfers) in addition to the overall mean or median income in an area. On the other hand, a number of additional components of economic well-being were also suggested for inclusion. For example, the opportunities for community residents to obtain financial resources would be influenced by characteristics of economic development in an area such as productivity, industrial mix, amount of area business lending, as well as exchanges of goods and services through the informal economy. The availability of financial services including banks and other sources of credit were considered important, as were local costs of living, patterns of redistribution through taxes and transfers, and the fiscal capacity of the area. One other seldom-considered aspect of the economic milieu concerns the degree to which segments of the community are differentially exploited, and thereby constrained in their access to monetary resources. Indicators of exploitation include the ratio of wages to corporate profits, as well as issues related to location of jobs, such as length of commute and commuter taxation.

In addition to the usual employment indicators, we include a number of other measures. Looking in detail at characteristics of the workforce, for example, along with the area business capacity and the geography of job growth, would facilitate assessment of job access, as well as the degree of spatial ‘mismatch’ which may adversely affect the employment opportunities of central city residents (Holzer 1991; Mouw 2000). Racial, gender-based, and antigay discrimination also limit access to employment, as well as having possible stress-related consequences for health (Williams 1999; Krieger and Sidney 1997; Yen et al. 1999). The degree to which occupational safety regulations and policies are in place and enforced is likely to influence the frequency and severity of work-related injuries (McQuiston, Zakocs, and Loomis 1998), while aspects of job quality including wage equity, family-friendly policies and demand/control characteristics of jobs can reduce or exacerbate job-related stress and its sequellae (Cheng et al. 2000; de Jonge et al. 2000; Saltzstein, Ting, and Saltzstein 2001). The presence of labor unions is also associated with more optimal working conditions and employee compensation (Hirsch and Macpherson 2001).

In these studies the contextual educational variable tends to be treated as a marker for a more generalized concept of community socioeconomic status and resources, rather than being considered in its own right. Our consultants suggested that a focused assessment of aspects of education that are likely to vary among communities is warranted. Multiple measures of the population's educational attainment and functioning are included in the recommended indicators. Moreover, the levels of funding, characteristics of school systems and curricula, and learning-related aspects of community life such as prevalence of television viewing and numbers of library books per capita, can provide insights into the priority placed on education and corresponding investment within an area, which itself may be related to health outcome.

Specific aspects of the curriculum also have implications for the health of children at school ages and throughout their lives. For instance, bullying and violence is a serious problem among children and adolescents (Nansel et al. 2001), and the presence of violence prevention programs has been found to be effective in decreasing physically aggressive behavior (Twemlow et al. 2001; Grossman et al. 1997). Similarly, obesity in children is approaching epidemic proportions, and is related to adult obesity and levels of lipids, cholesterol, triglycerides, insulin, blood pressure, and to risk of coronary heart disease (Styne 2001). Incorporation of nutrition modification programs and optimal physical education curricula in schools can be effective in modifying these risks (Stone et al. 1998; Snyder et al. 1999).

More broadly, political participation has been of recent interest as an indicator of embeddedness in the institutions of civil society. As such, it is considered to be a reflection of social capital within a community (Kawachi 1999). Social capital, measured in several different ways, has been associated with positive health outcomes (Subramanian, Kawachi, and Kennedy 2001; Kawachi, Kennedy, and Glass 1999; Kawachi et al. 1997) (see the Psychosocial Dimension section below for further discussion of the social capital concept).

Within the political contextual dimension, we include aspects of political participation such as voting and political party membership, as well as donations to parties and candidates, which are known to influence public policy (Ferguson 1995). Likewise the degree to which elected officials are representative of their areas in terms of gender and race/ethnicity may be an important factor in their responsiveness to constituents' needs (Whitby 1997; Bratton and Haynie 1999). The percent of the local budget devoted to public health investments can be considered an indication of the priority placed on health by the community as well as a measure of available fiscal resources. We also include the number and influence of various politically active community groups.

In addition, this contextual dimension encompasses structural features of communities such as physical design of streets, sidewalks, and safety structures that are associated with level of injury risk (Navin, Zein, and Felipe 2000; Agran et al. 1996). Aspects of land usage are also considered, such as public spaces and parks that may facilitate greater physical activity levels (French, Story, and Jeffery 2001), as are services related to environmental quality like waste disposal and recycling programs.

Home ownership has been associated with reduced morbidity and mortality risk (Filakti and Fox 1995; O'Campo et al. 1997). In most cases this housing variable is regarded as a marker for general material well-being. It has been suggested, however, that long-term exposure to specific health-promoting or health-damaging characteristics of housing itself is likely to account for some of the observed health effects (Macintyre et al. 1998; Ellaway and Macintyre 1998).

There is also some evidence that poor housing conditions during childhood can adversely affect health in later life. For example, Barker and colleagues (1990) found an association between domestic crowding during childhood and later stomach cancer mortality rates, suggesting that overcrowding may promote the transmission of causative organisms among children that exert negative health effects later in life. Similarly, Dedman et al. (2001) noted aspects of poorer childhood housing conditions were associated with increased mortality risk from common diseases in adulthood.

Our consultants suggested that we include characteristics of housing discussed above in our framework, as well as other aspects of residential patterns within communities. Homelessness, for example, has known associations with differentially poorer health (Barrow et al. 1999; Hwang 2001). Segregation by race has been associated with adverse health outcomes among blacks (Williams and Collins 2001; Jackson et al. 2000), as well as among whites in some cases (Collins and Williams 1999). Similarly, concentration of poverty has been found to be associated with elevated mortality risk (Waitzman and Smith 1998a).

We also include two other components within the housing dimension. Regulations such as zoning and industrial/residential segregation can affect the degree to which residential areas are exposed to industrial pollution and other health threats such as increased traffic. Financial issues specific to housing are also considered, such as housing costs, the availability and characteristics of low-income housing, mortgage lending practices, and community reinvestment initiatives.

We also look specifically at aspects of access to health care services, which is related to health status and known to vary among population groups (U.S. Department of Health and Human Services 2000). Some of the factors included, such as insurance coverage and the availability of indigent care, are well-known determinants of access (Baker, Shapiro, and Schur 2000; Newacheck et al. 1998; Nelson et al. 1999). The racial/ethnic makeup of medical staff in relation to the patient population and the cultural competence of providers and institutions may also be important in encouraging utilization of health care resources that are present in an area (Flores et al. 1998; Langer 1999). In addition to traditional measures of access, we also include rates of hospitalization for ambulatory care sensitive conditions. These are conditions considered to be manageable on an outpatient basis given access to high quality primary care, and therefore higher hospitalization rates can be used as an indicator of poorer access to appropriate care (Institute of Medicine 1993).

Another aspect of local governance with potential health relevance concerns the degree to which there is municipal fragmentation. This term refers to a situation in which large numbers of smaller governmental entities exist within a metropolitan area. It has been argued that in cases where there are high levels of municipal fragmentation and no single government empowered to act for the good of the entire region, a host of problems result, including resource and public service imbalances within the area and the protection of privilege (Mitchell-Weaver, Miller, and Deal 2000; Ross and Levine 1996).

More recently, aspects of social interactions and relationships have been increasingly conceptualized as forms of social capital, although there is widespread disagreement about the meaning of the term and the level of aggregation at which it operates (Lynch, Due et al. 2000; Woolcock 2001). Portes (1998, p. 6) defines social capital as “the ability of actors to secure benefits by virtue of social membership in other networks or social structures.”Coleman (1988, p. 598) sees social capital as a resource for organizations as well as individuals: “Social capital is defined by its function. It is not a single entity but a variety of different entities, with two elements in common: they all consist of some aspect of social structures, and they facilitate certain actions of actors—whether persons or corporate actors—within the structure. Like other forms of capital, social capital is productive, making possible the achievement of certain ends that in its absence would not be possible.”Putnam et al. (1993, p. 167) consider social capital broadly as “features of social organization, such as trust, norms, and networks that can improve the efficiency of society by facilitating coordinated actions.”

Social capital has been operationalized in different ways in health-related empirical research. Per capita membership in groups and associations has been used to assess civic engagement (Kawachi et al. 1997; Kawachi, Kennedy, and Glass 1999), as has political participation (Blakely, Kennedy, and Kawachi 2001). Several studies have considered greater mistrust to be indicative of lower levels of social capital (Kawachi et al. 1997; Kawachi, Kennedy, and Glass 1999; Subramanian, Kawachi, and Kennedy 2001). Mistrust is generally defined as the percentage in an area who agree with the second part of the statement: “Generally speaking, would you say that most people can be trusted or that you can't be too careful in dealing with people”. A related indicator is perceived lack of fairness, indexed by percent agreeing that “most people would try to take advantage of you if they got the chance” (Kawachi et al. 1997). Perceived helpfulness/reciprocity has also been used as a gauge of social capital, based on answers to the question “Would you say that most of the time people try to be helpful, or are they mostly looking out for themselves?” (Kawachi et al. 1997). It has also been hypothesized that crime level is an indicator of collective well-being that is influenced by the degree of cohesiveness in social relations or social capital (Sampson, Raudenbush, and Earls 1997; Kawachi, Kennedy, and Wilkinson 1999).

Within the psychosocial dimension we include theorized aspects of social capital such as civic engagement via political participation, membership in voluntary organizations and unions, and charitable giving. Crime as a marker for social cohesion is assessed through expenditures on jails and incarceration rates. Collection of information on lawsuits and the presence and utilization of protective services was also suggested as an indicator of the level of trust in communities.

For each of the behaviors, we examine specific characteristics of communities that might influence the degree to which they will be adopted by residents. In the case of tobacco use, these characteristics include current smoking rates, the presence of cessation and preventive education programs, workplace smoking restrictions, the cost and accessibility of cigarettes, and targeted advertising. In the area of physical activity, we include reported activity levels, physical education requirements in schools, participation in local sports and recreational activities, as well as availability of exercise facilities in the workplace and in the area more generally. We also consider indicators of sedentary activities such as television viewing patterns and video game sales and use. Regarding diet and obesity, we look at consumption patterns of healthy foods such as fruit and vegetables as well as high-fat and high-sugar foods. The quality, availability, and cost of a range of different foods is of interest, as is the availability of generally less-nutritious “fast food” as indexed by the number of such establishments in the area. We also include aspects of nutrition in the schools, including the prevalence of subcontracting to vendors of nonnutritious items and the presence of nutrition education programs. In the area of alcohol and illicit drug use, we consider availability as indicated by number of liquor stores and marketing laws as well as the nature of public advertising. We also include drug and alcohol treatment service availability, and the presence of syringe laws and exchange programs. Violence in the community is indicated by factors such as the availability of guns and the level of exposure to violence perceived by residents.

In our framework we include measures within each of these health-related aspects of transportation. Vehicle occupant and pedestrian safety factors are included, as well as characteristics of the infrastructure of roads, sidewalks, and bike lanes. We examine characteristics and numbers of vehicles, and aspects of the public transportation system. Finally, we include economic issues such as government transportation spending priorities, funding for public transportation, and personal insurance rates.

With this conceptual “road map” as a framework, the next phase was to attempt to find out what measures of these recommended dimensions actually existed. Thus, we proceeded with building a data library with datasets containing information for Metropolitan Statistical Areas (MSAs) relevant to each dimensional component. We chose to focus on MSAs as our unit of analysis in quantifying contextual factors because they are the smallest census-defined geographic unit for which a broad range of data is routinely collected and geocoded. We recognize, however, that there is no single ideal level at which to measure context in terms of its relationship to health-damaging and health-promoting factors. In fact different characteristics may operate at different levels. An argument can be made that using more localized units, such as county, zip code, census tract, and census block, increases the likelihood that certain aspects of the social and physical environment actually experienced by individuals are being measured. Conversely, there are considerably more richly detailed contextual data sources available for larger units such as states and MSAs. Given the inevitable tradeoffs between data availability and proximity to lived experience, we find the MSA to be the most satisfactory alternative for our purposes.

An important aspect of our project is the exploration of data sources not traditionally used by public health researchers but which can provide valuable information about the health enhancing aspects of the local environment. Tables 2–13 provide lists of such data sources related to each of the 12 dimensions. The majority of these sources provide data for MSAs or smaller geographic units such as counties; a few sources at the state level have been included when more localized data did not exist.

Table 2 Selected Data Sources Relevant to the Economic Dimension

Table 3 Selected Data Sources Relevant to the Employment Dimension

Some sources are familiar to many researchers and practitioners, such as the Census Bureau decennial data files and the State and Metropolitan Area Data Book, a resource that is compiled by the Bureau and incorporates a wide variety of data from governmental agencies and the private sector. Others may not be so well known, such as aggregate databases assembled by research, governmental, or private sector organizations with particular interests in MSA- or county-level information. One example is Demographics USA, a commercially available database containing measures of disposable income, income capacity, consumer expenditures, and information on area retail trade establishments and retail sales. Another is the State of the Nation's Cities, which was compiled by the Center for Urban Policy Research in collaboration with the U.S. Department of Housing and Urban Development. This database has measures of area income inequality, gross metropolitan product, and fiscal health. A third such source is the Contextual Data Archive, containing indices of economic segregation. We have also obtained economic indicator data from organizations including the Brookings Institution, the Federal Financial Institutions Examination Council, the American Chamber of Commerce Researchers Association, and the Economic Policy Institute. Further information about accessing these and other data sources listed in Tables 2–13 is available from the authors.

Table 13 Selected Data Sources Relevant to the Transportation Dimension

Having provided details of potential characteristics of communities that may affect health and the process through which they were generated, we will conclude this paper by describing three general principles that we think can help guide the development of future research and advocacy on the importance of community contextual characteristics for health.

A key question underlying much of the literature on contextual characteristics and health is whether local variations in outcomes are attributable to the composition of the population in an area, or whether aspects of the areas themselves influence health status. While authors of several earlier studies concluded that area effects were not significant once individual factors had been taken into account (Duncan, Jones, and Moon 1993; Sloggett and Joshi 1994), reviews of more recent literature incorporating more complex multilevel modeling have concluded that there is fairly consistent evidence for the presence of so-called “area effects” for a range of health outcomes (Pickett and Pearl 2001; Macintyre, Ellaway, and Cummins 2002).

We propose three related guiding principles that we think are useful for understanding how aspects of contextual characteristics can affect health and disparities in health.

Greater understanding of how community contextual characteristics may affect health will require thinking about mechanisms linking contextual characteristics to specific health outcomes. The contextual characteristics associated with falls among the elderly are not necessarily the same characteristics implicated in respiratory disease in children, or in suicide. Likewise, the uneven or broken pavements implicated in falls among the elderly do not cause respiratory disease, even though we may be forced to pragmatically use a count of uneven or broken pavements as a surrogate for the causal exposures for respiratory disease. Indeed, tallies of dilapidated buildings or broken windows have been used as surrogate measures of contextual environment linked to certain health outcomes (Cohen et al. 2000). Nevertheless, it is important to be clear that these may not be the causal exposures relevant to different types of health outcomes.

The point is that unless we attempt to study rather specific health outcomes, it will be difficult to know which contextual characteristics are important for which health outcomes. A study by Merlo and colleagues (in press) is an example of showing how a specific aspect of contextual environment—levels of social participation and the communication networks it fostered—is plausibly related to use of hormone replacement therapy, but is not related to anti-hypertensive medication use, in spite of the fact that these are very similar outcomes.

This specific outcome/mechanism approach is, in our opinion, one way forward and needs to be applied more rigorously across a range of potential health outcomes relevant to contextual influences on health disparities. We do not wish to appear overly reductionist and would hold open the possibility that many contextual characteristics are likely to be empirically implicated in a range of health outcomes (e.g., poverty). Moreover, we should also be aware of the potential for clusters of exposures to interact in producing disease, so that it is the spatial aggregation of a series of negative contextual forces that act synergistically to magnify the total exposure load or increase susceptibility to certain outcomes (Diderichsen, Evans, and Whitehead 2001). However, the general principle of moving toward a search for greater conceptual and mechanistic specificity (both social and biological) of contextual influences on health is important.

Our point here is to suggest that there is no single pathway to health and therefore there is not a unique set of contextual characteristics that will be universally important for all health outcomes or at all stages of the outcomes. The two principles described so far are important in guiding the next phase of research on health-related contextual effects and this recognition is motivated by the overwhelming body of knowledge that suggests that there are indeed different pathways to different types and stages of health outcomes.

This does not mean that there are not important proximal contextual characteristics related to the triggering of events. However, we should conceptualize how measured exposures to contextual characteristics match up with the temporal logic of the stage of the outcome. This means we should start thinking about the potential effects of community contextual characteristics through a lifecourse perspective. It is possible that exposure to current environmental conditions per se has little to do with the current distribution of disparities in the incidence of chronic disease outcomes like rates of heart disease or cancer. Rather, they may more strongly reflect both contextual characteristics and individual influences from earlier in the lifecourse. This recognition means that both individual and contextual exposures over the lifecourse are of interest in better understanding contemporary levels and disparities in population health outcomes such as heart disease, cancer, chronic lung disease, and diabetes. Thus, it is important to remember that exposures in early life can have both direct and indirect long-term effects on health. Biological and social exposures co-evolve over the lifecourse so that contextual exposures at one point may be literally “embodied” to the extent they become an “individual characteristic” later in the lifecourse. (Kuh et al. in press; Aboderin et al. 2002; Davey Smith et al. 2000; Davey Smith, Harding, and Rosato 2000; Harding 2001; Ben-Shlomo and Kuh 2002; Leon 2000).

In this paper, we have documented a project designed to elucidate those aspects of community contextual characteristics that are plausibly linked to health. We hope the information presented here is useful to public health researchers and advocates in terms of broadening the conceptualization of potential contextual influences on health and understanding the health-promoting and health-endangering characteristics within which decisions about the provision of health care services are made. In addition to this very practical contribution, we have also attempted to provide some thoughts on how the research agenda on contextual health effects might move most profitably forward. We believe that it is important to adopt a lifecourse approach to more fully understand the effects of contextual characteristics on health and health disparities.

The authors wish to thank the workshop participants for their invaluable insights and suggestions.