Future Research Directions for Understanding Neighborhood Contributions to Health Disparities

Abstract

This paper proposes several promising future directions for neighborhood research to address health inequalities. First, there is a need to apply a Geography of Opportunity framework to understand how vast spatial (neighborhood, regional) inequality translates into health inequality. Such a framework highlights inequality that unfolds across an entire region, as well as the continuing significance of race/ethnicity for producing disparities in health and in the social determinants of health. The Geography of Opportunity framework also points to some of the methodological limitations of current neighborhood health studies, given the structure of neighborhood racial inequality in the US for estimating how important neighborhoods are for producing racial health disparities. Second, there is a need to incorporate life course concepts, data, and methods, including to model residential histories, neighborhood temporal change and residential mobility, starting early in life. A life course focus would help inform when in life neighborhoods matter most for health and health inequalities, as well as improve exposure assessment of residential contexts. Third, we must model mechanisms linking neighborhoods and health, including the role of individual and household socioeconomic status (SES). Lastly, we need to more meaningfully integrate social determinants of health, including drawing on policy evaluations that aim to improve neighborhood environments or that aim to expand household neighborhood choice. Doing so would inform how specific modifiable neighborhood exposures stimulated by policy may influence health and health disparities.

Introduction

The United States (US) Government sets 10-year health goals and tracks progress for the nation in its Healthy People initiative. Its 2020 goals include improving overall health, achieving health equity and eliminating health disparities, as well as creating social and physical environments that promote good health for everyone(1). Therefore, understanding health inequalities, including the contribution of neighborhood residential context, is a US governmental priority.

Neighborhood effects are defined as outcomes from a causal process of an exposure of living in a particular neighborhood (2, 3). A neighborhood effects literature has developed over the past few decades, of which health forms one part. Certain fields like epidemiology are increasingly focusing on contextual factors like neighborhoods, instead of exclusive focus on individual and family level explanations for understanding the health of populations (4). This locates the sources of illness external to the individual, implicating toxic context (5). Although the body of neighborhood-health literature is strengthening with time, it remains focused on cross sectional designs, on present-day exposures, and on adult populations, in lieu of probing how neighborhoods shape health across the long term or early in life (6–8). The health literature also has only a rudimentary understanding of how neighborhoods shape the health of different racial/ethnic populations, especially in the US, where both neighborhood context and health outcomes are markedly unequal by race. Such racial disparities in the social determinants of health have developed as a result of historical institutional practices and laws, as well as interpersonal and societal racist stereotypes and attitudes, that jointly have disenfranchised racial minorities (9, 10). Although race is widely recognized as a social, not a valid biological, construct (11, 12), it continues to be relevant for socially distributed and patterned goods and privileges (12).

In this paper, I propose several promising future directions for neighborhood research, specifically to address health inequalities, including those by race. First, there is a need to apply a Geography of Opportunity framework to understand how vast spatial (neighborhood, regional) inequality translates into health inequality. Such a framework highlights inequality that unfolds across an entire region, as well as focuses in the continuing significance of race/ethnicity for producing disparities in health and in the social determinants of health. The Geography of Opportunity framework also points to some of the methodological limitations of current neighborhood health studies, given the structure of neighborhood racial inequality in the US for estimating how important neighborhoods are for producing racial health disparities. Second, there is a need to incorporate life course concepts, data, and methods, including to model residential histories, neighborhood temporal change and residential mobility, starting early in life. A life course focus would help inform when in life neighborhoods matter most for health and health inequalities, as well as improve exposure assessment of residential contexts. Third, we must model mechanisms linking neighborhoods to health, including the role of individual and household socioeconomic status (SES). Lastly, we need to more meaningfully integrate social determinants of health, including drawing on neighborhood-related policy evaluations, e.g. those that aim to improve neighborhood environments or that aim to expand household neighborhood choice. Doing so would inform how specific modifiable neighborhood exposures stimulated by policy may influence health and health disparities.

Incorporating a Geography of Opportunity Framework

Although some social determinants of health are increasingly being studied for understanding racial/ethnic health disparities, the role of neighborhood context for originating racial/ethnic health disparities has garnered insufficient attention. In epidemiology, neighborhoods are examined in isolation from residential sorting or residential segregation processes that perpetuate substantially different distributions of neighborhood context for racial minorities and whites (13). Epidemiologic neighborhood research thus omits examination of the larger context within which neighborhood and residential move decisions are made – within regional housing markets – which function differentially along racial/ethnic lines (14) and thereby produce an unequal Geography of Opportunity by race/ethnicity (13, 15–17). The Geography of Opportunity premise is that residents are situated within a context of place-based opportunities that shape quality of life, and there are consistently large racial/ethnic disparities in access to neighborhoods of high opportunity (3, 18). Neighborhood epidemiologic research would therefore benefit from incorporating a Geography of Opportunity perspective to address health disparities.

Because high racial residential segregation is a central feature of inequality in America, it plays a key role in patterning opportunity by neighborhood. Segregation is defined as living in separate neighborhoods across a larger geographic area, for example, a metropolitan region (19). Racial segregation remains high for black households in the US today. For example, according to the dissimilarity measure, the most common segregation measure ranging from 0 (complete integration) to 1 (complete segregation), 59% of black or white Americans would have to move to a different neighborhood for there to be complete racial integration in the year 2010, on average in the largest metropolitan regions (20). In some of the highest-segregated regions (e.g. Detroit MI, New York NY, and Chicago IL), the 2010 black-white separation exceeds 75% (on the dissimilarity index). While racial segregation for Latino and Asian households compared to white households is lower than it is for black households, it remains moderate (20). Racial residential segregation generates large inequalities in neighborhood resources, services, and social contexts which are hypothesized to influence health, including school quality, safety, healthy food access, green space, exposure to violence and crime, sexual activity, and social networks (21, 22). Generally black and Latino households in the US are disproportionately exposed to harmful neighborhood exposures, while white households are disproportionately exposed to salubrious neighborhood exposures. And while household socioeconomic status accounts for some of these racial segregation patterns, other factors play prominent roles as well, including housing discrimination, race-based attitudes, foregone choice due to anticipated harassment, residential choice and constraint, and historical segregation patterns (23–27).

In addition to their higher probability of exposure to single hazardous environments, racial minority families living in highly segregated regions are also much more likely to be exposed to cumulative disadvantage. For example, in a quarter of black and Latino families in highly segregated US regions, children experienced the double jeopardy of living in impoverished families, compounded by simultaneous residence in impoverished neighborhoods. There is a sharp gradient by segregation level for Latino and black families, although this double jeopardy is not as prevalent for white families, since poor white families in the US often live in middle income neighborhoods (13, 28).

Several implications for neighborhood research derive from a Geography of Opportunity framework. It highlights some of the limitations of focusing neighborhood studies within one city or town, as opposed to considering a regional approach. Metropolitan regions proxy housing and labor markets (19), and sampling individuals throughout the region may better model the effects of those market forces, including housing market processes that are differential by race. The framework emphasizes the continuing significance of racial discrimination in US housing and labor markets; and it considers differential access to neighborhood resources by race, social class, nativity and other meaningful social stratification categories (13).

Methodological Challenges to Studying Neighborhood and Health Inequalities

Even though neighborhood context may be influential for producing health disparities, the Geography of Opportunity perspective highlights the methodological difficulties of trying to demonstrate empirically how important neighborhood context is for explaining health disparities, given such high racial residential segregation and racism in the US. Foremost among these issues are threats to causal inference, but internal validity is not the only issue. Study designs and measures, for example, have the potential for bias or limited generalizability (13).

Numerous threats to causal inference (internal validity) exist when observational data is used to estimate the association between neighborhood characteristics and health. Violations of causal inference assumptions may be especially problematic for neighborhood studies focused on racial health disparities given the high degree of racial segregation in the US. For example, the positivity (also known as overlap) assumption assumes there is comparable overlap across all covariates for an exposure of interest (neighborhood environment) (29, 30). Yet by virtue of high segregation, positivity is often violated across racial groups when examining racial differences in large US cities, especially in the most highly segregated cities like Chicago, Detroit, and New York City (13, 31). Positivity is violated because black households may not live in the best tail of the distribution to neighborhood quality, and white households may not live in the worst tail of the neighborhood quality distribution (3, 31). This illustrates a form of structural confounding that is a function of the social ecology of racial inequality of residence in America. Therefore this lack of positivity is not addressable by better sampling design since such comparable racial populations residing in different types of neighborhoods may just not exist (30).

The exchangeability assumption, another primary causal inference assumption, assumes lack of residual confounding between exposure (the neighborhood characteristic) and outcome (29). However confounding by race might be especially difficult for neighborhood studies given that willingness and motivation to live in different types of neighborhoods is differential by race (32, 33). This confounding related to neighborhood attitudes and aspirations operates above and beyond the fact that households of different racial groups may live in completely different neighborhoods(3). Moreover, the neighborhood effects literature for the most part does not integrate racialized housing market processes into the understanding of what causes neighborhood of residence, and therefore, what may confound neighborhood health associations that are race-specific (13). Indeed, residential selection patterns are not well modeled in neighborhood studies, and this has been deemed one of the fundamental threats to causal inference in neighborhood research (8, 29, 34, 35).

A third causal inference assumption, the reverse causation (or simultaneity) assumption, dictates that an exposure occurs prior to a health outcome (36). However if people are forced to move to worse neighborhoods because of an illness, this causes health-related selection into worse neighborhoods. Since housing discrimination cases due to disability status have been recently outnumbering discrimination cases due to race (37), it is feasible that the pattern of disabled individuals living in worse neighborhoods than non-disabled individuals, observed in cross sectional data, could be due to disabled individuals who are denied fair housing in better areas, rather than to neighborhood context causing the disability. Reverse causation of such neighborhood-health associations could be minimized given longitudinal data and incorporation of variables detailing residential moves.

Integrating a life course approach to neighborhood epidemiologic research

Life course epidemiology focuses on how exposures throughout life, especially during biologically or socially vulnerable periods, influence health at later ages (38, 39). Context refers to both place and time (38); however the neighborhood epidemiology and life course epidemiology literatures typically operationalize either place or time. Integrating both is a key challenge for future neighborhood epidemiologic research. Applying a life course perspective to neighborhood research highlights the importance of such phenomena as mechanisms, transitions, trajectories, developmental stage and age, childhood neighborhood, and dynamic assessments of residence into measures of neighborhood exposure (40).

Modeling Dynamic Neighborhoods: Time, Change, and Developmental Stage

Most prior evidence treats neighborhood contexts as a static construct modeled as a point-in-time exposure (41). But there is a strong need to model dynamic measures of neighborhood environment, to better capture neighborhoods as a source of risk or protection that can vary over time. These limitations are particularly important because childhood and adolescence are developmentally-sensitive periods for the establishment of long-term health risk (42). Knowing when exposure to neighborhood context matters most for health may guide the prioritization of policies to target certain populations or resources during such sensitive periods (43). Therefore, incorporating time, age, and developmental stages into neighborhood studies is important for assessing neighborhood effects, to not only understand time-varying neighborhood exposures, but also to contrast the neighborhood exposures as causal effects on health.

Although some studies are starting to leverage longitudinal models, studying neighborhoods as time-varying exposures may be challenging conceptually and empirically. There is some evidence that moves households make across the life course are among similar types of neighborhoods (41). If so, then there may be limited within-person variation to meaningfully model neighborhood transitions with longitudinal data in some populations. In these situations, cross sectional assessments of neighborhood may be good proxies for life course neighborhood exposure.

Other research is demonstrating strong stability of neighborhood environment across generations in highly racially residentially segregated regions, like Chicago. But the intergenerational stability may be strongly patterned by race. For example, while white families move out of Chicago central city to suburban neighborhoods which are safer and of higher SES, black families remain in inner city Chicago concentrated poverty areas (44). Such differential neighborhood advancement or attainment in the US by race has been well described in sociology literature, as an effect of institutional racism whereby racial minority groups experience barriers to translating individual socioeconomic mobility into residential moves to better neighborhoods. Such barriers may be due to housing discrimination by real estate professionals, to foregone choices to move to certain areas by minorities anticipating hostile treatment, to historical conditioning, or to access to different information which informs vastly different choice sets for households wishing to make a residential move. This literature on the racial barriers and differential choice sets to achieving residential mobility which differs by race suggests that the factors influencing moves differ by race (14, 27, 45, 46).

In addition to integrating measurement of neighborhood context and residential history across longer temporal periods such as decades, some research is modeling the daily mobility of populations within their neighborhood. Mobility as measured by Geographic Information Systems (GIS) technology may more accurately measure place-based exposures for understanding physical activity and obesity. Such methodology may better estimate the scale and spatial distance in which people interact with their neighborhoods, instead of relying on administrative definitions of neighborhoods which may not capture a conceptually meaningful unit that residents experience (8).

Early Life

Studying the role of neighborhoods from birth or early life may vastly inform our understanding of how important neighborhood exposure is for health disparities. One important research goal pursues whether neighborhood effects on health are different when exposure acts during different ages or developmental stages. Although the majority of neighborhood studies are among adults, the footprint of inequality may be imprinted from very early in life, including to launch children on different health and developmental trajectories based on their social class (47). Notably, unlike some more traditional exposures examined in epidemiology, exposure to social causes such as neighborhood context may be nonspecific, nonlinear, and recursive, and neighborhood exposures may reflect exposure to mundane stressors experienced over long periods of time, rather than to exceptional exposures (47).

Childhood and adolescence are developmentally-sensitive periods when exposure may imprint its effects for health risk far into the future (42, 48). For example, adolescents are experiencing extensive hormonal changes, social changes, and psychological changes, all of which may interact with social context (38, 49). For younger children, neighborhood effects are buffered and filtered through their parents. However as children age, their exposure to neighborhoods, including social interactions with peers, becomes more important. In sum, the paucity of longitudinal studies from early life is an impediment to examining developmentally-specific neighborhood research questions. Moreover since the family is a core enduring context for understanding life course effects, integrating multilevel studies of children and their families within neighborhoods will be necessary to understand the multiple and contingent effects of neighborhoods early in life.

Place based opportunity and neighborhoods as causes of socioeconomic position

One element of life course epidemiology highlights mechanisms linking exposure at each period in the life course to subsequent health outcomes. Therefore, mediation analyses are one essential need for neighborhood research (8), including identifying mediators that are modifiable by policy or interventions. Some pathways for some outcomes (for example, the role of the built environment for walking and/or physical activity (50, 51)) have been better studied than others. However other pathways linking neighborhood with health remain understudied, including the role of individual and household level socioeconomic status attainment.

The contribution of neighborhood environment to racial differences in socioeconomic attainment is often omitted in both the epidemiology literatures on socioeconomic health disparities and on racial health disparities. Yet social class differences by race are partially determined by neighborhood and metropolitan opportunity structures(52). The limitations of modeling a neighborhood as a more distal cause of health may be challenging for documenting these associations empirically with traditional epidemiologic methods (including more measurement error, which contributes to smaller effect sizes)(53). Yet the social class and health literature should recognize the contributions of place towards the generation of social class that have been established in the demography literature and elsewhere (52, 54).

Racial segregation is one key factor producing differential educational attainment and employment outcomes by race (52) and thus socioeconomic status (SES) is on the causal pathway between residential segregation and health (9). Household wealth and housing tenure (renting vs. owning), are two further dimensions of SES closely tied to neighborhoods and housing markets, and patterned by race, especially since homeownership is the primary source of American household wealth (55). Yet wealth and housing tenure are much less likely to be modeled as SES indicators in health studies compared to such indicators as education and income.

SES may be especially important as a mediator of segregation and health early in the life course, and may be less important later in life. For example, education is a common indicator of SES that is established early in life. Some strong causes of educational attainment and quality are spatially and racially patterned in America. Because schools are funded locally and public school assignment is for the most part based on local residence, racial residential segregation is very highly correlated with racial school segregation, particularly at the elementary school level (56, 57). School segregation is one linchpin of American racial inequality since education is so central for social advancement (57). Some literature has moreover documented that school desegregation policies may influence health outcomes (58, 59).

Additionally, racially separate distributions of neighborhood poverty result in disproportionate exposure of minorities to poor neighborhood environments which may compound the health effects of low socioeconomic status at the individual level. For example, while only about 1% of poor white children live in poor neighborhoods, 17% of poor black children and 21% of poor Hispanic children do (28). These facts combine with results of other literature to suggest that place may matter more for minorities than for whites. This is based on evidence that variance in health outcomes is larger for minorities than whites (60, 61) and on other evidence that distributions of neighborhood context are much broader for minorities than whites in the US. For blacks and Hispanics, due to residential and school segregation, some metropolitan areas have vastly better opportunity spaces than others, whereas anywhere whites live seems to be relatively good (small variance around a good mean) (3, 62).

Yet, most studies of health disparity focus on causes operating at the individual level, and as such, they adjust for socioeconomic status at the individual level but fail to account for neighborhood exposures. Given the substantial separation of neighborhood quality distributions by race (3), omitting neighborhood information is a serious limitation, and may explain why many studies of health disparities have found that “race” has residual statistical associations with health after adjusting for SES at the individual level. Additionally, since minorities (and poor minorities) are much more likely to live in high poverty neighborhoods than whites (and poor whites)(3, 28), then failing to adjust for neighborhood environment may have differential effects by race.

Relevant levers to modify neighborhood exposures

Although social determinants of health are being consistently documented as associated with and causing a range of health conditions (63), understanding how to change health via the social determinants of health is more challenging. This may be partially due to the mismatch between an exposure measured in an observational study, compared an exposure that is manipulated in an intervention study. Interventions may manipulate much more discrete changes in social factors, compared to observational social factors which may remain constant over long periods of time, or which may change as a result of any number of endogenous factors. Moreover, causal inference presumes not only the ability to modify an exposure, but also modifying it within the relevant etiologic window for disease. For example, although improving social support may be important to prevent chronic disease, the effects of social support may accumulate across the life course such that intervention late in life is less effective for improving late-life health (64).

Although social and economic policies are not typically considered part of US health services infrastructure, increasingly research is documenting that such policies may affect health, even if they are not originally intended to (28, 65, 66), via the social determinants of health. Generally, two different policy or intervention approaches are taken to improve neighborhood context. One is place-based neighborhood revitalization policy, which attempts to improve different dimensions of neighborhood context, generally targeting underclass neighborhoods. A second approach is housing mobility, which in America facilitated by housing mobility policy, whereby the choice of housing and neighborhoods is expanded for low income populations by virtue of receiving a government sponsored voucher which subsidizes rent in the private rental market. Both approaches are recommended by housing policy experts to address neighborhood inequality (67), with each approach offering strengths and challenges (13).

Some housing mobility programs, like the Housing Choice Voucher (formerly known as Section 8) housing policy in the US, have been evaluated with experimental designs. Therefore, an exogenous move, stimulated by a randomly-assigned offer to move to a subsidized rental unit with a rental voucher, can be studied with strong internal validity among the policy-relevant target population (which in the US, is the low-income population). The experimental design hence offers the opportunity to observe a clear counterfactual of the effects of voluntarily moving to certain types of housing or neighborhood (66).

Unfortunately, few housing policy experiments targeting low-income populations have measured health outcomes. But among those that have, a recent review of the literature among low-income populations in the US documented a pattern of improved mental health among those in the active treatment group, whose housing related social determinants improved positively with a move stimulated by a housing voucher, compared to a control group (66).

As one example of a housing mobility study, in the Moving to Opportunity (MTO) experiment, poor families living in public housing at baseline were voluntarily randomized to receive a housing voucher subsidy to subsidize rent in the private rental market. These rental units were also located in much lower poverty neighborhoods than the baseline neighborhood from which the families originated (68). While girls and their mothers experienced beneficial mental health effects of the MTO policy, boys experienced adverse mental health effects as a result of moving to lower poverty neighborhoods through MTO – striking qualitative effect modification by gender (69, 70). Moreover, there was also effect modification of the MTO treatment on mental health by baseline vulnerability, such that youth in families with baseline health problems, or who had experienced violent crime victimization, did not experience mental health benefits of the policy, and the mental health harm of the policy for boys was concentrated in the vulnerable families (69, 70). The heterogeneity of these effects of housing vouchers on health highlights the importance of the contingency of causal processes by health and the social determinants of health, including gender and baseline health status. Examining effect modification by theoretically relevant social determinants of health is warranted for neighborhood future studies to understand these causal processes, to identify who is benefiting (and who is not) from current housing policies, and to identify solutions to expand the benefits to all subgroups (69, 70).

These social experiments represent multimillion dollar investments (68), and housing mobility policy constitutes a large ongoing portion of the US federal housing budget (71). Even though housing vouchers are only one vehicle to manipulate neighborhood of residence, via individual household moves as opposed to place-based neighborhood revitalization, the experimental design of these exogenous moves may be a powerful way to understand how and why changes in neighborhood context emerge. Such manipulation is therefore one policy-relevant way to understand how changes in neighborhood related social determinants of health may cause health and illness, and inform future interventions that aim to do so.

Conclusion

I have argued here that the neighborhood epidemiologic literature would benefit from incorporating a Geography of Opportunity framework, to better understand the structure of social and locational inequality (namely: high racial residential segregation and racism in America) that bears on health inequality. A Geography of Opportunity approach also informs how the methods we typically apply in epidemiology for understanding the main effects of neighborhoods on health may be more challenging for estimating neighborhood health inequalities, given the spatial racial structure of inequality in America. Lessons from life course epidemiology suggest that the neighborhoods literature must advance beyond static measures of neighborhood context to operationalize exposures and model health effects across time and across the life span. The literature would benefit from incorporating neighborhood exposures in childhood particularly, from incorporating family processes as moderators of context, from examining the mediating role of SES for neighborhood –health associations. Lastly, the literature needs to meaningfully integrate social determinants of health into neighborhood –health disparities research, which may include social policy experiments as one way to manipulate and understand housing and neighborhood effects on health with strong internal validity.

In the United States, racial inequalities represent some of the most persistent, prevalent patterns in both population health and housing/neighborhood fields. Addressing health inequalities will therefore require moving beyond proximal causes of disease, to understand the upstream role of neighborhoods and housing markets, especially early in life when exposures to neighborhood context may set populations on unequal health and social trajectories that may be difficult to reverse.