Journal of Urban Health: Bulletin of the New York Academy of Medicine Vol. 80, No. 4, December 2003
2003 The New York Academy of Medicine
Residential Environments and Cardiovascular Risk
Ana V. Diez Roux
ABSTRACT The article begins with a discussion of the rationale for studying the relationship
between residential environments and cardiovascular health. Existing empirical
research relating residential environments to cardiovascular outcomes and risk factors
is summarized. The research areas discussed include neighborhood socioeconomic
characteristics and cardiovascular disease, the effects of residential environments on
physical activity, and the effects of residential environments on diet. Other mechanisms
through which residential environments may affect cardiovascular health are also
briefly noted. Key challenges in investigating the relationship between residential environments
and health are discussed. These challenges include characterizing environments
(including definition and geographic scale as well as conceptualization and measurement
of relevant features), the limitations of observational studies, and the need
to evaluate the health impact of interventions or “naturally” occurring changes in local
environments. The need for interdisciplinary work is emphasized.
Cardiovascular disease epidemiology is often considered to be the paradigm of risk
factor epidemiology. Indeed, many features of modern epidemiology, including
aspects of study design and data analysis and the use of large observational cohort
studies to identify multiple disease risk factors, have been developed or elaborated
in the study of cardiovascular disease. A key feature of this paradigm has been the
focus on individual-level risk factors, both behavioral and biological, often divorced
from the environments or social contexts that facilitated the emergence and persistence
of these risk factors. A correlate of this approach has been that cardiovascular
risk factors are often viewed as matters of individual choice and medical care. Consequently,
their prevention has, by and large, centered on two individual-based
strategies: (a) health education and motivation of individuals to change their habits,
and (b) the early detection of risk factors and their treatment through the medical
care system (as exemplified by campaigns for the early detection and treatment of
high blood pressure and high cholesterol).
In parallel with these individual-centered approaches, there is a long history of
attempts to view cardiovascular risk within environmental contexts. Perhaps one
of the most insightful examples of this approach has been the work of Geoffrey
Rose.1,2 Rose highlighted the need to examine factors related to interpopulation
differences in the distribution of risk factors, rather than factors related to interindividual
differences in risk factors within a population. Rose illustrated how population-
based strategies that shift the entire distribution of risk factors (e.g., that shift
The author is with the Department of Epidemiology and Center for Social Epidemiology and Population Health, University of Michigan. This work was supported in part by MD00206P60, Columbia Center for the Health of Urban Minorities.
Correspondence: Ana V. Diez Roux, Department of Epidemiology, University of Michigan, 1214 S. University, 2nd floor, Ann Arbor, MI 48104. (adiezrou@umich.edu)
569
570 DIEZ ROUX
the distribution of blood pressure in the entire United States to a lower mean blood
pressure level) are likely to have a much greater public health impact than strategies
that target individuals at the high end of the distribution within a population (the
usual individual-based approach). Implicit in Rose’s idea is that distributions are
likely to be influenced by environmental factors affecting the population as a whole.
Others have followed Rose in highlighting how environmental factors may shape
the distribution of cardiovascular risk. Syme3 has argued that the effectiveness of
individually oriented risk-factor modification approaches may be substantially reduced
in settings where environmental factors constrain the initiation or permanence
of behavior change. Although intervention trials have demonstrated that it is indeed
possible to modify individual behaviors with intensive one-on-one approaches,4–6
the overall success of these programs in achieving sustained behavior change has
been modest, particularly in view of the fact that many of these interventions were
quite resource intensive and focused on high-risk and highly motivated individuals.
Syme has argued that, contrary to an assumption implicit in much of the behavior
change literature, changing individual behaviors in isolation may be more difficult
than modifying the environment that facilitates and promotes them. The high prevalence
of these behaviors in the community also makes one-on-one approaches
highly inefficient. Indeed, the emphasis on individual-level education rather than
environmental change has been put forward as one possible explanation for the
lack of strong intervention effects in several US community trials to prevent cardiovascular
disease. The fact that in some cases differences in behavioral changes between
intervention and control communities were not observed despite differences
in knowledge suggests that broader environmental changes are required to support
and sustain behavior change7.
Empirical studies of geographic differences and temporal change also suggest
that environmental factors are likely to be important in shaping the distribution of
cardiovascular risk. Epidemiologic studies have documented important geographic
(county-level) differences in coronary heart disease mortality and in the decline of
coronary heart disease mortality over time.8–10 Researchers have argued for the need
to examine how features of communities shape the context that gives rise to the
presence of cardiovascular risk factors. The rapid increase in body mass index over
recent years11 also suggests that broad environmental factors affecting the population
as a whole (rather than genetic factors) are important determinants of cardiovascular
risk. Some have argued that the obesity epidemic may be linked to concomitant
changes in mass-produced foods and portion size12 as well as to changes
in physical activity associated with urban sprawl.13
It is important to note that emphasis on the environmental causes of disease
does not negate the importance of behaviors. As suggested by Wing,9 community
socioenvironmental characteristics should be included in the model for the causes
of cardiovascular disease, not as competing with risk factor explanations but rather
as providing the context that shapes the distribution of these risk factors. In the
environmental approach to cardiovascular disease prevention, lifestyle characteristics
and psychosocial factors are viewed not as exclusively individual characteristics
but as the patterned response of social groups to the external environment, the
material and social environment in which people live and work14. However, until
recently, empirical examinations of the specific ways in which the environment
influences cardiovascular risk have been rare.
Several trends have converged in recent years to stimulate interest in how residential
environments may influence cardiovascular risk. The first of these has been
RESIDENTIAL ENVIRONMENTS AND CARDIOVASCULAR RISK 571
increasing concern, in epidemiology, about how neighborhoods, or features of
neighborhood environments, may affect health. This interest in residential environments
has occurred within the context of revitalized emphasis within public health
on the socioecological determinants of health15, 16 or the multilevel determinants of
health more generally. A second factor has been the growing visibility of obesity as
a public health problem in industrialized nations (especially the United States); this
has stimulated research into the determinants of obesity, including the role of environmental
factors in shaping patterns of physical activity in children and adults. A
third trend has been increasing interest in interdisciplinary work in understanding
population health, motivated in part by the frustration experienced by researchers
and practitioners working within a single discipline, which has resulted in increased
collaboration and exchange between public health researchers and those in fields
such as sociology, geography, and urban planning. These converging trends have
resulted in a small but vibrant and growing literature relating neighborhood characteristics
and environmental factors generally to cardiovascular risk factors and cardiovascular
risk.
This article summarizes findings of this research, highlights its strengths and
limitations, and suggests avenues for future development. The first section summarizes
three areas of empirical research: work relating neighborhood socioeconomic
characteristics to cardiovascular risk, research on residential environments and
physical activity, and research on residential environments and diet. Work on other
mechanisms through which residential environments may influence cardiovascular
risk is also briefly noted. Although the focus is on neighborhoods or residential
environments, other environmental determinants (which do not necessarily correspond
to neighborhoods) is also mentioned. The emphasis is on the main findings
and gaps in knowledge rather than on details of specific studies. The second section
of the article outlines key conceptual and methodological challenges in the field,
and suggests areas where additional work is needed.
EMPIRICAL EVIDENCE
Neighborhood Characteristics and Cardiovascular Risk
Over the past few years, several studies have documented associations between
neighborhood socioeconomic characteristics and cardiovascular risk. Living in
socioeconomically disadvantaged neighborhoods (as assessed by area socioeconomic
characteristics derived from national censuses) has been linked to higher
prevalence of coronary heart disease,16–18 higher cardiovascular disease mortality,
17,19 and greater incidence of coronary heart disease.20 These associations persist
after controlling for personal measures of income, education, and occupation, suggesting
that the socioeconomic characteristics of residents do not entirely explain
them. Neighborhood socioeconomic characteristics have also been linked to the
prevalence of cardiovascular risk factors, including smoking, dietary patterns,
blood pressure, blood lipids, and body mass index.18,21–24
Although suggestive, this literature is far from conclusive regarding the presence
or strength of neighborhood effects on cardiovascular risk. A major limitation
has been the difficulty inherent in accounting for multiple confounders in an observational
setting. Thus, the possibility that associations reflect unmeasured individual-
level socioeconomic characteristics (rather than an effect of the neighborhood
per se) remains. A related issue pertains to the use of aggregate neighborhood socio572
DIEZ ROUX
economic indicators as the sole measure of neighborhood attributes. These measures
have been used because they are easily available from census data for the areas
to which individuals with measured risk factor information can be linked. The
rationale for their use has been that they serve as proxies for a variety of more
specific features of neighborhoods relevant to cardiovascular risk that have not
been directly measured. The inability to measure these specific attributes of neighborhoods
and empirically test their effects remains a major limitation of this work.
In addition, because the neighborhood characteristics investigated are measures of
aggregate socioeconomic status (SES), their use also raises questions regarding
whether any residual effects observed result simply from uncontrolled differences
in the SES of individuals.
Despite these limitations, the reported associations between neighborhood socioeconomic
characteristics and cardiovascular risk are compatible with a broad
range of causal processes. Possible pathways linking residential environments to
cardiovascular risk are shown in the Figure. Both physical and social features of
neighborhoods may be relevant. Accessibility of recreational facilities (including
both public and private resources), the presence of sidewalks and bike lanes, transportation,
and the design of public spaces may influence participation in sports
and other leisure-time physical activity. Availability of sidewalks and bike lanes,
design of public spaces, features of urban form (such as building design and setback),
land use mix, density of population and activities, patterns of street connectivity,
and the aesthetic quality of the areas may affect the extent to which individuals
walk as part of their daily lives. Neighborhood aesthetic quality (e.g., the
presence of green spaces, interesting features, and pleasant surroundings) may also
be related to the experience of stress or the ability to recover after exposure to
stressors. Features of the local environment such as availability and cost of healthy
FIGURE Schematic representation of possible pathways linking residential environments to cardiovascular
risk.
RESIDENTIAL ENVIRONMENTS AND CARDIOVASCULAR RISK 573
foods and tobacco products as well as food and tobacco advertising may affect
dietary patterns and smoking habits. Noise levels could be related to cardiovascular
risk through the effects of noise on sleep disturbance and possibly stress. Air pollution
may relate to a variety of biological factors that influence either the development
or clinical manifestation of cardiovascular disease.
Features of the social environment may also be relevant. Safety and violence as
well as levels of social support and cohesion may be related to the experience of
stress and the development of psychosocial factors for cardiovascular disease. Social
norms regarding acceptable behaviors may arise or be reinforced in the context of
neighborhoods. There are also multiple interrelationships between the domains discussed
above. The physical and social environments are likely to influence each
other. For example, features of street connectivity and urban design may enhance
or detract from social interactions among neighbors and may influence social cohesion
as well as safety and violence. Physical features of neighborhoods may contribute
to the development of social norms regarding appropriate behaviors. For example,
the presence of sidewalks may influence the likelihood of walking and jogging
in the area, and seeing others walk or jog may influence the likelihood that a given
person will walk or jog. Reciprocal pathways are also likely to be present. For example,
greater prevalence of bicycling to work may create advocacy for changes in
physical environments more conducive to cycling (such as bike lanes or trafficcalming
techniques).
Elucidating whether the differences across neighborhoods in cardiovascular disease
outcomes observed in epidemiologic studies do in fact reflect causal processes
will require the testing of hypotheses linking specific features of residential environments
to specific risk factors and investigating the pathways and processes mediating
these relationships. The investigation of these pathways is still in its infancy.
Environmental impacts on two specific cardiovascular risk factors, physical activity
and diet, have received special attention in the past few years. Existing research on
each of these pathways is summarized below. Evidence for other plausible mechanisms
linking residential environments to cardiovascular risk, including the effects
of residential environments on the development of psychosocial risk factors for
cardiovascular disease as well as the possible contributions of noise and air pollution,
are also briefly discussed.
Residential Environments and Physical Activity
The investigation of the causes of the epidemic of obesity experienced by many
industrialized nations over the past few years has reinvigorated research into the
determinants of physical activity. Part of this trend has been increasing interest in
identifying environmental and policy determinants of physically active lifestyles in
both adults and children.25, 26 The recognition that short daily episodes of moderate
physical activity may be sufficient to produce health benefits has also stimulated
research into the factors that promote being physically active in daily life.27 This
work has developed in parallel with (and independently from) research on neighborhoods
and cardiovascular risk, although both are obviously closely related.
Humpel et al.28 have conducted a systematic review of public health research
on the environmental determinants of physical activity in adults. They identified 19
quantitative studies that assessed the relationship between perceived and objectively
determined physical environment attributes and physical activity behavior. Humpel
et al. classified the environmental variables investigated into five domains: accessibility
of facilities, opportunities for activity, aesthetics, safety, and weather. Accessi574
DIEZ ROUX
bility refers to the presence of recreational facilities and distances to them. Opportunities
for activity includes awareness of facilities and satisfaction with them, the
availability of sidewalks and equipment, and residents’ perceptions regarding opportunities
for physical activity in their area. Aesthetics includes the extent to which
the local area is perceived as pleasant, attractive, and friendly. Safety includes crime,
perceived neighborhood safety, street lights, and traffic. Weather refers to the presence
of poor weather. The vast majority of studies reviewed were cross-sectional
(18 out of 19) and most (15 out of 19) focused exclusively on perceived features of
the environment as opposed to objective measures. Some studies asked participants
to report on perceived barriers to starting or increasing physical activity. Others
asked participants to report on the existence and location of physical activity facilities
in the environment.28 Only two of the studies reviewed by Humpel et al. used
geographic information systems (GIS) approaches to construct objective measures
of relevant environmental features.29,30
Humpel et al. conclude that the most consistent evidence regarding effects
of environmental factors on physical activity in adults is observed for accessibility
of facilities, opportunities for activity, and aesthetic qualities of the area. A few representative
examples are summarized below. In a population phone survey of nearly
3,400 adults in Australia, Ball et al.31 found that the men and women reporting
a more convenient environment (including proximity of a park or beach, a cycle
path, or shops) or a more aesthetically pleasing environment (a friendly, attractive,
or pleasant neighborhood) were more likely to report walking for exercise or recreation.
Booth et al. also found that access to local facilities was positively associated
with physical activity participation in a sample of 450 Australian elderly adults.32
The presence of enjoyable scenery and the frequency of seeing others exercise were
positively associated with being physically active in a population-based multiethnic
sample of 2,900 middle-aged US women.33 Sternfeld et al.34 found that self-reported
lack of facilities was associated with decreased sports and exercise in a sample of
5,000 US women enrolled in a large health maintenance organization. Brownson
et al.35 conducted a national phone survey of nearly 2,000 US adults and found that
the odds of meeting physical activity recommendations were significantly higher in
persons who reported access to places to exercise, walking or jogging trails, or a
park. The presence of sidewalks and enjoyable scenery in the neighborhood was
also associated with increased odds of meeting physical activity recommendations.
These associations persisted after adjustment for socioeconomic indicators. Awareness
of facilities, satisfaction with facilities, and the perception that the area offers
opportunities to be physically active were also found to be associated with greater
physical activity in other studies.36–38
Studies using objective measures have also documented associations between
accessibility of facilities or aesthetic quality of the area and physical activity. For
example, Sallis et al. found that local density of pay-for-use facilities was positively
associated with frequency of exercise.39 In a probability sample of 1,800 Australian
adults, a composite measure of spatial access to built and natural recreational facilities
(assessed using GIS methods) and observer ratings of the functional environment
(proximity of footpaths and stores) and of the appeal of the environment (low
traffic and presence of trees) was positively associated with the likelihood of being
physically active.29 GIS-based measures of distance to bikeways were negatively associated
with bikeway use in a cross-sectional study of 400 adults in Massachusetts.30
In contrast to the findings for accessibility, opportunity, and aesthetic quality,
findings for weather and safety have been mixed. The few studies that have examRESIDENTIAL
ENVIRONMENTS AND CARDIOVASCULAR RISK 575
ined weather as a predictor have found no associations with physical activity.33, 40
Among approximately 13,000 persons in five US states who responded to a special
module included in the Behavioral Risk Factor Surveillance System, those who reported
that their neighborhood was unsafe were more likely to be physically inactive.
41 Safety of footpaths was found to be positively associated with physical activity
in elderly Australians,32 but other studies based on US samples have failed to
find a relationship between safety or crime and physical activity.33, 42
Despite the findings highlighted above, evidence for the impact of residential
environments on physical activity remains limited. Associations are often weak and
results are not always consistent.28 Some studies, for example, have found either no
positive associations or weak associations between neighborhood features and
physical activity,34,40,43,44 although this may be due to limitations in the measures
used. It is also impossible to assess consistency in the strength of associations observed
due to differences in the environmental features measured (as well as how
they were measured) and in the variables for which results are adjusted.
A major limitation of existing work pertains to the conceptualization and measurement
of the environmental features of interest. Most evidence to date comes
from isolated questionnaire items as opposed to validated scales. A few studies have
combined often disparate items into scores that are hard to interpret. Objective (as
opposed to self-reported) measures of environmental features are extremely rare. A
limitation of self-reported measures is the possibility that residents who report low
levels of physical activity are more likely to also report poor recreational resources
in their local area, resulting in spurious associations. On the other hand, poor measurement
of the relevant environmental features is likely to result in important
underestimates of associations. Promising trends in the development of objective
measures include the use of GIS-based measures of resource accessibility29 and systematic
ratings of areas by trained observers on prespecified criteria.45, 46 More fundamentally,
additional work is needed on the conceptualization of the features of
residential environments that are likely to be relevant to different types of physical
activity (for example, walking, sports, or other leisure time activities).
The development of measurement instruments for theoretically relevant
environmental constructs that may be linked to specific types of physical activity
remains a key challenge in the field. Thus, future work will need to combine improved
environmental assessment with greater specificity regarding the aspect of
physical activity studied. One example of this type of approach used multiple observers
to rate neighborhoods on attributes related to their “walkability” and related
the scores obtained to the likelihood that residents would walk to work.45 The
measurement was enhanced through the use of a scale and multiple raters, whose
ratings were used to estimate area attributes using recently developed ecometric
techniques.47 This dimension was then investigated in relation to a specific type of
physical activity.
Another important limitation of existing work pertains to its cross-sectional
nature. This makes it vulnerable to the critique that the associations observed result
from the fact that persons who are more likely to be physically active may choose
to live in areas that are more conducive to physical activity. Longitudinal studies,
especially studies that relate change in environmental conditions to changes in behavior,
are important to strengthen the evidence of a relationship between the environment
and physical activity.
In parallel with the public health research summarized above, a growing body
of work in urban planning and transportation has investigated how features of the
576 DIEZ ROUX
built environment influence the travel behaviors of individuals, including walking
and cycling.48 Although the determinants of transportation choices and transportation
behavior (especially automobile use) have long been of interest to transportation
planners, interest in physical activity itself as an outcome is relatively new.
The built environment, as defined in this literature, comprises urban design, land
development and land use, and the transportation system.48 Urban design refers to
the design of the city and the physical elements within it, including their arrangement
and appearance, such as street design and building design, orientation, and
setback. Land development and land use refer to the distribution of activities across
space, including the location and density of residential, commercial, office, industrial,
and other activities. The transportation system includes the physical infrastructure
of roads (including street connectivity), sidewalks, bike lanes, railroads,
and characteristics of public transport systems. Extensive reviews of how land use
and transportation patterns may influence travel choices and physical activity are
available (see, e.g., Frank and Engelke,49 Ewing and Cervero,50 and Handy51) and
only selected aspects will be highlighted here.
A large number of studies have focused on relating land development patterns
(such as density and land use mix) and transportation systems (such as transit availability
and street characteristics) to transportation choices, including walking.52–56
In a cross-sectional study with Puget Sound census tracts as the units of analysis,
Frank and Pivo54 showed that population density, employment density, and mixed
land use are positively related to transit use and walking for shopping and workrelated
trips. Using data from a household survey of five selected neighborhoods
in the San Francisco Bay area, Kitamura et al.53 showed that residence in a highdensity
area was positively related to the fraction of nonmotorized trips, and distances
to the nearest park and bus stop were negatively related to the fraction of
nonmotorized trips. These associations persisted after controlling for household
socioeconomic characteristics. Other researchers have also found less travel by car
and more nonmotorized travel (such as walking and cycling) in neighborhoods
characterized by a mix of land uses, high density, and pedestrian-friendly designs.55,57
Neighborhoods characterized by proximity and connectivity of trip origins and destinations
also appear to enhance nonmotorized travel such as walking.58,59
Other aspects of the built environment such as building design, building orientation
toward the street, and details of the distribution of homes, stores, and workplaces
have also been investigated in relation to walking and travel behavior.60–63
For example, Cervero and Kockelman63 found that pedestrian-oriented design (including
features of sidewalks, street lighting, and planted strips) encouraged nonauto
travel. Cervero and Kockelman63 and Handy61 found that residents of “traditional”
neighborhoods walked to the store much more frequently than did those in
“late modern” neighborhoods. Age of home (as a proxy measure of features of
urban form related to pedestrian friendliness) has been found to be positively associated
with walking in a national sample of US adults.64
Overall, although results from the urban planning and transportation literature
suggest that features of the built environment are related to travel and transportation
choices, important questions remain.50,51,59,65 As in the case of public health
research on environmental determinants of physical activity, most of the work has
been cross-sectional.66 Until recently, the bulk of the work has focused on decisions
regarding automobile travel rather than walking or cycling. Travel choices are influenced
by many different variables,53,57,66 and the complexities of the full range of
factors affecting travel choice and behavior are often not fully addressed in analyRESIDENTIAL
ENVIRONMENTS AND CARDIOVASCULAR RISK 577
ses.50,51,65–67 The relative importance of features of the built environment compared
to individual attitudes and other individual-level characteristics is still a matter of
research and debate. Features such as urban design, transportation characteristics,
and land use are often interrelated, making it difficult to disentangle their effects.
Different studies have arrived at different conclusions regarding which variables are
most important.49
From the public health perspective, an additional limitation of the urban planning
and transportation literature is that it has focused on transportation choices
(of which walking and cycling are two examples) rather than on physical activity
itself.59 Although travel diaries, activity diaries, and time use surveys48 have allowed
detailed quantification of travel behaviors, they often do not provide adequate estimates
of physical activity (which may take place as part of travel, e.g., with the
purpose of arriving at a specific destination, or not). Handy et al.48 noted how the
models used to predict transportation behavior (both number of trips and choice
of transportation mode) have been largely based on the desire to study predictors
of automobile use and may not be entirely adequate when studying predictors of
walking or cycling. Moreover, in most of the urban planning literature, walking
and cycling are studied as modes of travel, not as walking or cycling per se. Predictors
of walking and cycling for travel may be different from predictors of walking
or cycling during leisure or for physical activity. In addition, walking and cycling
are only two of the many forms of physical activity. Other forms of physical activity
such as sports or other leisure time activities may also be influenced by features of
the built environment. Despite the presence of often sophisticated measures of
travel behavior, measures of physical activity in these studies are often limited or
absent.
An additional difficulty in research on the effects of the built environment on
physical activity pertains to the measurement of relevant features. The measurement
of aspects of the built environment (such as land use mix or pedestrian friendliness)
that are especially relevant to walking and cycling (as well as other forms of physical
activity) is complex and often costly. Few operational measures that can be used
in health studies exist. The use of existing data sources and GIS methods to develop
measures of specific features of the built environment for both neighborhoods and
regions (such as density, land use mix, street connectivity, street scale, aesthetic
qualities) is a promising trend in the field.48,59 Greater specificity in the aspect of physical
activity examined (e.g., differentiating leisure time from non–leisure time physical
activity) is also important, since specific aspects of residential environments (or
of the built environment generally) may be related to some types of activity but not
others. Different types of physical activity may show different trends over time68
and different patterns by neighborhood, and may have made differential contributions
to recent increases in obesity rates.
Residential Environments and Diet
There has been growing interest in the ways in which features of the local food
environment may be related to the dietary habits of individuals. There is some
evidence that the dietary patterns of individuals differ across neighborhoods and
that these differences may not be wholly attributable to individual-level socioeconomic
characteristics.21,23
Neighborhoods may differ in the cost and availability of different types of
foods. Small-scale (and often exploratory) studies in the United Kingdom and the
United States, usually focusing on the comparison of small numbers of contrasting
578 DIEZ ROUX
neighborhoods, have suggested that healthy foods are less available in deprived communities
compared to more affluent ones.69–71 In the United States, reports from
consumer groups and government agencies have also documented differences in
the availability and cost of foods across neighborhoods.70,72–74 Recent research has
suggested that in the United States, supermarkets are more likely to be located in
wealthier and predominantly white areas,75 and fruit and vegetable intake was found
to be positively associated with the presence of a supermarket in the area, after
controlling for personal socioeconomic factors.76 Other researchers have also found
that the availability of healthful (low-fat and high-fiber) products (as assessed by
shelf space occupied in community stores) is related to the consumption of healthful
foods by individuals living in the area.77
Despite the findings noted above, research examining variability in food availability
and cost across neighborhoods remains rare, possibly due to difficulties in
measuring these constructs in a systematic fashion across multiple areas. Findings
may also be highly context specific (differing by city, country, and time period).
For example, some studies have found that healthy foods cost more in poorer
areas,71 whereas others have not.69 In the United States, Morland et al.75 found more
supermarkets in wealthier areas, but in the United Kingdom, Cummins and Macintyre
found more large superstores (with greater food availability) in the more deprived
areas.78 Most work has focused on location of stores and, in some cases,
availability and costs of some foods. There has been little research relating specific
features of the local food environment to the actual dietary behaviors of individuals.
In particular, changes in dietary patterns in response to local changes in food
availability and cost have not been examined.
Other mechanisms through which the environment may influence dietary patterns
include food consumption in restaurants and food advertising. Some work
suggests that eating out and consumption of fast foods are associated with increased
body mass.79,80 Larger portion sizes have been shown to be associated with greater
energy intake.12 Food portion sizes are greater at fast-food restaurants than at other
types of restaurants and have increased in the United States over the last three
decades, with the greatest increases observed for food consumed at fast-food establishments.
81 However, the relation between local availability of different types of
eating places and dietary patterns has not been examined. Food advertising is yet
another mechanism through which the local environment may affect dietary patterns.
82 Differences in tobacco advertising across neighborhoods have been documented.
83–85 The extent to which these differences also exist for food advertising
has not been determined.
Complexities in studying the effects of the local food environment on dietary
patterns include disentangling the types and number of stores available from the
actual foods offered for sale at these stores. Although some of the research looking
at distribution of types of stores is predicated on the notion that large stores often
offer better prices and more choice, this may not always be empirically true. Moreover,
the location of these large stores (as well as what they offer compared to
smaller stores) may differ from context to context (or in different countries) as well
as over time, based on economic conditions and the locational strategies of large
store chains.86 Availability of foods and their consumption are likely to be reciprocally
related and mutually reinforcing. Cultural factors may play a role in shaping
neighborhood consumption patterns, and neighborhood consumption patterns may
in turn influence food availability and cost.77,87 This makes it difficult to empirically
RESIDENTIAL ENVIRONMENTS AND CARDIOVASCULAR RISK 579
isolate the effects of food cost and availability on dietary patterns in observational
studies.
Other Mechanisms Possibly Linking Residential
Environments to Cardiovascular Disease
There is some evidence that social support and depression, two psychosocial factors
for cardiovascular disease, may be affected by neighborhood physical environments.
It has been hypothesized that features of the built environment of neighborhoods
(such as the presence and attractiveness of public spaces or the design of homes)
may facilitate or discourage social interactions among neighbors.88,89 The presence
of multiple activities within a neighborhood (including residential, commercial, and
recreational activities, for example) is likely to enhance connections between residents.
People’s sense of attachment to their neighborhood may also be influenced
by the quality of the built environment. All these factors may influence the social
support experiences by residents. Limited data also suggest that living in socioeconomically
disadvantaged neighborhoods and other features of the built environment
may be associated with increased risk of depression, even after individual-level
factors are taken into account.90–92 Features of residential environments, including
aesthetic quality, absence of natural environments, and noise, may serve as stressors.
Studies have suggested that exposure to natural environments (as opposed to
urban environments) has a stress-reducing or restorative influence.93,94 Some work
has linked exposure to noise with cardiovascular risk factors such as elevated blood
pressure and with risk of myocardial infarction, but evidence of this relationship is
mixed.95 Noise and lighting may also affect quantity and quality of sleep, and sleep
disturbance has been linked to glucose metabolism96 as well as cardiovascular risk
factors and mortality.97 Evidence is mounting regarding the possible effects of air
pollution (especially small airborne particulates) on the development of cardiovascular
disease and the clinical expression of underlying disease.98,99 Thus, although
research remains scant, it is conceivable that noise, air pollution, and traffic could
contribute to area differences in cardiovascular risk.
CHALLENGES TO FUTURE RESEARCH
Although existing research suggests that features of residential environments are
related to cardiovascular risk, important questions remain regarding whether the
observed associations reflect causal processes. Addressing a series of conceptual and
methodological challenges in future research may help strengthen causal inferences
regarding the presence of these effects. A first set of challenges, which pertains
to identifying the relevant environments and characterizing them, is closely related
to the need to better specify the theoretical models and processes that may link
residential environments and cardiovascular health and empirically test specific
hypotheses derived from these models. A second set of challenges is related to study
design and the inherent limitations of observational studies in drawing causal inferences.
Characterizing Environments
Definition and Geographic Scale A key issue is defining the appropriate geographic
scale for which features of the environment should be measured. In study580
DIEZ ROUX
ing the effects of residential environments on health, it is important to recognize
that geographic areas of different sizes may be relevant for different processes and
outcomes. Features of the immediate residential environment (e.g., the block) may be
important for some outcomes (such as social interactions), whereas much broader
areas (e.g., the geographic area within which shopping is done) may be relevant for
other outcomes (such as dietary habits). Area sizes, features of the area, processes,
and outcomes are therefore linked. Different scales may thus be relevant for different
research questions.
Most research to date has paid little attention to issues of the relevant scale in
studying the effects of residential environments on health, due in part to the necessary
reliance on existing data, which is often available only for certain levels of
aggregation (and is often not available for the most theoretically relevant level).
The need to pay more attention to issues of scale has also been noted in the urban
planning literature. In studying transportation choices, for example, Handy et al.48
note how features of the region may be more important to automobile trips than
features of the immediate neighborhood, whereas features of the neighborhood may
be more relevant to decisions regarding walking. Thus, comprehensive examinations
of how residential environments (or the built environment) are related to
cardiovascular risk may need to incorporate multiple measures at multiple scales.
A priori theorizing on what constructs and what scales are most relevant to a specific
outcome will be key. Qualitative investigations of how individuals relate to
their immediate and more distant environments and how features of these environments
may affect their decisions and behaviors may also provide guidance on what
specific features of built environment are likely to be especially relevant, and therefore
worthy of investigation in large-scale quantitative studies.
Although neighborhoods have been the focus of recent research, other built environments
such as the work or school environments may be especially relevant for
some population groups. The importance of the immediate residential environment
and the relevance of other environments to daily life may vary by personal characteristics,
time, and place. Thus, research on environmental determinants of cardiovascular
risk will need to include the multiple environments that people interact
with in their daily lives. It is also important to note that certain factors may be
pervasive across a society and thus difficult to identify as determinants of cardiovascular
risk in a study that focuses on interindividual differences within a society,
even if their impact on the risk of cardiovascular disease is large.100 For example,
mass production and marketing of foods and reliance on the automobile for transportation
(with its associated effects on public transportation and facilities for walking
and cycling) are pervasive in many industrialized societies. Thus, almost everyone
is exposed to environments that stimulate consumption of large amounts of
food often high in fat and salt, and everyone is exposed to environments that hinder
walking and cycling in daily life. The relevant scale in this case may be the country
as a whole, rather than the neighborhood. Studies that focus on interindividual
differences within a country will be unable to detect the effects of these ubiquitous
conditions.
Conceptualization and Measurement of Relevant Features Another important limitation
of existing work pertains to the conceptualization and measurement of the
relevant area features. As has been noted in both the public health and urban planning
and transportation literatures,48,59,101,102 a large part of existing empirical work
has relied on existing databases to characterize areas. Available data is often used
RESIDENTIAL ENVIRONMENTS AND CARDIOVASCULAR RISK 581
as a crude proxy for the relevant variables of interest. Promising trends in the measurement
of area-level constructs include systematic social observation,103,104 surveys
of area residents, and the use of GIS. Systematic social observation is the technique
by which trained observers are sent to the areas of interest to rate them on a series
of dimensions in a systematic fashion. Special statistical techniques (ecometrics)47 are
then used to aggregate the ratings of multiple observers per area in order to obtain
more valid and reliable measures. Surveys can also be used to obtain information
on the characteristics of areas from area residents using validated scales. The responses
of different residents within a given area to a series of scale items are then
also aggregated to obtain a measure for the area as a whole. An advantage of the
use of surveys is that information on area characteristics, such as the aesthetic quality
of the area or the availability of recreational resources, can be obtained from a
different set of respondents than the outcome measures (e.g., reported frequency of
walking or physical activity), thus avoiding the possibility of same-source bias (i.e.,
when the same person is asked about the availability of recreational resources in
an area and the extent to which he or she is physically active). Both of these techniques
have been successfully applied in the social sciences to study neighborhood
effects on violence and criminal behavior.105 Their use to investigate how features
of the built environment may be related to health outcomes is a promising trend in
the field.
The enormous development of GIS technology106 over the past few years and
the increasing availability of geographically coded information have generated enormous
possibilities for the study of residential environments and health. The flexibility
of GIS measures allows estimation of measures for areas of varying sizes, depending
on the research question. A wide variety of measures can be investigated.
Examples include measures of density of resources, average distances to resources,
and street connectivity.107,108 However, the conceptualization and development of
GIS-derived measures relevant to a specific outcome remains a key challenge.
Study Design and the Limitations
of Observational Studies
A large part of existing research linking residential environments or features of
the built environment to health outcomes has been cross-sectional, relating health
behaviors to features of places of residence at a single point in time. Although some
studies have investigated how area characteristics (usually aggregate measures of
the socioeconomic environment) are related to mortality, the incidence of disease,
or changes in health over time,17,20,109,110 longitudinal studies that relate changes in
area characteristics to changes in health behaviors are still rare. The longitudinal
approach would help reduce the possibility of reverse causation as an explanation
for the associations observed if changes in residential environments can be related
to changes in outcomes. For example, existing work reporting cross-sectional associations
between the local food environment and dietary habits is prone to the
criticism that food availability may result from dietary patterns, rather than the
other way around. Demonstrating a relationship between changes in food availability
and subsequent changes in dietary habits in a longitudinal study would help
strengthen inferences regarding a causal relationship between food availability and
diet.
Clearly, studies of the effects of residential environments on health must include
data on both areas and individuals. The individual-level data are of crucial importance
for several reasons. Because the persons residing in different types of areas
582 DIEZ ROUX
differ in their individual-level characteristics, it is important to control for these
characteristics in estimating area effects. For example, because income is known to
be inversely related to physical activity, and persons of low income may live in
areas with less resources, it is important to control for personal income in studying
whether the area availability of recreational resources is related to physical activity.
Another reason for including information on individual characteristics is that area
effects may be especially important for some types of individuals and less important
for others. For example, the aesthetic quality and safety of the local area may be
especially relevant to physical activity in the elderly, who may derive most of their
activity from walking in a relatively restricted geographic area. Features of the local
area may be less relevant to working adults, who may be less likely to spend time
around their homes. The fact that many health-related behaviors are established
early in life highlights the importance of studying environmental effects on children,
who may be especially vulnerable to the effects of the environments they are exposed
to at home or at school.111–113
Additional complexity results from the fact that individuals may compensate
for deficiencies in their local areas through a variety of mechanisms. For example,
persons who live in areas with few recreational resources may still be physically
active because they use resources at the workplace. Wealthy persons may live in
suburbs with no sidewalks where walking to the store is virtually unheard of, but
they may get physical activity by joining expensive sports facilities or health clubs
near their workplaces. Relationships may sometimes be complex and extend in unexpected
directions. Despite having fewer and poorer quality recreational resources
in their local areas, low-income persons living in more disadvantaged neighborhoods
may be more physically active than wealthier persons in the suburbs, simply
because they walk more due to differences in transportation and features of land
use in rich and poor neighborhoods.114 Failure to consider these compensatory
mechanisms and complex relationships would result in an incorrect estimate of area
effects on health.
The vast majority of work linking residential environments to cardiovascular
risk has been observational. The key problem in observational studies in this field
is that persons may be selected into different types of residential areas based on
individual-level characteristics that may themselves be related to cardiovascular
risk. Although researchers have attempted to account for these individual-level differences
by statistically controlling for individual-level variables, the possibility that
the area differences that persist after adjustment are due to mismeasured or omitted
individual characteristics remains. (It is worth noting that this problem, often referred
to technically as residual confounding, is common to observational studies
generally and is not unique to this area of research.) This limitation, sometimes
referred to as the selection problem, has been repeatedly noted in the social sciences,
public health, and urban planning literatures.53,66,67 Another problem is that, because
there is often relatively little overlap between the personal characteristics of persons
living in different types of neighborhoods, the statistically adjusted estimates are
based on a series of assumptions that may not be empirically testable. One option
to avoid the confounding by SES issue noted above is the comparison of neighborhoods
selected to be different in built environments (e.g., different mixes of land
uses) but similar in socioeconomic composition.59 Unfortunately, this approach is
not always possible because socioeconomic and other features are often correlated.
The limitations of observational studies have led some researchers to call for
increasing the use of intervention studies and randomized community trials in the
RESIDENTIAL ENVIRONMENTS AND CARDIOVASCULAR RISK 583
study of the effects of residential environments on health.115 Although experimental
approaches would obviously solve many of the problems of observational studies,
they have their own set of methodological problems, as noted in past work on community
interventions to prevent cardiovascular disease.116–118 One crucial limitation
is that it may not always be logistically feasible to conduct a community trial of
how features of the residential environment (or of the built environment) affect
health. Although the reliance on observational studies is likely to continue, exploring
the use of natural experiments or community trials when feasible would be an
important step forward in this field. An area likely to yield useful information is
the evaluation of the behavioral and health effects of interventions that modify the
built environment (e.g., changes in transportation policies such as policies that reduce
automobile traffic or interventions that modify features of residential environments
such as sidewalks, traffic flow, or green spaces). This approach has been
used in the transportation literature, where, for example, case studies of the effects
of traffic-calming policies have shown that they generally increase walking and
bicycling.59 To date, the changes in the built environment that occur frequently in
many neighborhoods, as a result, for example, of community actions, commercial
interests, or government policy, have not been systematically evaluated with respect
to their health effects.
CONCLUSION
The recent interest in residential environments is part of a new paradigm in public
health which recovers the notion that the determinants of health can be defined at
multiple levels. Thus, health depends not only on the biological characteristics of
individuals but also on the environments where people live, and these environments
are in turn shaped by the economic and social policies driving residential differentiation,
and shaping the features of residential environments. This new paradigm views
health behaviors as constrained, at least in part, by the environmental contexts in
which they occur. Importantly, the word environment is used to encompass not
only physical and chemical exposures but also features of built environments at
multiple levels ranging from the walkability of neighborhoods to the organization
of transportation in a large metropolitan area. Under this paradigm, transportation
and urban planning policies are, in fact, health policy. Modifying these broad environmental
factors may be a more effective mechanism to increase physical activity
(e.g., increasing walking by making walking easier, more pleasant, and less costly
than driving) than encouraging individuals to go to an exercise facility once a day.15
Modifying the availability and relative cost of healthy foods may also do more for
improving diet than relying on health education campaigns. Environmental and
individual-based strategies are likely to be synergistic: environmental change is
likely to enhance and facilitate the effects of strategies that focus on individuals,
and individual-based strategies may increase awareness of the need for environmental
change and may enhance support for this change.
The need to consider environmental contexts in preventing cardiovascular risk
continues to emerge in policy documents119 and initiatives (e.g., the Active Community
Environments Initiative at CDC, www.cdc.gov/nccdphp/dnpa/aces.htm). Although
the impact of environmental contexts on cardiovascular health may appear
to many to be self-evident, rigorous empirical evidence is needed to support calls
for policy change and also to help identify the interventions and policies that are
likely to be most effective. Key areas for future work include: (a) increasing the
584 DIEZ ROUX
specificity of the hypotheses tested in terms of both environmental factors and
health outcomes; (b) improving the measurement of the environmental features of
interest; and (c) evaluating the health effects of “naturally” occurring environmental
changes (e.g., the creation of a new public space, changes in food availability,
or changes in automobile, transportation, or land use policies). In addressing these
research needs, work that transcends disciplinary boundaries, bringing together urban
planners, sociologists, geographers, behavioral scientists, environmental scientists,
and epidemiologists, will be key.
