Content uploaded by Catarina Kiefe
Author content
All content in this area was uploaded by Catarina Kiefe
Content may be subject to copyright.
American Journal of Epidemiology
ªThe Author 2010. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of
Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org.
Vol. 172, No. 3
DOI: 10.1093/aje/kwq098
Advance Access publication:
July 7, 2010
Original Contribution
Do Neighborhood Socioeconomic Deprivation and Low Social Cohesion Predict
Coronary Calcification?
The CARDIA Study
Daniel Kim*, Ana V. Diez Roux, Catarina I. Kiefe, Ichiro Kawachi, and Kiang Liu
*Correspondence to Dr. Daniel Kim, Department of Society, Human Development, and Health, Harvard School of Public Health, 677
Huntington Avenue, 7th Floor, Boston, MA 02115 (e-mail: dkim@hsph.harvard.edu).
Initially submitted December 23, 2009; accepted for publication March 30, 2010.
Growing evidence suggeststhat neighborhood characteristics may influence the risk of coronary heart disease. No
studies have yet explored associations of neighborhood attributes with subclinical atherosclerosis in younger adult
populations. Using data on 2,974 adults (1,699 women, 1,275 men) aged 32–50 years in 2000 from the Coronary
Artery Disease Risk Development in Young Adults (CARDIA) Study and 2000 US Census block-group-level data, the
authors estimated multivariable-adjusted associations of neighborhood socioeconomic deprivation and perceived
neighborhood cohesion with odds of coronary artery calcification (CAC) 5 years later. Among women, the quartiles of
highest neighborhood deprivation and lowest cohesion were associated with higher oddsof CAC after adjustment for
individual-level demographic and socioeconomic factors (for deprivation, odds ratio ¼2.49, 95% confidence interval:
1.22, 5.08 (Pfor trend ¼0.03); for cohesion, odds ratio ¼1.87, 95% confidence interval: 1.10, 3.16 (Pfor trend ¼
0.02)). Associations changed only slightly after adjustment for behavioral, psychosocial, and biologic factors. Among
men, neither neighborhood deprivation nor cohesion was related to CAC. However, among men in deprived neigh-
borhoods, low cohesion predicted higher CAC odds (for interaction between neighborhood deprivation and cohesion,
P¼0.03). This study provides evidence on associations of neighborhood deprivation and cohesion with CAC in
younger, asymptomatic adults. Neighborhood attributes may contribute to subclinical atherosclerosis.
atherosclerosis; coronary disease; residence characteristics; risk factors; social environment
Abbreviations: CAC, coronary artery calcification; CARDIA, Coronary Artery Disease Risk Development in Young Adults; CHD,
coronary heart disease; CI, confidence interval; OR, odds ratio; SEP, socioeconomic position.
Growing evidence suggests that neighborhood socioeco-
nomic context may influence the risk of coronary heart dis-
ease (CHD) (1–6). Neighborhood socioeconomic position
(SEP), reflecting the relative social and economic position
of neighborhoods, may be closely linked to a variety of
material/physical amenities and resources relevant to CHD
risk (7). For instance, the local availability and quality of
nutritious foods and green spaces plausibly contribute to
CHD behavioral risk factors (8–12).
Aside from the material/physical environment, an attribute
of the neighborhood social environment that may potentially
shape CHD risk is social cohesion, defined as the presence
of strong social bonds, including interpersonal trust (13).
Posited mechanisms for health effects include diffusion of
knowledge about health-related behaviors (e.g., dietary prac-
tices), maintenance of healthy behaviors through informal
social control, generation of psychosocial processes includ-
ing social support, and greater collective efficacy in improv-
ing local amenities and services (13). Social cohesion (or the
related concept of social capital) at the state and community
levels has predicted wide-ranging health outcomes, including
obesity (14), self-rated health (15, 16), smoking (17), depres-
sion (18), and CHD incidence (19).
While researchers have explored the associations between
neighborhood socioeconomic deprivation, social cohesion/
capital, and individual health endpoints, few investigators
288 Am J Epidemiol 2010;172:288–298
at University of Michigan on November 29, 2010aje.oxfordjournals.orgDownloaded from
have studied these neighborhood attributes in relation to cor-
onary artery calcification (CAC), a marker for underlying
CHD; to our knowledge, no investigators have done so in
younger adult populations. Prospective analyses of asymp-
tomatic adults have shown CAC to predict CHD (20–22).
The relations between neighborhood deprivation, cohe-
sion, and individual CAC may vary by gender and race/
ethnicity, as findings for related endpoints suggest. For in-
stance, several studies have found stronger associations
of neighborhood characteristics with CHD outcomes in
women than in men (3, 5, 23) and in whites than in blacks
(24). Neighborhood deprivation and low social cohesion
may further act synergistically to affect health (25).
We used data from a population-based sample of younger
adults to investigate associations of neighborhood depriva-
tion and low perceived neighborhood cohesion with CAC.
We hypothesized that higher deprivation and lower cohe-
sionwouldeachbeassociatedwithhigherCACpreva-
lence. We looked for heterogeneous associations by
gender (hypothesizing stronger associations in women)
and race/ethnicity (positing stronger associations in
whites) and explored whether associations of neighbor-
hood cohesion with CAC were modified by neighborhood
deprivation. We also investigated the extent to which psy-
chosocial, behavioral, and biologic factors could explain
the neighborhood associations.
MATERIALS AND METHODS
Study population
The Coronary Artery Disease Risk Development in
Young Adults (CARDIA) Study is a prospective cohort
study exploring predictors of the development of CHD risk
factors in young adults (26). At the initial examination in
1985, the cohort consisted of 5,115 black and white men
and women aged 18–30 years living in 4 US urban areas:
Birmingham, Alabama; Chicago, Illinois; Minneapolis,
Minnesota; and Oakland, California. Baseline response
rates were 36.2% in Birmingham, 50.3% in Chicago,
58.7% in Minneapolis, and 65.0% in Oakland (26). In each
area, the recruitment goal was to enroll adults in approxi-
mately equal numbers of blacks and whites, women and
men, persons aged <25 years and 25 years, and persons
with a high school education or less and persons with more
than a high school education. Compared with eligible per-
sons who did not participate at baseline, study subjects
were more likely to be white, male, older, and more edu-
cated (26). The baseline examination included measures of
established and putative CHD risk factors. Follow-up ex-
aminations took place in 1987, 1990, 1992, 1995, 2000, and
2005.
Outcome variable
CAC in the main coronary arteries (left main, left circum-
flex, left anterior descending, and right) was assessed at the
2005 examination using 2 computed tomography scans
taken 1–2 minutes apart (3-mm slice thickness for electron-
beam computed tomography and 2.5-mm slice thickness for
multidetector computed tomography). A cardiovascular radi-
ologist scored the scans with imaging software. A blinded
expert adjudicated each scan set with at least 1 nonzero score
and a random sample of scan sets. For scan sets with only
1 nonzero score judged to be artifactual, the score was set to
zero. In 2000, there was high interreader agreement based on
a similar adjudication process; 2 reviewers agreed in 91% of
cases (95% of concordant scan sets, 82% of discordant scan
sets).
The study outcome was a dichotomous variable corre-
sponding to the presence or absence of CAC in 2005.
Predictor variables
Participants’ residential addresses in 1995 were linked to
2000 US Census variables, including aggregate income, ed-
ucation, and occupational data, at the Census block-group
level (as proxies for neighborhoods, with each containing
1,000 residents on average). Neighborhood characteristics
in 2000 were examined in relation to CAC in 2005 to
incorporate a plausible lag period for hypothesized effects
of neighborhood characteristics. Prospective analyses with
follow-up periods of less than a decade have found signif-
icant relations between higher neighborhood deprivation
and CHD incidence (3, 5, 23). We used 6 block-group-level
variables corresponding to dimensions of income/wealth
(log of median household income, log of median value of
housing units, and percentage of households receiving in-
terest/dividend/net rental income), education (percentage of
adults aged 25 years who had completed high school,
percentage of adults aged 25 years who had completed
college), and occupation (percentage of employed persons
aged 16 years in executive/managerial/professional spe-
cialty occupations). For each variable, we derived a stan-
dardized zscore. Neighborhood SEP scores were calculated
by taking the mean value across all zscores, with lower
scores indicating lower neighborhood SEP and higher so-
cioeconomic deprivation (3).
Perceived neighborhood cohesion was measured in the
2000 CARDIA follow-up questionnaire as the individual’s
mean score on a 5-point Likert scale using the following 5
items (the last 2 items being reverse-coded): ‘‘People around
here are willing to help their neighbors,’’ ‘‘this is a close-
knit neighborhood,’’ ‘‘people in this neighborhood can be
trusted,’’ ‘‘people in this neighborhood generally don’t
get along with each other,’’ and ‘‘people in this neighbor-
hood do not share the same values.’’ This scale has exhibited
acceptable internal consistency reliability (27, 28).
Covariates
Data on individual-level covariates were gathered from
the 2000 CARDIA examination and consisted of age, gen-
der, race/ethnicity, income, education, health-care access
(i.e., having a usual source of medical care), and study site.
Neighborhood-level covariates and potential confounders
consisted of the percentage black, immigrant concentration
(mean of standardized percentages of Hispanics and foreign-
born residents), and residential stability (mean of standard-
ized percentages of persons living in the same house over
Neighborhood Attributes and Subclinical Atherosclerosis 289
Am J Epidemiol 2010;172:288–298
at University of Michigan on November 29, 2010aje.oxfordjournals.orgDownloaded from
the past 5 years and owner-occupied housing). These data
were derived from the 2000 US Census.
Potential mediators
The following psychosocial, behavioral, and biologic
factors measured in 2005 were explored as potential me-
diators of the associations of neighborhood characteristics
with CAC: psychosocial factors—depression (depressive
symptom scores of 0–100 on the Center for Epidemiologic
Studies Depression Scale (29)); behavioral factors—
physical activity (weighted average of the intensity of
moderate and vigorous physical activity over the past year,
in metabolic equivalents (30)) and smoking (number of
cigarettes smoked per day); biologic factors—systolic/
diastolic blood pressure (120/80 mm Hg), serum low
density lipoprotein cholesterol (130 mg/dL), serum high
density lipoprotein cholesterol (<40 mg/dL), fasting blood
glucose (110 mg/dL), and body mass index (weight (kg)/
height (m)
2
;30). The biologic factors have previously
predicted CAC (31).
Statistical methods
We used multivariable logistic regression models to esti-
mate the associations between higher neighborhood socio-
economic deprivation (equivalent to lower neighborhood
SEP) and the odds of CAC and between lower perceived
Table 1. Individual- and Neighborhood-Level Demographic and Socioeconomic
Characteristics of Participants in 2000 and the Presence of Coronary Artery Calcification in 2005,
by Sex, Coronary Artery Disease Risk Development in Young Adults (CARDIA) Study
a,b
Women Men
Individual characteristics (n¼2,974)
No. of participants 1,699 1,275
Coronary artery calcification in 2005, %
Yes 10.8 28.9
No 89.2 71.1
Perceived neighborhood cohesion
c
3.7 3.6
Mean age, years 45.3 (3.7) 45.3 (3.6)
Race/ethnicity, %
Black 48.6 41.2
White 51.4 58.8
Marital status, %
Married/cohabitating 52.6 57.8
Divorced/separated/widowed 18.2 12.2
Never married 18.8 19.8
Other/missing data 10.4 10.2
Educational attainment, %
High school or less 17.5 22.0
Some college/college 52.2 48.1
Graduate school 20.3 20.1
Unspecified/missing data 9.9 9.9
Annual household income ($US), %
<16,000 8.5 6.2
16,000–34,999 15.2 12.2
35,000–49,999 14.4 13.6
50,000–74,999 20.7 18.2
75,000–99,999 12.0 15.4
100,000 18.4 23.5
Unspecified/missing data 10.9 10.9
Access to health care
d
,%
Yes 87.2 81.8
No 2.9 8.2
Unspecified/missing data 9.9 10.0
Table continues
290 Kim et al.
Am J Epidemiol 2010;172:288–298
at University of Michigan on November 29, 2010aje.oxfordjournals.orgDownloaded from
neighborhood cohesion and the odds of CAC. Given the
overall average of 1.4 participants per neighborhood and
the small intraclass correlation (~0), multilevel models/
methods to account for within-neighborhood correlations
were not applied. Findings from the multivariable regression
models are reported.
In all models, the highest levels of neighborhood SEP and
cohesion served as the reference categories. An indicator
variable was used to code for missing values on cohesion
(comprising approximately 10% of observations in each of
women and men). To test for trend, we modeled quartiles
of higher neighborhood deprivation/lower cohesion (after
excluding missing values) as an ordinal variable and noted
the associated Pvalue.
To assess the presence of mediation, we looked for
attenuation in the odds ratio estimates for neighborhood
deprivation/low cohesion after each set of risk factors was
added to the respective model.
All reported findings are stratified by gender. A stronger
positive association was observed between higher neighbor-
hood deprivation and CAC in women than in men. In the
gender-combined model, cross-product terms correspond-
ing to the interactions between the higher quartiles of neigh-
borhood deprivation and gender were jointly significant
(P¼0.001). No significant interactions were observed be-
tween race/ethnicity and neighborhood SEP or cohesion; in
models stratified by race/ethnicity, similar but less precise
point estimates were observed for blacks and whites. There-
fore, our reported results are stratified by gender and adjusted
for race/ethnicity.
We further tested for effect modification of the association
between low neighborhood cohesion and CAC by
Table 1. Continued
Women Men
Study center, %
Minneapolis, Minnesota 25.3 28.5
Birmingham, Alabama 22.1 24.5
Chicago, Illinois 22.7 22.8
Oakland, California 29.8 24.2
Neighborhood characteristics (n¼2,185)
No. of distinct neighborhoods 1,237 948
Socioeconomic deprivation
Median household income, 1,000 $US 52.8 (27.9) 52.6 (26.1)
Median house value, 1,000 $US 189.1 (146.1) 188.1 (139.4)
Mean % of households receiving interest,
dividend, or rental income
37.9 (20.1) 39.1 (19.3)
Mean % of adults (aged 25 years)
with high school education or more
83.1 (14.1) 85.3 (12.4)
Mean % of adults (aged 25 years)
with college education or more
39.3 (22.9) 41.4 (21.8)
Mean % of persons aged 16 years
employed in executive, managerial,
or professional occupations
38.9 (18.4) 39.9 (17.6)
Mean % black 25.5 (33.1) 21.0 (30.2)
Mean % immigrant
Hispanic 8.9 (13.5) 8.1 (12.0)
Foreign-born 12.1 (12.5) 11.9 (12.0)
Residential stability
Mean % living in same household
during the past 5 years
53.9 (15.3) 52.8 (15.8)
Mean % living in owner-occupied housing 62.8 (27.3) 61.8 (27.5)
a
For continuous variables, mean values (with standard deviations in parentheses) are dis-
played. For categorical variables, the percentage of the sample (women or men) in each category
is shown.
b
Data on individual-level characteristics (except for coronary artery calcification) were col-
lected in 2000. The presence of coronary artery calcification was ascertained in 2005. Data on
neighborhood characteristics were derived from the 2000 US Census (with addresses ascertained
in 1995).
c
Range of scores was 1–5 in both women and men; higher values indicate higher perceived
neighborhood cohesion.
d
Defined as having a usual source of medical care.
Neighborhood Attributes and Subclinical Atherosclerosis 291
Am J Epidemiol 2010;172:288–298
at University of Michigan on November 29, 2010aje.oxfordjournals.orgDownloaded from
neighborhood deprivation (dichotomized using the median
neighborhood SEP score) in both women and men.
In supplementary analyses, we repeated the main analyses
for participants who were living in the same homes in 1995
and 2000. In addition, to check the robustness of our find-
ings, we employed multiple imputation methods to impute
missing values for CAC and cohesion (by creating multiply
imputed data sets using logistic regression and combining
the results) (32). We also repeated the main analyses
using linear regression, modeling the outcome as continuous
(logarithm of CAC score plus 1).
RESULTS
Addresses in 1995 for 3,320 of the 3,549 study partici-
pants (93.5%) seen at the 2005 CARDIA examination were
linked to 2000 US Census block-group-level variables. Ex-
clusion of participants with missing CAC values in 2005
yielded a total sample of 2,974 persons (1,699 women and
1,275 men, of whom 184 and 369 had CAC, respectively) in
2,185 neighborhoods (Table 1). Approximately two-thirds
of women (63%) and men (64%) lived in the same homes in
1995 and 2000.
The mean age of the overall sample in 2000 was 45.3 years
(range, 32–50 years); 57.1% were female, 54.6% were white,
and 45.4% were African-American. Women were more
likely to be African-American and to report lower income
and education, although gender differences by income and
education were not large. There were no substantial differ-
ences in neighborhood characteristics between men and
women. CAC prevalence was substantially lower in women
(10.9% in women vs. 28.9% in men). A comparison of the
2005 and 1985 baseline samples of women and men (by race/
ethnicity, income, and education) suggested greater selective
attrition in men.
Internal consistency reliability estimates for the neighbor-
hood SEP and cohesion measures were high (Cronbach’s
avalues were 0.94 and 0.82, respectively). In the overall
sample, neighborhood SEP scores were strongly inversely
correlated with the percentage of black residents (r¼0.63),
weakly inversely correlated with the percentage of immi-
grants (r¼0.13), and positively correlated with residen-
tial stability (r¼0.21). Higher neighborhood SEP and
residential stability were positively correlated with neigh-
borhood cohesion (r¼0.26 and r¼0.17, respectively).
Being female, being white, having a higher income, being
widowed, divorced, or separated, and never being married
were also related to higher cohesion.
In women (Table 2), the highest quartile of neighborhood
deprivation was associated with 2.49 times’ higher odds of
CAC, controlling for covariates and perceived neighbor-
hood cohesion (Pfor trend ¼0.03; model 2). Persons in
the quartile corresponding to the lowest level of cohesion
had 1.87 times’ higher adjusted odds of CAC than persons
in the quartile of highest cohesion (Pfor trend ¼0.02;
model 2). Associations of neighborhood deprivation and
cohesion with CAC were very similar before and after
adjustment for individual sociodemographic and socioeco-
nomic characteristics (compare model 1 with model 2).
Additional adjustment for physical activity and smoking
slightly attenuated associations with neighborhood depri-
vation (model 3), and separate adjustment for Center for
Epidemiologic Studies Depression Scale score slightly
attenuated associations with cohesion (model 4), although
both main effects remained statistically significant. Adjust-
ment for biologic factors (model 5) also resulted in minor
changes. When all potential risk-factor mediators were
added, the neighborhood associations remained but were
no longer statistically significant for deprivation (for quar-
tile of highest deprivation, odds ratio (OR) ¼2.02, 95%
confidence interval (CI): 0.95, 4.32 (Pfor trend ¼0.16); for
quartile of lowest cohesion, OR ¼1.80, 95% CI: 1.02, 3.16
(Pfor trend ¼0.04)).
Neighborhood deprivation was not associated with CAC
in men (Table 3). Men in quartiles 2, 3, and 4 of neighbor-
hood cohesion (indicating lower levels of cohesion) had
higher CAC odds than those in the lowest quartile, but
confidence intervals were wide and only the estimate for
quartile 2 was statistically significant; there was no dose-
response trend. These estimates were largely unchanged
after adjustment for individual-level risk factors.
Table 4 shows associations of neighborhood cohesion with
CAC for men in deprived and nondeprived neighborhoods
separately. The interaction between neighborhood depriva-
tion and cohesion was statistically significant (P¼0.03). In
deprived neighborhoods, all 3 lower quartiles of cohesion
were associated with higher odds of CAC in comparison with
the highest quartile, with a marginally statistically significant
trend (P¼0.07). No association between social cohesion
and CAC was seen in men in nondeprived neighborhoods
(model 1). For women, lower cohesion predicted higher odds
in both deprived and nondeprived neighborhoods, although
associations were significant only in nondeprived neighbor-
hoods (data not shown; for interaction between neighborhood
deprivation and cohesion, P¼0.34).
After restricting the main analyses to persons living in the
same homes in 1995 and 2000, neighborhood deprivation
associations adjusted for demographic/socioeconomic cova-
riates and neighborhood cohesion became stronger in
women (for quartile of highest deprivation, OR ¼3.93,
95% CI: 1.52, 10.13) and were relatively unchanged in men.
In multiple imputation analyses, odds ratio point esti-
mates for neighborhood deprivation and cohesion differed
by less than 10% in comparison with corresponding
estimates from models without imputation.
With the logarithm of (CAC score þ1) as the outcome,
we observed qualitatively similar patterns. For example, in
women, the highest quartile of neighborhood deprivation
was associated with a 26% higher CAC score (P¼0.02);
the lowest level of cohesion was associated with a 14%
higher CAC score (P¼0.08).
DISCUSSION
We found independent, graded associations of higher
neighborhood deprivation and lower social cohesion with
the presence of CAC 5 years later in women. No associa-
tions were observed in men overall. However, low social
cohesion was associated with higher odds of CAC among
292 Kim et al.
Am J Epidemiol 2010;172:288–298
at University of Michigan on November 29, 2010aje.oxfordjournals.orgDownloaded from
men living in poorer neighborhoods, although no clear dose-
response was found.
A key strength of our analysis was the investigation of an
established marker of subclinical atherosclerosis in young
adults. This helped us to avoid biases that may occur in
cross-sectional analyses of prevalent clinical outcomes
when clinical symptomatic disease influences residential
location. Moreover, the investigation of CAC in a young
adult sample allows the detection of associations with very
early disease, long before it becomes symptomatic. Thus,
our analyses demonstrated that, particularly in women,
neighborhood characteristics predict the presence of very
early disease.
Our finding that neighborhood characteristics predict very
early CHD agrees with past studies of neighborhood depri-
vation and subclinical atherosclerosis (33–36). To our knowl-
edge, this study is among the first to document this relation in
younger adults, and represents the first US study of these
characteristics and coronary calcification. Our findings are
also consistent with the magnitude of associations between
neighborhood deprivation and CHD incidence/mortality
documented in prospective studies (3, 5, 36). They add
Table 2. Odds Ratios for Coronary Artery Calcification Among Women in 2005 Associated With Neighborhood- and Individual-level
Characteristics in 2000 (n¼1,699), Coronary Artery Disease Risk Development in Young Adults (CARDIA) Study
Model 1 Model 2 Model 3 Model 4 Model 5
OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI
Neighborhood-level predictors
Quartile of socioeconomic
deprivation
a
2 1.89*1.13, 3.18 1.81*1.06, 3.11 1.80*1.05, 3.09 1.75*1.01, 3.01 1.74 0.99, 3.04
3 1.47 0.83, 2.60 1.44 0.79, 2.63 1.38 0.75, 2.53 1.29 0.70, 2.38 1.45 0.78, 2.70
4 2.46*1.26, 4.82 2.49*1.22, 5.08 2.39*1.17, 4.91 2.22*1.07, 4.58 2.43*1.16, 5.09
Pfor trend 0.02 0.03 0.046 0.08 0.04
Quartile of low perceived
neighborhood
cohesion
a
2 1.25 0.74, 2.11 1.24 0.73, 2.11 1.23 0.72, 2.09 1.18 0.68, 2.02 1.30 0.76, 2.24
3 1.56 0.96, 2.55 1.47 0.89, 2.43 1.44 0.87, 2.40 1.43 0.86, 2.38 1.53 0.91, 2.58
4 1.84*1.11, 3.07 1.87*1.10, 3.16 1.77*1.03, 3.02 1.89*1.11, 3.22 1.83*1.06, 3.17
Pfor trend 0.01 0.02 0.04 0.02 0.03
Individual-level potential
mediators
Center for Epidemiologic
Studies Depression
Scale score
b
1.10 0.94, 1.28
Intensity of physical activity
b,c
0.88 0.74, 1.06
Smoking, cigarettes/day
b
1.37** 1.21, 1.55
Blood pressure
(120/80 mm Hg)
1.84** 1.30, 2.60
Low density lipoprotein
cholesterol
(130 mg/dL)
1.49*1.04, 2.15
High density lipoprotein
cholesterol
(<40 mg/dL)
2.24** 1.37, 3.65
Fasting glucose
concentration
(110 mg/dL)
1.30 0.79, 2.14
Body mass index
d
(30) 1.20 0.85, 1.72
*P<0.05; **P<0.01.
Abbreviations: CI, confidence interval; OR, odds ratio.
a
Quartile 1 for deprivation (reference category) corresponds to the quartile with the least socioeconomically deprived (i.e., ‘‘richest’’) neighbor-
hoods. Quartile 1 for cohesion (reference category) contains the persons with the highest perceived level of neighborhood cohesion. In all models,
results were also adjusted for the neighborhood percentage black, percentage immigrant, and residential stability and for individual age, race/
ethnicity, marital status, educational attainment, household income, access to health care, and study center—except for model 1 (results were
adjusted for neighborhood characteristics and study center only).
b
Odds ratios correspond to a 1-standard-deviation change in the risk factor.
c
Weighted average of the intensity of moderate and vigorous physical activity over the past year, in metabolic equivalents (30).
d
Weight (kg)/height (m)
2
.
Neighborhood Attributes and Subclinical Atherosclerosis 293
Am J Epidemiol 2010;172:288–298
at University of Michigan on November 29, 2010aje.oxfordjournals.orgDownloaded from
substantially to the limited existing research on the relation
between social cohesion/capital and CHD, which is largely
confined to ecologic studies and non-US studies that used
single-item measures of social cohesion/capital (37, 38).
The present study advances the literature by showing that
social cohesion predicts subclinical atherosclerosis, even in
a relatively young adult sample.
We estimated neighborhood characteristics 5 years before
CAC assessment, making the temporal relation in our data
compatible with causation. Strengths of our analysis include
the use of a demographically and socioeconomically diverse
sample. Furthermore, we adjusted for multiple neighborhood-
and individual-level potential confounders and predictors of
CAC, which should have limited residual confounding.
In several prospective studies of neighborhood deprivation
and CHD incidence and mortality, researchers have found
stronger associations in women than in men (3, 5, 23). Similar
differences have been observed in analyses of subclinical
carotid artery intima-media wall thickness (34, 36). In a
German study of neighborhood deprivation and coronary
calcification, Dragano et al. (35) found a less clear
pattern, although they primarily used the neighborhood
Table 3. Odds Ratios for Coronary Artery Calcification Among Men in 2005 Associated With Neighborhood- and Individual-level Characteristics
in 2000 (n¼1,275), Coronary Artery Disease Risk Development in Young Adults (CARDIA) Study
Model 1 Model 2 Model 3 Model 4 Model 5
OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI
Neighborhood-level predictors
Quartile of socioeconomic
deprivation
a
2 0.71 0.49, 1.03 0.75 0.51, 1.10 0.77 0.52, 1.14 0.74 0.50, 1.10 0.73 0.49, 1.10
3 0.88 0.59, 1.30 1.00 0.65, 1.54 0.96 0.62, 1.48 0.99 0.64, 1.52 0.94 0.60, 1.47
4 1.04 0.62, 1.76 1.18 0.67, 2.10 1.19 0.67, 2.13 1.13 0.63, 2.02 1.20 0.66, 2.18
Pfor trend 0.85 0.47 0.56 0.57 0.53
Quartile of low perceived
neighborhood
cohesion
a
2 1.55*1.05, 2.28 1.54*1.03, 2.30 1.52*1.01, 2.29 1.54*1.03, 2.30 1.61*1.06, 2.46
3 1.32 0.91, 1.92 1.46 0.98, 2.16 1.54*1.03, 2.29 1.46 0.98, 2.16 1.46 0.97, 2.21
4 1.35 0.91, 2.01 1.45 0.94, 2.22 1.49 0.96, 2.31 1.48 0.96, 2.27 1.48 0.95, 2.32
Pfor trend 0.24 0.14 0.09 0.12 0.15
Individual-level potential
mediators
Center for Epidemiologic
Studies Depression
Scale score
b
1.13 0.99, 1.29
Intensity of physical activity
b,c
1.01 0.88, 1.15
Smoking, cigarettes/day
b
1.10 0.97, 1.25
Blood pressure
(120/80 mm Hg)
1.32 1.00, 1.74
Low density lipoprotein
cholesterol
(130 mg/dL)
1.49*1.12, 1.98
High density lipoprotein
cholesterol
(<40 mg/dL)
1.06 0.79, 1.44
Fasting glucose
concentration
(110 mg/dL)
1.15 0.78, 1.69
Body mass index
d
(30) 1.58*1.16, 2.14
*P<0.05.
Abbreviations: CI, confidence interval; OR, odds ratio.
a
Quartile 1 for deprivation (reference category) corresponds to the quartile with the least socioeconomically deprived (i.e., ‘‘richest’’) neighbor-
hoods. Quartile 1 for cohesion (reference category) contains the persons with the highest perceived level of neighborhood cohesion. In all models,
results were also adjusted for the neighborhood percentage black, percentage immigrant, and residential stability and for individual age, race/
ethnicity, marital status, educational attainment, household income, access to health care, and study center—except for model 1 (results were
adjusted for neighborhood characteristics and study center only).
b
Odds ratios correspond to a 1-standard-deviation change in the risk factor.
c
Weighted average of the intensity of moderate and vigorous physical activity over the past year, in metabolic equivalents (30).
d
Weight (kg)/height (m)
2
.
294 Kim et al.
Am J Epidemiol 2010;172:288–298
at University of Michigan on November 29, 2010aje.oxfordjournals.orgDownloaded from
Table 4. Odds Ratios for Coronary Artery Calcification Among Men in 2005 Associated with Perceived Neighborhood Cohesion in 2000,
According to Residence in a Deprived Neighborhood, Coronary Artery Disease Risk Development in Young Adults (CARDIA) Study
Model 1 Model 2 Model 3 Model 4
OR 95% CI OR 95% CI OR 95% CI OR 95% CI
Living in a socioeconomically
deprived neighborhood
(n¼668)
a
Neighborhood-level predictors
Quartile of low perceived
neighborhood cohesion
2 3.66** 1.77, 7.55 3.35** 1.61, 6.95 3.68** 1.78, 7.62 4.46** 2.01, 9.91
3 3.29** 1.62, 6.67 3.32** 1.63, 6.74 3.32** 1.63, 6.74 3.77** 1.73, 8.24
4 3.02** 1.46, 6.25 2.89** 1.38, 6.04 3.08** 1.48, 6.42 3.77** 1.69, 8.43
Pfor trend 0.07 0.06 0.06 0.04
Individual-level potential
mediators
CES-D score
b
1.10 0.91, 1.34
Intensity of physical activity
b
,
c
0.97 0.80, 1.17
Smoking, cigarettes/day
b
1.12 0.96, 1.32
Blood pressure
(120/80 mm Hg)
1.30 0.87, 1.94
LDL cholesterol (130 mg/dL) 1.31 0.86, 2.00
HDL cholesterol (<40 mg/dL) 1.26 0.82, 1.94
Fasting glucose concentration
(110 mg/dL)
1.23 0.73, 2.09
Body mass index
d
(30) 1.39 0.90, 2.14
Not living in a socioeconomically
deprived neighborhood
(n¼607)
a
Neighborhood-level predictors
Quartile of low perceived
neighborhood cohesion
2 0.97 0.57, 1.65 0.98 0.57, 1.68 0.97 0.57, 1.65 0.99 0.57, 1.72
3 0.97 0.57, 1.65 1.00 0.59, 1.71 0.98 0.57, 1.66 1.00 0.58, 1.74
4 1.04 0.57, 1.90 1.14 0.61, 2.10 1.06 0.58, 1.96 0.95 0.50, 1.80
Pfor trend 0.84 0.65 0.80 0.98
Individual-level potential
mediators
CES-D score
b
1.18 0.97, 1.42
Intensity of physical activity
b
1.01 0.82, 1.24
Smoking, cigarettes/day
b
1.10 0.86, 1.40
Blood pressure
(120/80 mm Hg)
1.29 0.85, 1.94
LDL cholesterol (130 mg/dL) 1.61*1.07, 2.42
HDL cholesterol (<40 mg/dL) 0.90 0.57, 1.42
Fasting glucose concentration
(110 mg/dL)
1.05 0.57, 1.94
Body mass index (30) 1.96** 1.24, 3.10
*P<0.05; **P<0.01.
Abbreviations: CES-D, Center for Epidemiologic Studies Depression Scale; CI, confidence interval; HDL, high density lipoprotein; LDL, low
density lipoprotein; OR, odds ratio.
a
Quartile 1 for cohesion (reference category) contains the persons with the highest perceived level of neighborhood cohesion. In all models,
results were also adjusted for the neighborhood percentage black, percentage immigrant, and residential stability and for individual age, race/
ethnicity, marital status, educational attainment, household income, access to health care, and study center.
b
Odds ratios correspond to a 1-standard-deviation change in the risk factor.
c
Weighted average of the intensity of moderate and vigorous physical activity over the past year, in metabolic equivalents (30).
d
Weight (kg)/height (m)
2
.
Neighborhood Attributes and Subclinical Atherosclerosis 295
Am J Epidemiol 2010;172:288–298
at University of Michigan on November 29, 2010aje.oxfordjournals.orgDownloaded from
unemployment rate as the measure of neighborhood SEP.
Empirical evidence also suggests that dietary factors, physi-
cal activity, and smoking behaviors are more responsive to
neighborhood socioeconomic environments in women than
in men (39, 40). These gender differences could result from
gender differences in health-related behavioral responses to
neighborhood perceptions (e.g., varying perceptions of
crime/physical safety contributing to differential levels of
physical activity). Furthermore, the gender discrepancies
could be due to differences in the degree and type of neigh-
borhood exposures, which in turn may be shaped by domestic
and work-related gender roles. For example, women may be
less likely to be employed full-time and may plausibly spend
greater proportions of time in the neighborhood due to child
care and domestic chores. Conversely, men may be more likely
to work full-time and to be exposedto psychosocial stressors in
the workplace, some of which may be linked to CHD inci-
dence (e.g., job strain) (41, 42). In a supplementary explor-
atory analysis, employment status modified the estimated
effects ofneighborhood deprivation in women (in the full-time
stratum, highest-quartile OR ¼1.12, 95% CI: 0.44, 2.82; in the
non-full-time stratum, highest-quartile OR ¼7.15, 95% CI:
1.49, 34.4 (Pfor interaction ¼0.02)). No effect modification
was seen in men, although the sample size in the non-full-time
stratum was limited (n¼161) (data not shown).
Our study suggested only partial mediation of neighbor-
hood effects by risk factors, at best. There were suggestions
of modest mediation of neighborhood deprivation effects by
behavioral factors (i.e., smoking, physical activity) and of
social cohesion effects by depression. Evidence of media-
tion by biologic factors was also limited. While previous
studies have similarly found limited evidence of mediation
by modifiable risk factors (3), our ability to examine medi-
ation may have been compromised by measurement error,
the timing of measures, and the particular potential media-
tors considered. In addition, while we hypothesized these
sets of risk factors primarily as potential mediators, our
study design could not allow us to distinguish mediation
from confounding. A fuller account of mediating pathways
would require data different from those available to us (in-
cluding longitudinal assessments of mediators and assess-
ments of confounders of the mediator-CAC relations).
Lower neighborhood cohesion predicted higher CAC
prevalence among men in poorer (but not richer) neighbor-
hoods. Plausibly, resources available in richer neighborhoods
(e.g., abundant green spaces for leisure) may buffer the ad-
verse effects of low cohesion on CAC. The similar associa-
tions observed for the 3 lowest quartiles versus the highest
quartile might reflect a threshold effect. We lack a clear ex-
planation for why this effect modification was observed in
men only, and replication in other studies is needed. We did
not find that high family income versus low family income
had a similar modifying effect on the associations for low
cohesion in men (data not shown), suggesting that individual
income was not driving the observed interaction.
Previous studies, including the Whitehall studies, have
found graded associations between individual SEP and
CHD which persist after controlling for behavioral and bi-
ologic risk factors (43). The persistence of this gradient
across places and time periods suggesting multiple pathways
to disease has characterized individual SEP as a ‘‘fundamental
cause’’ of healthand disease (44). To the extent that neighbor-
hood SEP and cohesion may mobilize and shape more prox-
imal specific neighborhood dimensions (such as access to
healthy foods and recreation (45, 46) or neighborhood sour-
ces of stress) over time, which may in turn shape individual-
level risk factors, neighborhood SEP and cohesion may be
considered contextual ‘‘fundamental causes’’ of health (44,
47). Therefore, explaining the associations between these key
neighborhood attributes and disease according to selected
risk factors may be particularly challenging. Nevertheless,
in future work, investigators should attempt to carefully elu-
cidate the specific pathways through which these distal
causes may operate in order to identify promising interven-
tions for CHD prevention. Furthermore, corresponding to
these factors as ‘‘fundamental causes,’’ more fundamental
approaches (e.g., mixed-income housing initiatives) to reduce
gaps in neighborhood SEP and social capital should be ex-
plored for their potential to reduce CHD inequalities.
Our study had several limitations. First, participants’ res-
idential addresses were ascertained in 1995 and characterized
on the basis of 2000 US Census data. Some study participants
had moved into other neighborhoods by the time of CAC
assessment in 2005. If the relevant exposure time frame for
the development of CAC is between neighborhood assess-
ment and CAC assessment, misclassification related to resi-
dential mobility subsequent to the 1995 assessment could
have led to underestimation of the true associations. On the
basis of information corresponding to residence in 1995 and
2000, we did in fact observe stronger associations for neigh-
borhood deprivation among women who did not move. How-
ever, if neighborhood of residence assessed in 1995 is
a reliable proxy for prior (life-course) exposures related to
CAC development, subsequent residential mobility may not
have introduced substantial bias. Second, despite the diverse
characteristics of the cohort at inception, both nonresponse
and cohort attrition may have limited the generalizability of
the findings to younger-to-middle-aged US adult urban pop-
ulations. Third, because of insufficient numbers of partici-
pants per neighborhood, perceived neighborhood cohesion
was modeled at the individual level, not the neighborhood
level. Unadjusted individual-level characteristics (such as af-
fective states) may have influenced such perceptions while
also determining CAC, resulting in residual confounding.
In summary, this study offers novel evidence on the asso-
ciations of neighborhood deprivation and low cohesion with
CAC in younger, asymptomatic adults. The associations ap-
pear to be relatively uniformly present in women, whereas in
men the adverse effects of low cohesion seem confined to
those living in deprived neighborhoods. Future investigations
should build on these findings, including gender differences
and mediating pathways, to better elucidate the contextual
and individual-level determinants of CHD and thereby opti-
mize the design of effective prevention strategies.
ACKNOWLEDGMENTS
Author affiliations: Department of Society, Human Devel-
opment, and Health, Harvard School of Public Health, Boston,
296 Kim et al.
Am J Epidemiol 2010;172:288–298
at University of Michigan on November 29, 2010aje.oxfordjournals.orgDownloaded from
Massachusetts (Daniel Kim, Ichiro Kawachi); Department of
Epidemiology, School of Public Health, University of
Michigan, Ann Arbor, Michigan (Ana V. Diez Roux); De-
partment of Quantitative Health Sciences, University of
Massachusetts Medical School, Worcester, Massachusetts
(Catarina I. Kiefe); and Department of Preventive Medicine,
Feinberg School of Medicine, Northwestern University, Evan-
ston, Illinois (Kiang Liu).
This work was supported by the John D. and Catherine T.
MacArthur Foundation Research Network on Socioeco-
nomic Status and Health. The CARDIA Study was funded
by National Heart, Lung, and Blood Institute contracts
N01-HC-05187, N01-HC-45134, N01-HC-48047, N01-
HC-48048, N01-HC-48049, N01-HC-48050, and N01-HC-
95095. Dr. Daniel Kim is supported by a Pathway to
Independence Award through the National Heart, Lung,
and Blood Institute (grant K99HL089459).
This work was presented in part at the Jeremiah and Rose
Stamler Research Award for New Investigators finalists’ ses-
sion of the American Heart Association’s 50th Cardiovascular
Disease Epidemiology and Prevention Annual Conference in
association with the Council on Nutrition, Physical Activity
and Metabolism, San Francisco, California, March 2–5,2010.
Conflict of interest: none declared.
REFERENCES
1. Smith GD, Hart C, Watt G, et al. Individual social class,
area-based deprivation, cardiovascular disease risk factors, and
mortality: the Renfrew and Paisley Study. J Epidemiol
Community Health. 1998;52(6):399–405.
2. Waitzman NJ, Smith KR. Phantom of the area: poverty-area
residence and mortality in the United States. Am J Public
Health. 1998;88(6):973–976.
3. Diez Roux AV, Merkin SS, Arnett D, et al. Neighborhood of
residence and incidence of coronary heart disease. N Engl J
Med. 2001;345(2):99–106.
4. Kawachi I, Berkman LF. Introduction. In: Kawachi I,
Berkman LF, eds. Neighborhoods and Health. New York, NY:
Oxford University Press; 2003:1–19.
5. Sundquist K, Malmstro
¨m M, Johansson SE. Neighbourhood
deprivation and incidence of coronary heart disease: a multi-
level study of 2.6 million women and men in Sweden. J Epi-
demiol Community Health. 2004;58(1):71–77.
6. Chaix B, Rosvall M, Merlo J. Neighborhood socioeconomic
deprivation and residential instability: effects on incidence of
ischemic heart disease and survival after myocardial infarc-
tion. Epidemiology. 2007;18(1):104–111.
7. Diez Roux AV. Residential environments and cardiovascular
risk. J Urban Health. 2003;80(4):569–589.
8. Macintyre S, Ellaway A. Methodological and conceptual
approaches to studying neighborhood effects on health. In:
Kawachi I, Berkman LF, eds. Neighborhoods and Health.New
York, NY: Oxford University Press; 2003:20–44.
9. Sooman A, Macintyre S, Anderson A. Scotland’s health—a
more difficult challenge for some? The price and availability
of healthy foods in socially contrasting localities in the West of
Scotland. Health Bull (Edinb). 1993;51(5):276–284.
10. Morland K, Wing S, Diez Roux A, et al. Neighborhood char-
acteristics associated with the location of food stores and food
service places. Am J Prev Med. 2002;22(1):23–29.
11. Morland K, Diez Roux AV, Wing S. Supermarkets, other food
stores, and obesity: the Atherosclerosis Risk in Communities
Study. Am J Prev Med. 2006;30(4):333–339.
12. Lee RE, Cubbin C, Winkleby M. Contribution of neighbour-
hood socioeconomic status and physical activity resources to
physical activity among women. J Epidemiol Community
Health. 2007;61(10):882–890.
13. Kawachi I, Berkman LF. Social cohesion, social capital,
and health. In: Berkman LF, Kawachi I, eds. Social Epidemi-
ology. New York, NY: Oxford University Press; 2000:
174–190.
14. Kim D, Subramanian SV, Kawachi I. Bonding versus bridging
social capital and their associations with self rated health:
a multilevel analysis of 40 US communities. J Epidemiol
Community Health. 2006;60(2):116–122.
15. Kawachi I, Kennedy BP, Lochner K, et al. Social capital,
income inequality, and mortality. Am J Public Health. 1997;
87(9):1491–1498.
16. Kim D, Subramanian SV, Gortmaker SL, et al. US state- and
county-level social capital in relation to obesity and physical
inactivity: a multilevel, multivariable analysis. Soc Sci Med.
2006;63(4):1045–1059.
17. Lindstrom M. Social capital and health-related behaviors. In:
Kawachi I, Subramanian SV, Kim D, eds. Social Capital and
Health. New York, NY: Springer Publishing Company; 2007:
215–238.
18. Gary TL, Stark SA, LaVeist TA. Neighborhood characteristics
and mental health among African Americans and whites living
in a racially integrated urban community. Health Place.
2007;13(2):569–575.
19. Sundquist J, Johansson SE, Yang M, et al. Low linking social
capital as a predictor of coronary heart disease in Sweden:
a cohort study of 2.8 million people. Soc Sci Med. 2006;62(4):
954–963.
20. Sangiorgi G, Rumberger JA, Severson A, et al. Arterial
calcification and not lumen stenosis is highly correlated
with atherosclerotic plaque burden in humans: a histologic
study of 723 coronary artery segments using nondecalcifying
methodology. J Am Coll Cardiol. 1998;31(1):126–133.
21. Pletcher MJ, Tice JA, Pignone M. Use of coronary calcification
scores to predict coronary heart disease. JAMA. 2004;291(15):
210–215.
22. Arad Y, Spadaro LA, Goodman K, et al. Predictive value of
electron beam computed tomography of the coronary arteries:
19-month follow-up of 1173 asymptomatic subjects. Circula-
tion. 1996;93(11):1951–1953.
23. Winkleby M, Sundquist K, Cubbin C. Inequities in CHD
incidence and case fatality by neighborhood deprivation. Am
J Prev Med. 2007;32(2):97–106.
24. Franzini L, Spears W. Contributions of social context to in-
equalities in years of life lost to heart disease in Texas, USA.
Soc Sci Med. 2003;57(10):1847–1861.
25. Skogan WG. Disorder and Decline: Crime and the Spiral of
Decay in American Neighborhoods. Berkeley, CA: University
of California Press; 1990.
26. Friedman GD, Cutter GR, Donahue RP, et al. CARDIA:
study design, recruitment, and some characteristics of the
examined subjects. J Clin Epidemiol. 1988;41(11):1105–
1016.
27. Barnes J, Sampson RJ, Kindlon D, et al. A community
survey approach to ecological assessment: results from a pilot
survey. In: Earls F, Buka SL, eds. Project on Human Devel-
opment in Chicago Neighborhoods. (Report no. NCJ 163495).
Washington, DC: National Institute of Justice, US Department
of Justice; 1997:34–45.
Neighborhood Attributes and Subclinical Atherosclerosis 297
Am J Epidemiol 2010;172:288–298
at University of Michigan on November 29, 2010aje.oxfordjournals.orgDownloaded from
28. Echeverria SE, Diez-Roux AV, Link BG. Reliability of self-
reported neighborhood characteristics. J Urban Health.
2004;81(4):682–701.
29. Radloff LS. The CES-D Scale: a self-report depression scale
for research in the general population. Appl Psychol Meas.
1977;1(3):385–401.
30. Jacobs DR, Hahn LP, Haskell WL, et al. Validity and reliability
of short physical activity history: CARDIA and the Minnesota
Heart Health Program. J Cardiopulm Rehabil. 1989;9:
448–459.
31. Loria CM, Liu K, Lewis CE, et al. Early adult risk factor levels
and subsequent coronary artery calcification: the CARDIA
Study. J Am Coll Cardiol. 2007;49(20):2013–2020.
32. Rubin DB. Multiple Imputation for Nonresponse in Surveys.
New York, NY: John Wiley & Sons, Inc; 1987.
33. Nordstrom CK, Diez Roux AV, Jackson SA, et al. The asso-
ciation of personal and neighborhood socioeconomic indica-
tors with subclinical cardiovascular disease in an elderly
cohort. The Cardiovascular Health Study. Soc Sci Med.
2004;59(10):2139–2147.
34. Lemelin ET, Diez Roux AV, Franklin TG, et al. Life-course
socioeconomic positions and subclinical atherosclerosis in the
Multi-Ethnic Study of Atherosclerosis. Soc Sci Med. 2009;
68(3):444–451.
35. Dragano N, Hoffmann B, Stang A, et al. Subclinical coronary
atherosclerosis and neighbourhood deprivation in an urban
region. Eur J Epidemiol. 2009;24(1):25–35.
36. Carson AP, Rose KM, Catellier DJ, et al. Cumulative
socioeconomic status across the life course and subclinical
atherosclerosis. Ann Epidemiol. 2007;17(4):296–303.
37. Kim D, Subramanian SV, Kawachi I. Social capital and
physical health: a systematic review of the literature. In:
Kawachi I, Subramanian SV, Kim D, eds. Social Capital and
Health. New York, NY: Springer Publishing Company;
2007:139–190.
38. Chaix B, Lindstrom M, Rosvall M, et al. Neighborhood social
interactions and risk of acute myocardial infarction. J Epide-
miol Community Health. 2008;62(3):62–68.
39. Wang MC, Kim S, Gonzalez AA, et al. Socioeconomic and
food-related physical characteristics of the neighbourhood
environment are associated with body mass index. J Epidemiol
Community Health. 2007;61(6):491–498.
40. Cubbin C, Hadden WC, Winkleby MA. Neighborhood context
and cardiovascular disease risk factors: the contribution of
material deprivation. Ethn Dis. 2001;11(4):687–700.
41. Hammar N, Alfredsson L, Johnson JV. Job strain, social
support at work, and incidence of myocardial infarction.
Occup Environ Med. 1998;55(8):548–553.
42. Kuper H, Marmot M. Job strain, job demands, decision lati-
tude, and risk of coronary heart disease within the Whitehall II
Study. J Epidemiol Community Health. 2003;57(2):147–153.
43. Marmot MG, Rose G, Shipley M, et al. Employment grade and
coronary heart disease in British civil servants. J Epidemiol
Community Health. 1978;32(4):244–249.
44. Link BG, Phelan J. Social conditions as fundamental causes of
disease. J Health Soc Behav. 1995;(spec no.):80–94.
45. Moore LV, Diez Roux AV. Associations of neighborhood
characteristics with the location and type of food stores. Am J
Public Health. 2006;96(2):325–331.
46. Lee RE, Booth KM, Reese-Smith JY, et al. The Physical
Activity Resource Assessment (PARA) instrument: evaluating
features, amenities and incivilities of physical activity
resources in urban neighborhoods. Int J Behav Nutr Phys Act.
2005;2:13. (doi: 10.1186/1479-5868-2-13).
47. Kim D. Blues from the neighborhood? Neighborhood char-
acteristics and depression. Epidemiol Rev. 2008;30:101–117.
298 Kim et al.
Am J Epidemiol 2010;172:288–298
at University of Michigan on November 29, 2010aje.oxfordjournals.orgDownloaded from
