Objectives. To strengthen existing evidence on the role of neighborhoods in chronic disease onset in later life, we investigated associations between multiple neighborhood features and 2-year onset of 6 common conditions using a national sample of older adults.

Methods. Neighborhood features for adults aged 55 years or older in the 2002 Health and Retirement Study were measured by use of previously validated scales reflecting the built, social, and economic environment. Two-level random-intercept logistic models predicting the onset of heart problems, hypertension, stroke, diabetes, cancer, and arthritis by 2004 were estimated.

Results. In adjusted models, living in more economically disadvantaged areas predicted the onset of heart problems for women (odds ratio [OR] = 1.20; P < .05). Living in more highly segregated, higher-crime areas was associated with greater chances of developing cancer for men (OR = 1.31; P < .05) and women (OR = 1.25; P < .05).

Conclusions. The neighborhood economic environment is associated with heart disease onset for women, and neighborhood-level social stressors are associated with cancer onset for men and women. The social and biological mechanisms that underlie these associations require further investigation.

Currently, 8 out of 10 older adults in the United States have at least 1 chronic condition.1 Reports of many common chronic conditions, such as heart disease, arthritis, diabetes, and some cancers, have been increasing, as have the costs associated with their treatment.2,3 Although the etiology of such conditions varies greatly, a rapidly growing literature has documented associations between characteristics of the neighborhoods in which older people live and late-life morbidity.

The most studied neighborhood feature in this context is economic disadvantage. Studies have established that living in economically deprived areas is associated with higher risks of heart disease,411 stroke,1214 hypertension,6,15,16 and a greater number of chronic conditions,17,18 but lower cancer incidence.1921 These effects, which appear to be greater for women than for men,6,911,19,22 often attenuate but are not completely eliminated after individual-level factors are taken into account. Numerous mechanisms have been postulated as underlying the linkage between economic deprivation and chronic conditions. In reviewing cardiovascular disease mechanisms, for example, Diez Roux discussed how social and physical aspects of poor neighborhoods may influence individual risk factors (e.g., physical activity, diet, smoking, and the ability to recover from stress), which in turn may influence more proximate biological risk factors (e.g., blood pressure, diabetes, body mass index, blood lipids, and inflammation).23

Recently, the literature has begun to address noneconomic features of neighborhoods, such as the social and built environments, and their relation to health in later life.16,20,2428 The social environment refers to relations among people living in a particular area and encompasses concepts such as connectedness to and similarity with neighbors and social disorder. The built environment refers to factors related to man-made elements including housing quality, businesses, street design, pollution, and crowding. Like the linkage between poor neighborhoods and cardiovascular disease, the relationship between the social and built environments and later-life morbidity is likely to be complex, operating through physiologic stress as well as health behaviors such as physical activity and diet and access to providers. However, measures of the social and built environment have typically been absent from analyses that include measures of economic disadvantage, thus making it difficult to sort out these influences.

More generally, conclusions that can be drawn from the growing number of studies devoted to neighborhood influences on late-life health remain limited in several respects. First, most studies highlight the relationship between a single neighborhood facet and 1 chronic condition. This approach precludes comparisons across conditions, and also limits interpretation, because aspects of the economic, social, and built environment are likely to be correlated. Second, only about one third of studies to date have examined disease incidence,4,5,10,11,15,20,22,29,30 whereas remaining studies have focused on prevalence. The latter confounds influences of the neighborhood on disease onset with its effects on survival and therefore provides only limited insight into disease etiology. Third, with few exceptions,17,18 studies focusing on the United States have drawn data from a limited number of communities, thus the generalizability of the results is uncertain. Fourth, indicators of individual-level circumstances often have been quite limited; consequently, the effect of neighborhood-level factors may be confounded by unmeasured individual-level factors. Fifth, selection into neighborhoods along health dimensions may bias findings, yet studies to date have not attempted to control for circumstances that may act as a proxy for neighborhood exposures earlier in life. Finally, despite evidence that chronic disease etiology and expression may differ for older men and women,3133 gender-specific investigations have been the exception rather than the norm.

We aimed to enhance this literature by adopting a richer characterization of the neighborhood that draws on previously validated scales reflecting the economic, social, and built neighborhood environment. We explored associations between these scales and the reported onset of 6 of the most commonly reported late-life conditions: hypertension, heart problems, stroke, diabetes, cancer, and arthritis. We hypothesized that all 3 domains—the economic, social, and built environments—would contribute to increased risks of chronic conditions in later life. To explore these hypotheses, we used a large, nationally representative sample of US adults aged 55 years or older from the Health and Retirement Study (HRS).34 The HRS includes excellent contemporaneous measures of income and assets, as well as retrospective measures of health and wealth earlier in life. We therefore were better able than previous studies to isolate the contribution of neighborhood-level socioeconomic components. Moreover, the large sample sizes allowed us to stratify the analyses for men and women.

The HRS collects extensive information on health, demographic, and socioeconomic characteristics from a large sample of persons aged 50 years or older and their spouses at 2-year intervals. Economic data are of particularly high quality.35 We used the 2002 wave, which was the first wave coded to reflect 2000 US Census boundaries and which had an overall response rate of 86.9%. Additional details on sample composition are available elsewhere.36

We restricted our analyses to respondents aged 55 years or older in 2002 who survived and responded to the 2004 wave. After excluding cases that were missing geographic identifiers (about 2%) and values on outcomes and predictors (about 5%), we restricted the analytic sample for each condition of interest to respondents not reporting that particular condition in 2002. Depending on the given condition, the final sample sizes varied from 2731 to 7414 for women and from 2780 to 5363 for men (Table 1).


TABLE 1 Chronic Conditions Among US Adults Aged 55 Years and Older: Health and Retirement Study, 2002–2004

TABLE 1 Chronic Conditions Among US Adults Aged 55 Years and Older: Health and Retirement Study, 2002–2004

Condition% Reporting a Condition in 2002No. of Respondents Not Reporting a Condition in 2002 (No. of Census Tracts)% Reporting a Condition in 2004% Reporting a Condition in 2002No. of Respondents Not Reporting a Condition in 2002 (No. of Census Tracts)% Reporting a Condition in 2004
Heart problems27.14194 (2335)6.321.96265 (3092)5.4
High blood pressure50.22815 (1743)10.252.53680 (2181)12.3
Stroke8.35363 (2722)1.97.77414 (3415)2.1
Diabetes17.94748 (2519)3.515.06727 (3264)3.3
Cancer13.45055 (2601)3.513.66945 (3280)2.4
Arthritis49.82780 (1748)11.264.82731 (1735)15.2

Note. Respondents were asked whether a doctor had ever told them that they had a given condition. In 2002, there were 6580 male respondents aged 55 years and older who survived to 2004 in 3153 tracts; there were 8794 female respondents in 3858 tracts.


Conditions of interest for this analysis included coronary heart disease, angina, congestive heart failure or other heart problems (which we referred to collectively as “heart problems”); high blood pressure or hypertension; stroke; diabetes; cancer or a malignant tumor, excluding minor skin cancers; and arthritis or rheumatism. HRS respondents were asked at their first interview whether a doctor had ever told them that they had a given condition. In subsequent interviews, respondents who had not previously reported a given condition were asked, “Since we last talked to you, has a doctor told you that you have [given condition]?” Although we did not have the rich clinical detail more typical of community-based epidemiologic studies, previous studies of the validity and reliability of self-reported chronic conditions have suggested excellent agreement with claims or medical records for most of the conditions considered here.37,38

Neighborhood Scales

Neighborhoods were characterized by using 8 previously validated scales: 2 reflecting the economic environment (economic advantage and disadvantage), 3 representing the social environment (immigration concentration, crime and segregation, and residential stability), and 3 measuring the built environment (connectivity, air pollution, and density). Details of scale formation are found elsewhere39; we therefore provide only a brief description here.

By use of factor analysis, we identified items that loaded together at 0.40 or higher, which we then standardized and added together. With few exceptions (discussed below), the items scaled into substantively unique and coherent factors. The resulting Cronbach α for these scales fell in the range of 0.89 to 0.96 for tracts represented in the HRS and 0.89 to 0.94 for all US Census tracts, demonstrating high internal validity and robustness. Scales were also found to be at most modestly correlated, with all pairwise correlations falling between −0.46 and 0.43, and most correlations having an absolute value less than 0.30. Each scale was standardized (with mean of zero) for ease of interpretation and comparison across scales; thus, a 1-unit change in a given scale represented a change of 1 standard deviation.

Economic environment.

Two scales reflecting the economic circumstances at the tract level were formed by using the 2000 Census.40 The economic disadvantage scale included the percentage of the total population in poverty; the percentage of the population aged 65 years or older in poverty; the percentage of households receiving public assistance income; the unemployment rate among persons aged 16 years or older; the percentage of housing units without a vehicle; and the percentage of the population that was Black. A scale reflecting economic advantage included the upper quartile of the percentage of owner-occupied housing units in the tract; the percentage of families with a total annual income of $75 000 or more; and the percentage of adults with a college degree.

Social environment.

The immigrant concentration scale included 4 standardized indicators from the 2000 Census: the percentage of the tract that was Hispanic, foreign-born, and with limited English skills; and an index reflecting Hispanic isolation (the probability that members of a given group will meet members of their own group in their Census tract).

Crime and segregation.

County-level measures of crime were drawn from the 2002 Uniform Crime Reporting Program Data41 and included aggravated assaults, burglaries, larcenies, motor vehicle thefts, murders, and robberies, divided by the size of the county population. These crime measures were loaded with 2 measures of segregation calculated from the 2000 Census: an isolation index for Blacks (described in “Social environment”) and a Black–White dissimilarity index (the proportion of Blacks in a tract that would have to move for the tract to have the same racial distribution as the surrounding county). Although substantively distinct, the crime and segregation items shared substantial variance and we found that subscales were correlated above 0.55, so we combined the 2 subscales into a single scale and investigated the effects of each measure individually in sensitivity analyses. Residential stability was reflected by the percentage in 2000 living in the same house since at least 1995 and by the median number of years of residence for the Census tract.

Built environment.

A scale reflecting connectivity of tracts was primarily drawn from the 2000 Topologically Integrated Geographic Encoding and Referencing system.42 The scale included the number of street segments per square mile, the number of nodes per square mile, and 2 connectivity measures, α and γ, for which higher values indicated more connectivity (where α was the ratio of the actual number of complete loops to the maximum number of possible loops given the number of intersections and γ was the ratio of actual street segments to the maximum possible street segments). The average age of the housing units was included to capture differences in street design.43 An air pollution scale was derived from the Environmental Protection Agency's Air Quality System.44 It included a set of quarterly measures of Particulate Matter of 10 μm or less (reflecting both fine and coarse dust particles) and a measure of summertime ozone averages. Finally, density was captured by county-level information on the number of food stores, restaurants, and housing units per square mile from the 2002 Economic Census (US Census Bureau, Geographic Area Series, unpublished data, 2002) and by tract-level population density measures from the 2000 Census.

Health care delivery environment.

In sensitivity analyses, we also explored several measures of the health care delivery environment. Three county-level indicators from the 2003 Area Resource File—total number of physicians, short-term hospital beds, and home health care agencies per 1000 persons in the population—were found not to reach a high enough threshold (greater than 0.40) in the factor analyses to be included in any of the aforementioned scales, nor did they form their own factor.39 However, because the health care delivery system has important theoretical relevance for determining both diagnosis and survival with chronic conditions, we retained these variables individually in the sensitivity analyses.

Individual-Level Predictors

In adjusted models, we included contemporaneous and retrospective individual-level characteristics that we expected to be related to chronic conditions in later life and to characteristics of current neighborhoods (Table 2). Contemporaneous measures included age (in 5-year age groups), race/ethnicity (non-Hispanic White, non-Hispanic Black, other non-Hispanic, and Hispanic), marital status (widowed, divorced or separated, or never married versus married), current region (south, midwest, or west versus northeast), whether the interview was conducted with a proxy respondent, smoking status (current or former versus never), completed education (less than 8 years, 9 to 11 years, high school, or more than high school), total assets of the respondent (and spouse if married), and income in relation to poverty (income-to-needs ratio). Variables reflecting experiences earlier in life included: retrospective reports of childhood health (fair or poor, good, or very good versus excellent), socioeconomic status (poor or varied versus about average or better), and region of birth (foreign-born, south, midwest, or west versus northeast).


TABLE 2 Individual-Level Characteristics of US Adults Aged 55 Years and Older: Health and Retirement Study, 2002

TABLE 2 Individual-Level Characteristics of US Adults Aged 55 Years and Older: Health and Retirement Study, 2002

Men (n = 6580), %Women (n = 8794), %
Age, y
    ≥ 854.56.9
    Non-Hispanic White84.082.7
    Non-Hispanic Black8.29.5
    Other non-Hispanic2.32.1
    Hispanic origin5.55.7
Marital status
    Divorced or separated9.213.7
    Never married3.13.3
Current region
Proxy response14.05.0
Smoking status
    Currently smokes15.213.0
    Formerly smoked71.349.7
    Never smoked13.537.3
Education completed, y
    ≤ 811.510.2
    ≥ 1345.837.7
Mean assets in $100 0004.33.4
Income categorya
    < 100% of poverty line5.510.1
    < 130% of poverty line3.46.1
    130% to < 185% of poverty line8.211.2
    185% to < 300% of poverty line17.920.9
    ≥ 300% of poverty line65.051.7
Health during childhood
    Very good25.325.2
    Fair or poor5.16.7
Childhood socioeconomic status
    Average or better66.469.0
    Poor or varied33.330.8
Region of birth

aPoverty level constructed using US Census Bureau methodology.

Statistical Methods

We initially entered each neighborhood scale individually into a series of unadjusted logistic models predicting the onset of specific conditions between 2002 and 2004 among those without a given condition in 2002. We then estimated a series of 2-level random intercept logistic models first with no predictors, then with individual- but no neighborhood-level predictors, and finally with both individual- and neighborhood-level characteristics. For each model we calculated a pseudo-intraclass correlation coefficient (ICC), which expresses the percentage of variability in the outcome attributable to between-neighborhood variation.45,46 By comparing pseudo-ICCs across models, we quantified the extent of the variance attributable to neighborhood characteristics before and after controlling for individual- and neighborhood-level factors. Finally, in a series of sensitivity analyses, we added measures of the health care delivery system to the models. All models were stratified by gender and estimated in Stata version 9.0 (Stata Corp LP, College Station, TX) by using the robust cluster feature, which accounts for clustering of respondents within tracts.

In the unadjusted logistic regression models (Table 3), only 3 scales emerged as being significantly related to the onset of chronic conditions for men: living in a more economically advantaged area was associated with a lower odds of experiencing the onset of diabetes [odds ratio (OR) = 0.80; 95% confidence interval (CI) = 0.69, 0.94], living in areas with a higher immigrant concentration was associated with an increased odds of developing high blood pressure (OR = 1.16; 95% CI = 1.04, 1.30), and living in an area with higher levels of crime and more segregation was associated with a higher odds of developing cancer (OR = 1.18; 95% CI = 1.03, 1.37). The 95% CIs for the 3 built environment scales (street connectivity, air pollution, and density) crossed 1.0.


TABLE 3 Odds Ratios (OR) of 2-Year Onset of 6 Chronic Conditions Among US Adults Aged 55 Years and Older: Health and Retirement Study, 2002–2004

TABLE 3 Odds Ratios (OR) of 2-Year Onset of 6 Chronic Conditions Among US Adults Aged 55 Years and Older: Health and Retirement Study, 2002–2004

Heart Problems, OR (95% CI)High Blood Pressure, OR (95% CI)Stroke, OR (95% CI)Diabetes, OR (95% CI)Cancer, OR (95% CI)Arthritis, OR (95% CI)
Economic disadvantage0.98 (0.87, 1.10)1.08 (0.95, 1.23)1.02 0.85, 1.22)1.10 (0.95, 1.28)1.00 (0.84, 1.19)1.03 (0.90, 1.18)
Economic advantage0.93 (0.81, 1.06)0.98 (0.87, 1.11)0.92 (0.76, 1.11)0.80** (0.69, 0.94)1.09 (0.96, 1.24)0.96 (0.83, 1.10)
High immigrant area0.97 (0.85, 1.11)1.16** (1.04, 1.30)1.10 (0.90, 1.36)1.04 (0.89, 1.21)0.91 (0.76, 1.08)0.88 (0.76, 1.02)
High crime and more segregation0.94 (0.83, 1.06)1.11 (0.99, 1.25)1.04 (0.85, 1.28)0.98 (0.84, 1.13)1.18* (1.03, 1.37)0.94 (0.82, 1.07)
Residential stability1.07 (0.95, 1.20)1.03 (0.92, 1.15)1.05 (0.87, 1.28)1.02 (0.89, 1.16)0.99 (0.84, 1.16)1.01 (0.90, 1.13)
Connectivity1.08 (0.95, 1.21)0.97 (0.86, 1.09)0.99 (0.82, 1.19)1.04 (0.91, 1.20)1.00 (0.85, 1.17)1.03 (0.92, 1.16)
Air pollution0.98 (0.87, 1.11)1.11 (0.99, 1.24)1.07 (0.91, 1.24)0.97 (0.84, 1.12)0.95 (0.81, 1.11)1.01 (0.91, 1.13)
Density1.04 (0.94, 1.16)0.87 (0.71, 1.08)1.07 (0.92, 1.24)1.01 (0.88, 1.15)1.04 (0.93, 1.17)0.95 (0.84, 1.07)
Economic disadvantage1.15** (1.04, 1.28)1.11* (1.01, 1.22)1.11 (0.97, 1.27)1.25** (1.12, 1.39)0.93 (0.78, 1.11)1.05 (0.94, 1.16)
Economic advantage0.88* (0.78, 0.99)0.88* (0.80, 0.97)0.86 (0.71, 1.03)0.71** (0.60, 0.83)1.01 (0.87, 1.18)0.99 (0.90, 1.09)
High immigrant area1.02 (0.90, 1.16)0.99 (0.88, 1.10)1.07 (0.91, 1.26)1.06 (0.91, 1.23)0.83 (0.68, 1.00)0.97 (0.86, 1.09)
High crime and more segregation1.06 (0.95, 1.19)1.06 (0.96, 1.18)1.16 (1.00, 1.35)1.10 (0.96, 1.26)1.03 (0.89, 1.19)1.02 (0.92, 1.13)
Residential stability0.98 (0.88, 1.10)0.98 (0.89, 1.08)0.97 (0.84, 1.11)1.13* (1.00, 1.28)1.04 (0.89, 1.20)1.09 (0.98, 1.21)
Connectivity1.06 (0.94, 1.19)0.97 (0.88, 1.07)1.12 (0.96, 1.30)1.03 (0.90, 1.17)0.95 (0.81, 1.11)0.97 (0.87, 1.08)
Air pollution1.00 (0.90, 1.11)0.96 (0.87, 1.06)0.98 (0.82, 1.16)0.98 (0.85, 1.15)0.89 (0.75, 1.06)1.09 (0.99, 1.20)
Density1.01 (0.93, 1.09)0.95 (0.85, 1.05)1.09 (0.99, 1.19)1.00 (0.89, 1.13)0.96 (0.86, 1.08)1.05 (0.97, 1.14)

Note. CI = confidence interval. Models were unadjusted logistic regression models.

*P < .05; **P < .01.

For women, living in an economically disadvantaged area was associated with higher odds of developing heart problems (OR = 1.15; 95% CI = 1.04,1.28), high blood pressure (OR = 1.11; 95% CI = 1.01, 1.22), and diabetes (OR = 1.25; 95% CI = 1.12, 1.39), whereas living in an economically advantaged area was associated with reduced odds of developing these 3 conditions (ORs were 0.88 or lower and 95% CIs did not cross 1.00). In addition, residential stability was associated with an increased odds of diabetes onset (OR = 1.13; 95% CI = 1.00, 1.28; P = .048). Two neighborhood scales reached borderline statistical significance (just over P = .05): living in a highly segregated, high-crime area was associated with an increased odds of having a stroke (OR = 1.16; 95% CI = 1.00, 1.35; P = .052) and living in an area with a high immigrant concentration was inversely associated with the odds of developing cancer (OR = 0.83; 95% CI = 0.68, 1.00; P = .051).

Adjusted Models

When all neighborhood factors were included in 2-level models with individual-level controls, neighborhood features remained significantly associated with the onset of only 2 conditions: heart problems for women and cancer for men and women (Table 4). The OR for the economic disadvantage scale increased in size (OR = 1.20) and remained statistically significant at P = 0.05 in predicting heart problems for women. Highly segregated, high-crime areas appeared to put both men and women at risk for developing cancer (OR = 1.31 for men and OR = 1.25 for women).


TABLE 4 Adjusted Odds Ratios (AOR) of 2-Year Onset of 6 Chronic Conditions Among US Adults Aged 55 Years and Older: Health and Retirement Study, 2002–2004

TABLE 4 Adjusted Odds Ratios (AOR) of 2-Year Onset of 6 Chronic Conditions Among US Adults Aged 55 Years and Older: Health and Retirement Study, 2002–2004

Neighborhood ScaleHeart Problems, AOR (95% CI)High Blood Pressure, AOR (95% CI)Stroke, AOR (95% CI)Diabetes, AOR (95% CI)Cancer, AOR (95% CI)Arthritis, AOR (95% CI)
Economic disadvantage0.98 (0.79, 1.22)1.06 (0.85, 1.32)0.81 (0.57, 1.14)0.97 (0.75, 1.25)1.17 (0.91, 1.50)1.08 (0.88, 1.33)
Economic advantage0.94 (0.80, 1.12)1.11 (0.94, 1.30)0.92 (0.70, 1.21)0.88 (0.71, 1.11)1.04 (0.86, 1.26)1.05 (0.90, 1.22)
High immigrant area1.00 (0.82, 1.22)1.15 (0.95, 1.39)1.11 (0.83, 1.49)1.01 (0.80, 1.28)0.77 (0.59, 1.01)0.90 (0.73, 1.09)
High crime and more segregation1.01 (0.87, 1.17)1.15 (0.99, 1.33)1.05 (0.84, 1.32)0.97 (0.81, 1.16)1.31** (1.10, 1.56)0.91 (0.79, 1.05)
Residential stability0.97(0.85, 1.12)1.07 (0.93, 1.23)1.02 (0.82, 1.27)0.97 (0.82, 1.15)1.01 (0.86, 1.20)0.98 (0.86, 1.12)
Connectivity1.05 (0.89, 1.23)0.89 (0.75, 1.06)0.89 (0.68, 1.17)1.06 (0.86, 1.29)0.94 (0.77, 1.15)1.14 (0.97, 1.34)
Air pollution1.06 (0.90, 1.24)1.02 (0.89, 1.18)1.04 (0.81, 1.34)0.96 (0.80, 1.16)0.93 (0.78, 1.11)1.09 (0.94, 1.26)
Density1.03 (0.91, 1.16)0.76 (0.55, 1.05)1.06 (0.89, 1.26)1.05 (0.89, 1.24)1.04 (0.88, 1.21)0.94 (0.79, 1.13)
Economic disadvantage1.20* (1.00, 1.43)1.13 (0.95, 1.34)0.78 (0.60, 1.02)0.99 (0.80, 1.22)1.16 (0.89, 1.52)1.04 (0.86, 1.25)
Economic advantage0.95 (0.81, 1.11)0.92 (0.81, 1.05)0.85 (0.66, 1.08)0.86 (0.69, 1.07)0.92 (0.74, 1.15)0.99 (0.87, 1.13)
High immigrant area1.04 (0.87, 1.24)0.94 (0.80, 1.10)1.08 (0.85, 1.37)0.82 (0.66, 1.02)0.89 (0.66, 1.19)0.93 (0.78, 1.12)
High crime and more segregation1.06 (0.93, 1.22)1.05 (0.94, 1.18)1.16 (0.96, 1.40)1.07 (0.91, 1.26)1.25* (1.04, 1.52)1.07 (0.94, 1.21)
Residential stability0.99 (0.87, 1.12)0.99 (0.89, 1.10)0.95 (0.80, 1.13)1.15 (0.99, 1.33)1.09 (0.91, 1.31)1.10 (0.97, 1.23)
Connectivity0.90 (0.77, 1.05)0.92 (0.81, 1.06)1.02 (0.83, 1.25)1.01 (0.84, 1.20)0.82 (0.66, 1.01)0.90 (0.78, 1.04)
Air pollution0.97 (0.84, 1.12)0.98 (0.87, 1.10)0.93 (0.77, 1.13)0.88 (0.74, 1.04)0.85 (0.70, 1.03)1.10 (0.97, 1.25)
Density0.95 (0.83, 1.09)0.97 (0.84, 1.12)1.10 (0.96, 1.27)0.99 (0.83, 1.17)1.00 (0.82, 1.23)1.05 (0.95, 1.16)

Note. CI = confidence interval. Multilevel logistic regression models were adjusted for age, race/ethnicity, marital status, region of residence, smoking status, years of education completed, mean assets, income category, childhood health, childhood socioeconomic status, region of birth, and all neighborhood scales.

*P < .05; **P < .01.

Variance Attributable to Neighborhoods

Pseudo-ICCs (not shown) for models with no predictors were less than 5% for 8 of the 12 models. For men, ICCs reached 8% for heart problems and cancer and 16% for stroke. For women, ICCs were above 5% only for heart problems (ICC = 6%). Controlling for individual-level factors reduced the ICCs to less than 0.01% for heart problems for men and to less than 5% for heart problems for women, increased the ICCs for stroke for women from 4% to 8%, but in other cases had negligible effects. Adding neighborhood scales to models with ICCs at least 5% (after individual controls were introduced) further reduced the ICCs by 0.5 to 0.8 percentage points.

Sensitivity Analyses

Model results were highly robust to the addition of indicators of the health care environment. In almost all cases, the number of physicians, short-stay beds, and home health agencies per county population did not emerge as significant predictors. An exception was that, for men, the number of physicians was inversely associated at P = 0.01 with the risk of developing heart disease (OR = 0.83; 95% CI = 0.70, 0.95). Adding health care indicators did not eliminate findings previously reported as being statistically significant.

Using a national sample of older adults, we investigated associations between the economic, social, and built environment and self-reported onset of 6 common chronic conditions. We found associations between the onset of selected conditions (heart disease, cancer) and aspects of the economic and social environment in which older adults live, but no linkage with measures of the built environment that we included (connectivity, air pollution, and density).

We found that women living in economically disadvantaged areas in the United States have an increased risk of incident heart disease. Previous epidemiologic studies of coronary heart disease in community-based samples yielded strikingly similar results6; however, ours was the first study to demonstrate the robustness of this relation in national US data after control for indicators of the social and built environment and extensive individual-level characteristics. For other conditions such as hypertension and diabetes, we showed a lack of persistence in the association between neighborhood-level economic status and the onset of disease once differences in other neighborhood and individual-level factors were taken into account.

Our study was also the first to suggest that living in more highly segregated areas with higher crime rates is associated with an increased risk of developing cancer. The literature on cancer and the environment has emphasized lifestyle factors such as tobacco use, diet and exercise, and exposure to cancer-causing agents,47 but attention to socioeconomic aspects of the environment has been limited. At the same time, others have argued that racial segregation is a fundamental cause of disparities between Blacks and Whites in health and mortality48,49 and that Black segregation and crime are closely related.50 However, we could find no previous study in which a link to cancer was identified.

Although we drew upon national survey data, which offer the advantages of generalizability and relatively large sample sizes, our study had some limitations. Disease assessments were limited to self-reports; we therefore could not determine the clinical nature or severity of conditions, and undiagnosed cases may be differentially misreported. In addition, neighborhood boundaries were limited to US Census tracts and counties, which may not provide the most relevant construct of neighborhood. However, Krieger et al.21 showed that tracts were valuable in predicting mortality and cancer incidence and preferable to other constructs such as zip code. We were also unable to explore the health care system in detail. In particular, measures of health care quality may be necessary to sort out why disease rates vary across areas.

We also acknowledge that because residents may choose the neighborhoods in which they live on the basis of health-related characteristics,51 the associations we find may not necessarily be causal.52 However, unlike prior studies, we included retrospective indicators of childhood health and wealth, which (albeit limited) were intended to control to some extent for the confounding effects of neighborhood selection. Moreover, other studies that used the HRS found no relation between self-assessed health and the propensity to move.53

Despite such limitations, this exploratory study points to potentially new pathways through which the neighborhood environment may influence chronic disease etiology. To date, the most common explanation for the link between Black segregation and mortality has been that segregation influences socioeconomic deprivation and individual socioeconomic attainment.48 We found, however, that a scale composed of highly correlated indicators of Black segregation and crime was predictive of the onset of cancer for both men and women after controlling for neighborhood and individual-level socioeconomic resources. Although we controlled for certain aspects of the built environment, including air pollution levels, we could not rule out that this finding may reflect greater exposure to other environmental toxins (e.g., hazardous waste, poor water quality) in areas with higher concentrations of African Americans.54 However, in additional analyses (not shown), we disaggregated the index and found that, when included individually, each crime and segregation measure predicted cancer onset. Moreover, we found that the ORs for this scale were not significantly different for Blacks and Whites. The remarkable similarity in coefficient size and strength for men and women was surprising given the heterogeneity in cancer cites by gender and suggests that a nonspecific biological mechanism may be involved. For instance, a stress response that interrupts the body's ability to fight cancer cell development might be at work.

With respect to heart disease, a more complex set of potential pathways between the neighborhood environment and the onset of cardiovascular disease has been proposed.23 We cannot pinpoint the biological mechanisms in our study. However, we note that for 3 markers—self-reported blood pressure and diabetes in our study and, in previously published work,55 obesity—the link with neighborhood economic disadvantage was unique to women and that these relations dissipated in adjusted models. In contrast, the relation between economic disadvantage and heart problems for women was strengthened in the adjusted models. These findings hint that perhaps another biological mechanism that is unique, or at least more pronounced, among women may be involved. Moreover, although both physical and social aspects of the environment have been proposed as pathways through which the economic environment influences cardiovascular health, we did not find evidence to support an intervening role of density, street connectivity, availability of food stores, pollution, safety and violence, or social cohesion.

To further elucidate the mechanisms underlying these associations, studies linking public health outcomes with biological indicators are needed. National studies linking specific attributes of neighborhoods with cardiovascular disease would benefit from the introduction of cardiovascular disease biomarkers such as fractionated cholesterol, measured blood pressure, body mass index calculated from measured height and weight, glycated hemoglobin, and markers of inflammation. Similarly, studies of neighborhood influences on cancer onset would benefit from the introduction of biological measures of stress (e.g., cortisol and other reflections of hypothalamic-pituitary-adrenal axis activity) with more detailed clinical diagnoses. Advances in biosocial surveys and the ability to link clinical studies to rich geocoded data will soon make these public health investigations possible.


This work was supported by the National Institutes of Health through the National Institute on Aging (grant R01 AG024058), and by the National Institute of Environmental Health Sciences (grant P50 ES12383 to RAND).

An earlier version of this paper was presented at the annual meeting of the Gerontological Association of America; November 16–20, 2007; San Francisco, CA.

We thank Rizie Kumar for her programming assistance.

Note. The views expressed here are those of the authors alone and do not represent the views of their employers or the funding agency.

Human Participant Protection

This study was approved by the institutional review board of the University of Medicine and Dentistry of New Jersey.


1. Centers for Disease Control and Prevention. Public health and aging: trends in aging—United States and worldwide. MMWR Morb Mortal Wkly Rep. 2003;52(6):101106. Google Scholar
2. Freedman VA, Schoeni RF, Martin LG, Cornman JC. Chronic conditions and the decline in late-life disability. Demography. 2007;44(3):459477. Crossref, MedlineGoogle Scholar
3. Thorpe KE. Factors accounting for the rise in health-care spending in the United States: the role of rising disease prevalence and treatment intensity. Public Health. 2006;120(11):10021007. Crossref, MedlineGoogle Scholar
4. Chaix B, Isacsson SO, Rastam L, Lindstrom M, Merlo J. Income change at retirement, neighbourhood-based social support, and ischemic heart disease: results from the prospective cohort study “Men born in 1914”. Soc Sci Med. 2007;64(4):818829. Crossref, MedlineGoogle Scholar
5. Chaix B, Rosvall M, Merlo J. Neighborhood socioeconomic deprivation and residential instability: effects on incidence of ischemic heart disease and survival after myocardial infarction. Epidemiology. 2007;18(1):104111. Crossref, MedlineGoogle Scholar
6. Diez-Roux AV, Nieto FJ, Muntaner C, et al.. Neighborhood environments and coronary heart disease: a multilevel analysis. Am J Epidemiol. 1997;146(1):4863. Crossref, MedlineGoogle Scholar
7. Nordstrom CK, Diez-Roux AV, Jackson SA, Gardin JM. The association of personal and neighborhood socioeconomic indicators with subclinical cardiovascular disease in an elderly cohort. The Cardiovascular Health Study. Soc Sci Med. 2004;59(10):21392147. Crossref, MedlineGoogle Scholar
8. Smith GD, Hart C, Watt G, Hole D, Hawthorne V. Individual social class, area-based deprivation, cardiovascular disease risk factors, and mortality: the Renfrew and Paisley Study. J Epidemiol Community Health. 1998;52(6):399405. Crossref, MedlineGoogle Scholar
9. Stjarne MK, Fritzell J, De Leon AP, Hallqvist J. Neighborhood socioeconomic context, individual income and myocardial infarction. Epidemiology. 2006;17(1):1423. Crossref, MedlineGoogle Scholar
10. Sundquist K, Malmstrom M, Johansson SE. Neighborhood deprivation and incidence of coronary heart disease: a multilevel study of 2.6 million women and men in Sweden. J Epidemiol Community Health. 2004;58(1):7177. Crossref, MedlineGoogle Scholar
11. Winkleby M, Sundquist K, Cubbin C. Inequities in CHD incidence and case fatality by neighborhood deprivation. Am J Prev Med. 2007;32(2):97106. Crossref, MedlineGoogle Scholar
12. Brown P, Guy M, Broad J. Individual socio-economic status, community socio-economic status and stroke in New Zealand: a case control study. Soc Sci Med. 2005;61(6):11741188. Crossref, MedlineGoogle Scholar
13. Engstrom G, Jerntorp I, Pessah-Rasmussen H, Hedblad B, Berglund G, Janzon L. Geographic distribution of stroke incidence within an urban population: relations to socioeconomic circumstances and prevalence of cardiovascular risk factors. Stroke. 2001;32(5):10981103. Crossref, MedlineGoogle Scholar
14. Lisabeth LD, Diez-Roux AV, Escobar JD, Smith MA, Morgenstern LB. Neighborhood environment and risk of ischemic stroke: the brain attack surveillance in Corpus Christi (BASIC) Project. Am J Epidemiol. 2007;165(3):279287. Crossref, MedlineGoogle Scholar
15. Cozier YC, Palmer JR, Horton NJ, Fredman L, Wise LA, Rosenberg L. Relation between neighborhood median housing value and hypertension risk among black women in the United States. Am J Public Health. 2007;97(4):718724. LinkGoogle Scholar
16. Mujahid MS, Diez-Roux AV, Morenoff JD, et al.. Neighborhood characteristics and hypertension. Epidemiology. 2008;19(4):590598. Crossref, MedlineGoogle Scholar
17. Robert SA. Community-level socioeconomic status effects on adult health. J Health Soc Behav. 1988;39(1):1837. CrossrefGoogle Scholar
18. Robert SA, Li LW. Age variation in the relationship between community socioeconomic status and adult health. Res Aging. 2001;23(2):234259. CrossrefGoogle Scholar
19. Clarke CA, Glaser SL, Keegan TH, Stroup A. Neighborhood socioeconomic status and Hodgkin's lymphoma incidence in California. Cancer Epidemiol Biomarkers Prev. 2005;14(6):14411447. Crossref, MedlineGoogle Scholar
20. Eschbach K, Mahnken JD, Goodwin JS. Neighborhood composition and incidence of cancer among Hispanics in the United States. Cancer. 2005;103(5):10361044. Crossref, MedlineGoogle Scholar
21. Krieger N, Chen JT, Waterman PD, Soobader MJ, Subramanian SV, Carson R. Geocoding and monitoring of US socioeconomic inequalities in mortality and cancer incidence: does the choice of area-based measure and geographic level matter? The Public Health Disparities Geocoding Project. Am J Epidemiol. 2002;156(5):471482. Crossref, MedlineGoogle Scholar
22. Sundquist K, Winkleby M, Ahlen H, Johansson SE. Neighborhood socioeconomic environment and incidence of coronary heart disease: a follow-up study of 25 319 women and men in Sweden. Am J Epidemiol. 2004;159(7):655662. Crossref, MedlineGoogle Scholar
23. Diez Roux AV. Residential environments and cardiovascular risk. J Urban Health. 2003;80(4):569589. Crossref, MedlineGoogle Scholar
24. Leon Bluhm G, Berglind N, Nordling E, Rosenlund M. Road traffic noise and hypertension. Occup Environ Med. 2007;64(2):122126. Crossref, MedlineGoogle Scholar
25. Cagney KA, Browning CR. Exploring neighborhood-level variation in asthma and other respiratory diseases: the contribution of neighborhood social context. J Gen Intern Med. 2004;19(3):229236. Crossref, MedlineGoogle Scholar
26. Cagney KA, Browning CR, Wallace DM. The Latino paradox in neighborhood context: the case of asthma and other respiratory conditions. Am J Public Health. 2007;97(5):919925. LinkGoogle Scholar
27. Schootman M, Andresen EM, Wolinsky FD, et al.. The effect of adverse housing and neighborhood conditions on the development of diabetes mellitus among middle-aged African Americans. Am J Epidemiol. 2007;166(4):379387. Crossref, MedlineGoogle Scholar
28. Sundquist K, Theobald H, Yang M, Li X, Johansson SE, Sundquist J. Neighborhood violent crime and unemployment increase the risk of coronary heart disease: a multilevel study in an urban setting. Soc Sci Med. 2006;62(8):20612071. Crossref, MedlineGoogle Scholar
29. Diez Roux AV, Chambless L, Merkin SS, et al.. Socioeconomic disadvantage and change in blood pressure associated with aging. Circulation. 2002;106(6):703710. Crossref, MedlineGoogle Scholar
30. 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):99106. Crossref, MedlineGoogle Scholar
31. Hamil-Luker J, O'Rand AM. Gender differences in the link between childhood socioeconomic conditions and heart attack risk in adulthood. Demography. 2007;44(1):137158. Crossref, MedlineGoogle Scholar
32. Han MK, Postma D, Mannino DM, et al.. Gender and chronic obstructive pulmonary disease: why it matters. Am J Respir Crit Care Med. 2007;176(12):11791184. Crossref, MedlineGoogle Scholar
33. Seeman TE, Singer B, Wilkinson CW, McEwen B. Gender differences in age-related changes in HPA axis reactivity. Psychoneuroendocrinology. 2001;26(3):225240. Crossref, MedlineGoogle Scholar
34. Health and Retirement Study. HRS Core Final (version 2) and RAND HRS Data File (version G). Ann Arbor, MI: University of Michigan; 2002. Google Scholar
35. Juster TF, Cao H, Couper M, et al.. Enhancing the quality of data on the measurement of income and wealth. Ann Arbor, MI: University of Michigan Retirement Research Center; 2007. Available at: http://www.mrrc.isr.umich.edu/publications/papers. Accessed February 10, 2010. Retirement Research Center Research Paper No. WP 2007-151. Google Scholar
36. Heeringa SG, Connor JH. Technical description of the Health and Retirement Survey sample design. Ann Arbor, MI: University of Michigan; 1995. Available at: http://hrsonline.isr.umich.edu/sitedocs/userg/HRSSAMP.pdf. Accessed September 6, 2009. Google Scholar
37. Bush TL, Miller SR, Golden AL, Hale WE. Self-report and medical record report agreement of selected medical conditions in the elderly. Am J Public Health. 1989;79(11):15541556. LinkGoogle Scholar
38. Kehoe R, Wu SY, Leske MC, Chylack LT. Comparing self-reported and physician-reported medical history. Am J Epidemiol. 1994;139(8):813818. Crossref, MedlineGoogle Scholar
39. Freedman VA, Grafova I, Schoeni RF, Rogowski J. Neighborhoods and disability in later life. Soc Sci Med. 2008;66(11):22532267. Crossref, MedlineGoogle Scholar
40. Census Bureau. Census 2000 Population and Housing, Summary File 3 [United States]. Washington, DC: US Census Bureau; 2002. Google Scholar
41. US Dept of Justice, Federal Bureau of Investigation. Uniform crime reporting program data [United States]: county-level detailed arrest and offense data, 2002 [computer file]. Ann Arbor, MI: Inter-University Consortium for Political and Social Research; 2004. Google Scholar
42. US Census Bureau. 2000 Topologically Integrated Geographic Encoding and Referencing system. Available at: http://www.census.gov/geo/www/tiger. Accessed July 10, 2008. Google Scholar
43. Southworth M, Ben-Joseph E, eds. Streets and the Shaping of Towns and Cities. Chicago, IL: Island Press; 2003. Google Scholar
44. US Environmental Protection Agency. Technology transfer network, air quality control. Available at: http://www.epa.gov/ttn/airs/airsaqs/detaildata/downloadaqsdata.htm. Accessed July 29, 2008. Google Scholar
45. Guo G, Zhao H. Multilevel modeling of binary data. Annu Rev Sociol. 2000;26:441462. CrossrefGoogle Scholar
46. Snijder TAB, Bosker RJ. Multilevel analysis: an introduction to basic and advanced multilevel modeling. London, England: Sage; 1999. Google Scholar
47. Irigaray P, Newby JA, Clapp R, et al.. Lifestyle-related factors and environmental agents causing cancer: an overview. Biomed Pharmacother. 2007;61(10):640658. Crossref, MedlineGoogle Scholar
48. Acevedo-Garcia D, Lochner KA. Residential segregation and health. In: , Kawachi I, Berkman LF, eds. Neighborhoods and Health. New York, NY: Oxford University Press; 2003:265288. CrossrefGoogle Scholar
49. Williams DR, Collins C. Racial residential segregation: a fundamental cause of racial disparities in health. Public Health Rep. 2001;116(5):404416. Crossref, MedlineGoogle Scholar
50. Shihadeh ES, Flynn N. Segregation and crime: the effect of black social isolation on the rates of black urban violence. Soc Forces. 1996;74:13251352. CrossrefGoogle Scholar
51. Sampson RJ, Morenoff JD, Gannon-Rowley T. Assessing ‘neighborhood effects’: Social processes and new directions in research. Annu Rev Sociol. 2002;28:443478. CrossrefGoogle Scholar
52. Oakes JM. The (mis)estimation of neighborhood effects: causal inference for a practicable social epidemiology. Soc Sci Med. 2004;58(10):19291952. Crossref, MedlineGoogle Scholar
53. Longino CF, Bradley DE, Stoller EP, Haas WH. Predictors of non-local moves among older adults: a prospective study. J Gerontol B Psychol Sci Soc Sci. 2008;63(1):S7S14. Crossref, MedlineGoogle Scholar
54. Evans GW, Kantrowitz E. Socioeconomic status and health: the potential role of environmental risk exposure. Annu Rev Public Health. 2002;23:303331. Crossref, MedlineGoogle Scholar
55. Grafova IB, Freedman VA, Kumar R, Rogowski J. Neighborhoods and obesity in later life. Am J Public Health. 2008;98(11):20652070. LinkGoogle Scholar


No related items




Vicki A. Freedman, PhD, Irina B. Grafova, PhD, and Jeannette Rogowski, PhDAt the time of the study, Vicki A. Freedman, Irina B. Grafova, and Jeannette Rogowski were with the Department of Health Systems and Policy of the School of Public Health, University of Medicine and Dentistry of New Jersey, Piscataway. “Neighborhoods and Chronic Disease Onset in Later Life”, American Journal of Public Health 101, no. 1 (January 1, 2011): pp. 79-86.


PMID: 20299643