Neighborhood Context and Youth Cardiovascular Health Behaviors

 RESEARCH AND PRACTICE 

Neighborhood Context and Youth Cardiovascular Health Behaviors | Rebecca E. Lee, PhD, and Catherine Cubbin, PhD

Nearly a third of all women and half of all men younger than 40 years will develop cardiovascular disease (CVD) in the remaining years of their life.1 Poor dietary habits,2 physical inactivity,3 and tobacco smoking4 have been associated with an increased risk of CVD and named as the top 3 preventable causes of death in the United States.5 High-risk CVD profiles that include obesity, physical inactivity, and tobacco use are developing at younger ages.6–8 This trend is alarming in that poor cardiovascular health behaviors established during youth may contribute to CVD in adulthood.4,7,8 Among adults, relationships documenting cardiovascular health behaviors of racial or ethnic minority groups and those of low socioeconomic status (SES) are established.9–12 Racial or ethnic differences, or both, have been reported among youths in regard to body mass index,13 physical activity,3 and smoking6,13; however, these studies did not adequately account for differences in SES, which may lead to biased estimates of racial/ethnic differences in health behaviors.14–16 An additional problem is that individually measured SES does not account for exposures to neighborhood context that may affect human behavior. A newer approach to understanding racial/ethnic and SES differences and individual health behaviors has been to include measures of the social environment (e.g., the neighborhood context in which people live) to help explain why members of racial/ethnic minorities or persons of lower SES are more likely to develop high-risk cardiovascular behaviors. Recent studies involving adults have shown that neighborhood context may contribute independently to CVD outcomes after individual-level demographic and socioeconomic characteristics have been taken into account.17–19 In addition to demonstrat-

Objectives. This study sought to determine the relationships between race/ethnicity, socioeconomic status (SES), and cardiovascular health behaviors among youths and whether neighborhood characteristics are associated with such behaviors independently of individual characteristics. Methods. Linear models determined the effects of individual and neighborhood characteristics (SES, social disorganization, racial/ethnic minority concentration, urbanization) on dietary habits, physical activity, and smoking among 8165 youths aged 12 to 21 years. Results. Low SES was associated with poorer dietary habits, less physical activity, and higher odds of smoking. After adjustment for SES, Black race was associated with poorer dietary habits and lower odds of smoking. Hispanic ethnicity was associated with healthier dietary habits, lower levels of physical activity, and lower odds of smoking than non-Hispanic ethnicity. Low neighborhood SES and high neighborhood social disorganization were independently associated with poorer dietary habits, while high neighborhood Hispanic concentration and urbanicity were associated with healthier dietary habits. Neighborhood characteristics were not associated with physical activity or smoking. Conclusions. Changes in neighborhood social structures and policies that reduce social inequalities may enhance cardiovascular health behaviors. (Am J Public Health. 2002;92:428–436)

ing poor CVD outcomes, adults living in disadvantaged neighborhoods have reported poorer dietary habits,20 less physical activity,21 and more tobacco smoking17,18 than adults with similar characteristics living in advantaged neighborhoods. To our knowledge, no studies have examined these relationships among young people. Documenting a relationship between neighborhood characteristics and cardiovascular health behaviors would help guide public health interventions and policies aimed at changing health behaviors. The purpose of the current study was 2fold. First, we sought to determine, among youths, the relationships between race/ethnicity, SES, and cardiovascular health behaviors (dietary habits, physical activity, and tobacco smoking). Second, by linking youth data to census tract characteristics, we investigated whether characteristics of neighborhoods (SES, social disorganization, racial/ethnic minority concentration, and urbanization) were associated with individual cardiovascular health behaviors independently of individual characteristics.

428 | Research and Practice | Peer Reviewed | Lee and Cubbin

METHODS Data Analyses were based on 2 sources of data: the 1992 Youth Risk Behavior Survey (YRBS)22 and the 1990 US census.23 The YRBS was conducted as a follow back to the 1992 National Health Interview Survey (NHIS), a continuing annual household interview survey representative of the civilian noninstitutionalized population of the United States.24 Within each family interviewed for the NHIS, 1 child attending school and up to 2 children not enrolled in school (or whose school status was unknown) were selected for the YRBS sample (n = 10 645; age range: 12–21 years).22,25 The YRBS included questions about dietary habits, physical activity, and cigarette smoking. The NHIS included sociodemographic information for each youth respondent. The 1992 NHIS data included 1990 census geography codes (geocoded by NCHS based on household addresses), and these codes were used to link youths to their residential characteristics using census tract data.

American Journal of Public Health | March 2002, Vol 92, No. 3


About 20% (n = 2042) of the records for youths were not successfully geocoded and were dropped from the analysis. The remaining respondents, for whom geocoded records were available, resided in 3465 census tracts across the United States. Data from Summary Tape Files 1A and 3A of the 1990 census were used to construct the neighborhoodlevel variables. Census tracts were chosen to represent neighborhoods, because tracts are a good approximation of the neighborhood environment and because reliable social and economic data are available from the US Bureau of the Census. Census tracts include approximately 4000 people, and boundaries are delineated to encompass a relatively homogeneous population.26,27 Youths with missing data in regard to parental education (n = 37) and youths who were not Black, White, or Hispanic (n = 404) were excluded from the analysis. The final sample included the remaining youths without missing neighborhood information (n = 8165).

Dependent Variables Three cardiovascular health behaviors—dietary habits, physical activity, and tobacco smoking—were selected because of their strong association with preventable CVD risk. Diets higher in plant-based fiber and lower in saturated fats have been associated with lower CVD risk.2 Assessment of fruit and vegetable consumption may be more accurate in 24-hour recalls than in assessments made over longer periods of time.28 Six YRBS items focusing on respondents’ food consumption on the previous day were used in creating a summary index regarding dietary habits. These 6 items assessed consumption of (1) fruit; (2) green salad; (3) cooked vegetables; (4) hamburgers, hot dogs, or sausage; (5) french fries or potato chips; and (6) cookies, doughnuts, pie, or cake. The first 3 items were assigned a value of 1 if they were consumed once and a value of 2 if they were consumed more than once. The last 3 items were assessed in a similar way, except that they were assigned values of –1 and –2. Items not consumed were assigned a value of zero. The values for the 6 items were summed for each respondent, resulting in an overall dietary habits score; higher scores indicated healthier dietary habits.

Regular performance of physical activity on most days of the week3 and avoidance of tobacco smoking4 have been associated with a lower risk of developing CVD. Physical activity was measured through responses to the question “On how many of the last 7 days did you participate in any sports or exercise that made you sweat or breathe hard for at least 20 minutes at a time?” Tobacco use was defined through an assessment of the number of days in the previous 30 days on which the respondent had smoked at least 1 cigarette. This variable was dichotomized (0 days vs 1 or more days) to alleviate the skewed distribution.

tion of age were coded categorically. Correlations between individual-level variables were below 0.40. Neighborhood variables. Census tract variables were selected to measure neighborhood SES, social disorganization, racial/ethnic minority concentration, and urbanization. All variables (except urban–rural status) were coded via empirical quartiles to identify appropriate categories. Table 1 lists neighborhood variables and definitions. Correlations between neighborhood variables ranged from 0.10 to 0.90. Youths were assigned neighborhood variables according to their census tract of residence.

Independent Variables

Analysis

Individual variables. Correlates at the individual level included age, sex, self-reported race/ethnicity, income-to-needs ratio, and educational attainment of responsible adult family members. The income-to-needs ratio was created by taking the midpoint of the NHIS family income categories and dividing by family size. All variables with the excep-

To examine individual and neighborhood characteristics of those who engaged in unhealthy vs healthy cardiovascular behaviors, we dichotomized each of the 3 behaviors assessed. We identified unhealthy dietary habits were by scores ranging from –6 to 0 and healthy dietary habits were by scores ranging from 1 to 6. We dichotomized physical activity

TABLE 1—Neighborhood Variables and Definitions, 1990 US Census Tract Data Variable Socioeconomic status Family income Poverty Education Housing value Crowded housing Blue collar

Social disorganization Mobility Unemployment Housing tenure Female headship Poor female headship Divorced Racial/ethnic minority concentration Black Hispanic Urbanization Multi-unit housing Urban

March 2002, Vol 92, No. 3 | American Journal of Public Health

Definition

Median income for all households Proportion of persons whose annual income falls at or below 175% of the poverty line Proportion of persons 25 years and older with less than a high school education Median value of owner-occupied housing units Proportion of households with more than 1 person per room Proportion of employed persons in service occupations; farming and fishing occupations; precision production, crafts, and repair occupations; and operators, fabricators, and laborers Proportion of persons 5 years and older living in the same house for past 5 years Proportion of persons 16 years and older who are unemployed Proportion of occupied housing units that are rented Proportion of families headed by women Proportion of poor families that are headed by women Proportion of persons 15 years and older who are divorced or separated Proportion of all persons who are Black Proportion of all persons who are Hispanic Proportion of housing units with 5 or more units in structure Residence in an urban census tract

Lee and Cubbin | Peer Reviewed | Research and Practice | 429


as no days of activity vs any days of activity. We dichotomized smoking as described earlier. We used multiple linear (dietary habits, physical activity) and logistic (smoking) regression models in the analysis. Each neighborhood characteristic was tested in a separate model because of extreme multicollinearity between the neighborhood variables. First, individual demographic variables (age, sex, race/ethnicity) were entered in the baseline model. Second, the 2 individual SES variables (income-to-needs ratio, parental education) were added with the purpose of determining whether there were changes in racial or ethnic differences after adjustment for individual SES. Finally, neighborhood variables were added to the model to determine whether neighborhood characteristics were associated with each individual cardiovascular health behavior outcome after individual covariates had been taken into account. The NHIS is based on a complex multistage clustered sample design.29 The resulting survey design effects must be taken into account when one is producing parameter estimates generalizable to the national population.30 In these analyses, we accounted for design effects by using SUDAAN (version 7.11)31 with the final models to produce valid variance estimates. SUDAAN also alleviates similar difficulties with statistical inference introduced by multilevel research designs32,33; previous researchers have employed a similar approach.19,34–37 We did not use multilevel modeling techniques38,39 in this analysis, because, although the number of census tracts (or neighborhoods) was large (n = 3465), there were few youths per tract (more than 75% of the tracts included 3 or fewer sampled youths).40

RESULTS The sample was fairly evenly divided by sex; 49.5% of the respondents were male (n = 4038), and 50.5% were female (n = 4127). Nearly a fifth (19.1%; n = 1562) identified themselves as non-Hispanic Black, 12.5% (n = 12.5) identified themselves as Hispanic, and 68.4% (n = 5584) identified themselves as non-Hispanic White. The mean age was 16.5 years (SD = 4.5). Youth dietary habits clustered slightly below zero on the index

(mean = –0.35, SE = 0.03), physical activity frequency was lower than consensus recommendations for adolescents41 (mean = 3.04 days, SE = 0.03), and more than one fourth (28%) of the respondents smoked. Table 2 presents the individual and neighborhood characteristics of the sample. Relative to youths with less healthy dietary habits, youths with healthier dietary habits tended to be younger; were more likely to be White or Hispanic; and tended to be of higher SES and to live in neighborhoods characterized by higher SES, lower social disorganization, and lower Black concentrations. In comparison with youths who reported no physical activity, youths who were physically active tended to be younger, were more likely to be male and White, and tended to be of higher SES and to live in neighborhoods characterized by higher SES, lower social disorganization, and lower racial/ethnic minority concentrations. Youths who had smoked on at least 1 day in the previous month tended to be older, were more likely to be male or White, and tended to be of lower SES than nonsmokers. Neighborhood characteristics did not vary consistently among smokers and nonsmokers.

Individual Models Table 3 presents adjusted individual-level estimates for the 3 health behavior outcomes resulting from the multiple regression models. For each outcome, 2 models are included. The first models were adjusted for age, sex, and race/ethnicity, and the second models were adjusted for age, sex, race/ethnicity, and 2 measures of SES. Being male, Black, or a child of a parent or guardian with low educational attainment was associated with less healthy dietary habits, while Hispanic ethnicity was associated with healthier dietary habits. After adjustment for individual-level SES, Black–White differences in dietary habits were attenuated, while the Hispanic–White difference increased. Compared with their respective reference groups, youths who were older, Hispanic, and of lower SES were less likely to be physically active and male youths were more likely to be physically active. Being older, male, or a child of a parent or guardian with low educational attainment was associated with greater odds of smoking cigarettes at least 1 day in


the previous month. Even after adjustment for SES, Black and Hispanic youths were less likely to smoke.

Neighborhood Models Few significant neighborhood effects were found for physical activity and smoking, and these effects were inconsistent. Therefore, Table 4 presents neighborhood estimates (adjusted for individual-level demographic and socioeconomic characteristics) only for dietary habits. Only those models with statistically significant neighborhood effects are included. Individual estimates were generally consistent across the neighborhood-level models. Youths who resided in neighborhoods characterized by low income, high levels of poverty, low education, low housing values, and a high proportion of blue-collar workers were more likely to have poorer dietary habits than were youths living in higher-SES neighborhoods, independent of individual SES or demographic characteristics. In comparison with youths residing in neighborhoods with the lowest levels of mobility (i.e., proportions of residents relocating in the previous 5 years), youths residing in neighborhoods with higher levels of mobility had poorer dietary habits. Higher proportions of female-headed households in poverty were associated with poorer dietary habits. Residence in neighborhoods with the highest levels of Hispanic concentration was independently associated with healthier dietary habits. Residence in areas with the highest proportions of multi-unit housing (a proxy for urban residence) also was associated with healthier dietary habits. To determine whether the relationships between neighborhood characteristics and dietary habits were consistent for male and female youths and across ages, we stratified the final models (data not presented) by sex and age group (12–14, 15–17, 18–21 years). With the exception of the proportion of female-headed households living in poverty, neighborhood relationships in the models for male youths became weaker, while neighborhood relationships in the models for female youths became stronger, suggesting that the combined models may have been driven by the findings for female youths. With few exceptions, the neighborhood relationships in



TABLE 2—Individual and Neighborhood Characteristics for Respondents to the Youth Risk Behavior Survey, 1992, With Linkage to the 1990 US Census (n = 8165) Dietary Habitsa Unhealthy Individual characteristics Age, y mean Sex, % Male Female Race/ethnicity, % Black Hispanic Non-Hispanic White Parent/guardian education (%) Less than high school High school Some college College Income to needs, mean $ Neighborhood characteristics Socioeconomic status Family income, mean $ Poverty, % Education, % Housing value, mean $ Crowded housing, % Blue collar, % Social disorganization, % Mobility Unemployment Housing tenure Female headship Poor female headship Divorced Racial/ethnic minority concentration, % Black Hispanic Urbanization, % Multi-unit housing Urban

Physical Activity b

Smokingc

Healthy

None

Any

Yes

No

17

16

18

16

18

16

53 47

44 56

43 57

53 47

53 47

49 51

20 9 72

13 11 76

19 11 70

17 9 74

10 7 83

20 10 70

13 42 27 19 9 017

13 33 28 25 9 605

17 44 24 15 8 391

12 37 28 22 9 466

15 42 26 17 9 142

13 37 28 22 9 229

33 545 30 27 81 880 5 46

36 560 28 25 97 707 5 44

32 581 32 28 81 120 6 47

35 184 29 26 89 085 5 45

35 047 28 26 86 777 4 45

34 369 30 27 87 191 5 46

55 5 35 13 6 11

54 4 34 12 5 10

54 5 37 13 6 11

55 4 34 12 5 10

55 4 34 11 5 10

55 4 35 13 6 10

16 7

12 8

16 8

14 7

11 6

16 7

15 82

16 82

16 81

15 82

15 81

15 81

Note. Data were weighted and adjusted for the sample design. a Unhealthy = –6–0; healthy = 1–6. b None = 0 days; any = 1–7 days. c No = 0 days; yes = 1 or more days.

the younger (12–14 years) and older (18–21 years) age groups remained the same or became stronger, whereas there were no significant neighborhood characteristics in the middle age group (15–17 years).

DISCUSSION This study disentangled the effects of individual SES, race/ethnicity, and neighborhood social context on individual youth cardiovas-


cular health behaviors. The analysis described here is an important step toward understanding how social disadvantage, at both the individual and neighborhood levels, may increase the risk of developing behaviors that



TABLE 3—Weighted Individual-Level Estimates for Respondents to the Youth Risk Behavior Survey, 1992 Dietary Habits (n = 8090), Demoa Age, y Sex Male Female Race/ethnicity Non-Hispanic Black Hispanic Non-Hispanic White Education Less than high school High school Some college College Family income, $ Missing income 14 167

Physical Activity (n = 8113), Fullb

Smoking (n = 7993), Odds Ratio

Demoa

Fullb

Demoa

Fullb

–0.01

–0.01

–0.26***

–0.26***

1.21***

1.22***

–0.50*** Reference


0.76*** Reference

0.75*** Reference

1.21** Reference

1.22** Reference

–0.55*** 0.19* Reference

–0.51*** 0.21* Reference

–0.10 –0.40*** Reference

0.05 –0.23* Reference

0.41*** 0.62*** Reference

0.35*** 0.49*** Reference

–0.23* –0.38*** –0.18* Reference

–0.39** –0.38*** 0.07 Reference

1.86*** 1.39** 1.02 Reference

–0.12 –0.00 –0.11 –0.14 Reference

–0.29** –0.31** –0.30** –0.18 Reference

1.27* 1.19 1.14 1.07 Reference

Note. Estimates were adjusted for the sample design. Beta coefficients were derived from multiple linear models. a Demo = demographic model adjusting for age, sex, and race/ethnicity. b Full = full model adjusting for age, sex, race/ethnicity, responsible adult educational attainment, and income-to-needs ratio. *P < .05; **P < .01; ***P < .001.

contribute to CVD. Our findings suggest that low SES compromises young people’s ability to sustain health behaviors that are protective in regard to CVD. We found expected relationships between race/ethnicity, SES, and the 3 cardiovascular health behavior outcomes. After adjustment for SES, Black youths had poorer dietary habits and were less likely to smoke than White youths and Hispanic youths had better dietary habits, were less likely to smoke, and were less likely to be physically active than White youths. Residence in neighborhoods characterized by low SES and high social disorganization was related to poorer dietary behaviors, while residence in neighborhoods with high Hispanic concentrations or urban areas was related to better dietary habits, independent of individual demographic and socioeconomic characteristics. Black–White differences in dietary habits were attenuated after adjustment for neighborhood SES and social disorganization.

Although it is possible that residual confounding by SES exists even after adjustment for family income and parental educational attainment,42 it is likely that individual racial differences are explained in part by mechanisms that operate through the neighborhood environment.15 Neighborhood SES may influence the dietary habits of all residents regardless of their own individual race, ethnicity, or SES. Socioeconomically disadvantaged neighborhoods may have fewer large supermarkets and fewer nutritious food options, sold at higher costs,43,44 limiting the dietary choices of residents in terms of both availability and price. Lack of access to food sources, coupled with higher reliance on public transportation, may create significant barriers to consumption of fruits and vegetables. Residence in socially disorganized neighborhoods (indicated by high residential turnover and female-headed households living in poverty) was related to poorer youth dietary habits. A high degree of female-headed


households is a proxy for social disorganization and is associated with higher CVD risk among women.19 Greater social disorganization may indicate lower social control,45 fewer social resources,19 and higher crime rates,46 increasing individual levels of adult stress and inadequate monitoring of youths. In situations involving greater stress levels, adults may increase their reliance on “fast” food rather than allocating time and energy to shop for groceries and prepare food,19 and youths in environments with less social control may rely on unwholesome, convenient foods (e.g., candy and chips). Ethnically homogeneous neighborhoods may retain greater social organization and control while simultaneously retaining traditional dietary norms more readily than more diverse neighborhoods. Hispanic dietary norms may not include “American” foods such as hamburgers, hot dogs, french fries, and potato chips, reducing the demand for and availability of these kinds of foods.47 For



TABLE 4—Weighted Estimates of Dietary Habits for Respondents to the Youth Risk Behavior Survey, 1992, With Linkage to the 1990 US Census (n = 8090) 1

2

3

4

–0.01

–0.01

–0.01

5

6

7

8

9

Individual-level variables Age, y

–0.01

–0.01

–0.01

–0.01

–0.01

–0.01

Sex Male

–0.49***

Female

Reference






–0.49***

–0.50***

–0.50***

Reference

Reference

Reference

–0.52***

–0.52***

Race/ethnicity Non-Hispanic Black Hispanic Non-Hispanic White

–0.47***

–0.47***

–0.44***

0.21*

0.23*

0.25**

Reference

Reference

–0.47***

–0.47***

–0.50***

–0.43***

0.12

0.21*

0.18*

0.24**

0.07

Reference

Reference

Reference

Reference

Reference

Reference

Reference –0.21

0.18

Education Less than high school

–0.16

–0.18

–0.12

–0.16

–0.12

–0.21*

–0.19

–0.23*

High school

–0.32***

–0.33***

–0.31***

–0.30***

–0.30**

–0.37***

–0.36***

–0.37***

–0.37***

Some college

–0.15

–0.15*

–0.16*

–0.15

–0.15*

–0.19*

–0.17*

–0.19*

–0.18*

College or more

Reference

Reference

Reference

Reference

Reference

Reference

Reference

Reference

Reference –0.12

Income, $ Missing income

–0.04

–0.06

–0.06

–0.03

–0.06

–0.12

–0.09

–0.10

Less than 3750

–0.11

0.10

0.06

0.11

0.06

–0.03

0.05

–0.02

0.01

3750–7500

–0.02

–0.04

–0.05

–0.02

–0.05

–0.12

–0.07

–0.09

–0.11

7500–14 167 Greater than 14 167 Neighborhood level variables 1. Family income, $ 0–25 953 25 953–33 271 33 271–42 933 42 933–150 001 2. Poverty, % 0–13.6 13.6–24.8 24.8–38.1 38.1–100 3. Education, % 0–14.9 14.9–24.0 24.0–35.6 35.6–83.1 4. Housing value, $ 0–46 200 46 200–68 400 68 400–121 000 121 000–500 001 5. Blue collar, % 0–33.9 33.9–45.7 45.7–56.7 56.7–100

–0.08 Reference

–0.09

–0.09

–0.07

Reference

Reference

Reference

–0.09

–0.14

Reference

Reference

–0.11

–0.12

–0.13

Reference

Reference

Reference

–0.30** –0.36*** –0.25** Reference Reference –0.18* –0.31*** –0.24* Reference –0.23** –0.34*** –0.37*** –0.44*** –0.42*** –0.20* Reference Reference –0.25** –0.28** –0.41*** Continued




TABLE 4—Continued 6. Mobility, % 0–48.3 48.3–57.0 57.0–64.4 64.4–90.4 7. Poor female headship, % 0–1.0 1.0–2.6 2.6–6.3 6.3–74.4 8. Hispanic, % 0–0.5 0.5–1.7 1.7–6.7 6.7–98.0 9. Multi-unit housing, % 0–1.5 1.5–7.7 7.7–21.9 21.9–100

–0.13 –0.21** –0.22** Reference Reference –0.15* –0.16* –0.24* Reference –0.02** 0.20** 0.25** Reference 0.11 0.07 0.16*

Note. Values are data coefficients from multiple linear models, adjusted for the sample design. *P < .05; **P < .01; ***P < .001.

example, Mexican diets may rely on traditionally available foods (e.g., rice, beans, corn), in turn dictating which foods are available in local stores and subsequently eaten.48 Our findings suggest that girls may be more sensitive to neighborhood characteristics than are boys. Perhaps the socialization of girls encourages less independence in behavioral choices, which may be reflected in dietary behaviors. Some have suggested that, during adolescence, the behavior of girls may be more biologically driven than that of boys by pubertal, hormonal changes (e.g., onset of menses).49,50 Family-oriented hormonal changes may subtly encourage girls to be more reactive to environmental conditions. Our results also show that the dietary behavior of middle teens (those aged 15–17 years) may not be as strongly associated with neighborhood characteristics as are the dietary behavior of adolescents at other ages. The surge toward autonomy in adolescence reaches its zenith at this developmental stage, and perhaps these teens are less reactive to environmental conditions. Further research is needed to document and clarify this relationship.51

The main strength of this study was that individuals were linked to their neighborhoods through a nationally representative data set, allowing investigation of neighborhood effects on CVD risk behaviors and generalization to the US population of youths aged 12 to 21 years. Data for the 1992 YRBS were collected in the form of a follow-back survey to the 1992 NHIS rather than a school-based survey (as used in subsequent YRBS designs), including youths from a wide age range and those both in and out of school. Furthermore, SES was based on parent rather than youth responses, providing more complete and accurate measures. Multiple measures of neighborhood constructs (e.g., family income and crowded housing as measures of neighborhood SES) were tested to explore the specific characteristics of neighborhoods that were related to the outcomes and whether consistent results were found across multiple measures of the same construct. Neighborhood characteristics were measured in 1990, whereas the YRBS data were collected in 1992, presenting the possibility of simultaneity bias; however, neighborhoods


generally do not change significantly over the short time period involved.52 Of greater concern is that neighborhood measures are static. This analysis represents a snapshot of the relationship between neighborhood context and cardiovascular behaviors in youths for whom there is no information on neighborhood history, dynamics, or length of exposure to neighborhood environments.53 Additional analyses were conducted to assess how missing geocodes affected inferences (data not shown). Respondents whose records were missing geocodes were more likely to be Hispanic and of higher SES than respondents with geocoded records. We also ran the individual-level models with the full sample (including youths not geocoded) and found that the results were not notably biased by missing geocodes. The outcome measures were limited to the questions included on the YRBS. The dietary habits index does not necessarily reflect the breadth of foods available and eaten in less “Americanized” neighborhoods. Physical activity is notoriously difficult to assess accurately.54 The current assessment may have



underestimated or overestimated true physical activity patterns owing to poor memory or social desirability concerns. Previous research has shown associations between residence in a poverty area and physical activity over time among adults,21 and these relationships may not be evident in a cross-sectional assessment of youths. Youth smoking remains difficult to assess, both because of the inconsistent pattern of initiation and because of the covert nature of illegal behavior. Smoking among adults has been independently associated with neighborhood conditions17,18; Further research is needed to assess such associations among young people. It may be that relationships between neighborhood context and certain individual health behaviors are not apparent until adulthood. Neighborhood-level measures would also profit from refinement. Merely assessing income and other measures of SES may not result in adequate descriptions of the true nature of a particular neighborhood.20 Neighborhood characteristics drawn from other sources (e.g., topography, municipal policies) may be associated with youth health behaviors, but these relationships are unclear. Neighborhood data sources exist; however, topography and policy data are generally assessed at a local (i.e., city or county) level and are not standardized across municipalities, making comparisons involving large data sets challenging. Finally, the measure of individual Hispanic ethnicity included a heterogeneous group of youths who were combined into one category, because the numbers were not sufficiently large to allow a separate examination of each subgroup. Separate descriptive analyses (data not shown) indicated that 60% of the respondents were of Mexican origin (“Mexican–Mexicano, Mexican American, Chicano”), 14% were “Puerto Ricans,” 10% were “other Latin Americans,” 9% were “other Spanish,” and 2% were “Cubans.” In comparison with non-Hispanics, Hispanic subgroups had healthier dietary habits (except Puerto Ricans), lower levels of physical activity, and lower levels of smoking (except Puerto Ricans). US adults of Puerto Rican descent have been shown to score slightly lower on health

indicators and to smoke more than US adults of other Hispanic origins.55 Perhaps poorer health among Puerto Ricans originates at early ages as a result of environmental contexts. Nevertheless, our generally consistent patterns among Hispanic youths lend strength to our conclusions based on the Hispanic sample as a whole. Because the proportion of Puerto Ricans was relatively small, the differences shown in this group were not likely to seriously bias our results for Hispanic youths. The results of the present study show that both individual and neighborhood characteristics are independently related to dietary habits in an ethnically diverse, nationally representative sample of youths. Examining the relationship between environment and human behavior is a promising direction for future research attempting to provide an understanding of why poor health behaviors persist in the face of overwhelming evidence that such behaviors are detrimental to human health. Geographic methods that address neighborhood characteristics are recommended to understand and interpret these effects. Interventions are needed that consider the role of neighborhood environments in shaping individual health behaviors. Our findings suggest that changes in neighborhood social structures and policies that reduce social inequalities may assist in lessening the burden and cost of CVD.

About the Authors At the time of the study, Rebecca E. Lee was with the Stanford Center for Research in Disease Prevention, Stanford University School of Medicine, Palo Alto, Calif. Catherine Cubbin is with the Stanford Center for Research in Disease Prevention, Stanford University School of Medicine, and the Department of Family and Community Medicine, University of California, San Francisco. Requests for reprints should be sent to Rebecca E. Lee, PhD, Department of Preventive Medicine, University of Kansas School of Medicine, 3901 Rainbow Blvd, Kansas City, KS 66160 (e-mail: [email protected]). This article was accepted January 10, 2001.

Contributors R. E. Lee conceived of the research, with inspiration from C. Cubbin. C. Cubbin conducted analyses after substantial discussion with R. E. Lee. Both authors contributed to the interpretation and writing of the report.

institutional grant HL 58914 from the National Heart, Lung, and Blood Institute (Drs Cubbin and Lee). We gratefully acknowledge Drs Gregory J. Norman and Nina C. Schleicher for their insightful comments on an earlier version of the article.

References 1. Lloyd-Jones DM, Larson MG, Beiser A, Levy D. Lifetime risk of developing coronary heart disease. Lancet. 1999;353:89–92. 2. Srinivasan SR, Bao W, Wattigney WA, Berenson GS. Adolescent overweight is associated with adult overweight and related multiple cardiovascular risk factors: the Bogalusa Heart Study. Metabolism. 1996;45: 235–240. 3. Physical Activity and Health. Atlanta, Ga: Public Health Service, Centers for Disease Control and Prevention; 1996. 4. Preventing Tobacco Use Among Young People: A Report of the Surgeon General. Atlanta, Ga: Public Health Service, Centers for Disease Control and Prevention; 1994. 5. McGinnis JM, Foege WH. Actual causes of death in the United States. JAMA. 1993;270:2207–2212. 6. Centers for Disease Control and Prevention. Tobacco use among middle and high school students— United States, 1999. MMWR Morb Mortal Wkly Rep. 2000;49:49–53. 7. Bao W, Srinivasan SR, Valdez R, Greenlund KJ, Wattigney WA, Berenson GS. Longitudinal changes in cardiovascular risk from childhood to young adulthood in offspring of parents with coronary artery disease: the Bogalusa Heart Study. JAMA. 1997;278: 1749–1754. 8. Pierce JP, Gilpin E. How long will today’s new adolescent smoker be addicted to cigarettes? Am J Public Health. 1996;86:253–256. 9. Tobacco Use Among US Racial/Ethnic Minority Group—African Americans, American Indians and Alaska Natives, Asian Americans and Pacific Islanders, and Hispanics: A Report of the Surgeon General. Atlanta, Ga: Centers for Disease Control and Prevention; 1998. 10. Hahn RA, Heath GW, Chang MH. Cardiovascular disease risk factors and preventive practices among adults—United States, 1994: a behavioral risk factor atlas. MMWR Morb Mortal Wkly Rep. 1998;47(SS-5): 35–69. 11. Winkleby MA, Ahn DK. Blood pressure findings for Mexican-American women and men from the third National Health and Nutrition Examination Survey, 1988–1994. BioMedicina. 1998;1:321–325. 12. Winkleby MA, Kraemer HC, Ahn DK, Varady AN. Ethnic and socioeconomic differences in cardiovascular disease risk factors. JAMA. 1998;280:356–362. 13. Winkleby MA, Robinson TN, Sundquist J, Kraemer HC. Ethnic variation in cardiovascular disease risk factors among children and young adults. JAMA. 1999; 281:1006–1013.

Acknowledgments

14. Jones CP, LaVeist TA, Lillie-Blanton M. “Race” in the epidemiologic literature: an examination of the American Journal of Epidemiology, 1921–1990. Am J Epidemiol. 1991;134:1079–1084.

This work was supported in part by a joint Association of Schools of Public Health and Centers for Disease Control and Prevention fellowship (Dr Cubbin) and by

15. Lillie-Blanton M, Laveist T. Race/ethnicity, the social environment, and health. Soc Sci Med. 1996;43: 83–91.




16. Williams DR. Race/ethnicity and socioeconomic status: measurement and methodological issues. Int J Health Serv. 1996;26:483–505. 17. Diez-Roux AV, Nieto FJ, Nuntaner C, et al. Neighborhood environments and coronary heart disease: a multilevel analysis. Am J Epidemiol. 1997;146:48–63. 18. Davey-Smith G, 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;52:399–405. 19. LeClere FB, Rodgers RG, Peters K. Neighborhood social context and racial differences in women’s heart disease mortality. J Health Soc Behav. 1998;39:91–107. 20. Diez-Roux AV, Nieto FJ, Caulfield L, Tyroler HA, Watson RL, Szklo M. Neighbourhood differences in diet: the Atherosclerosis Risk in Communities (ARIC) Study. J Epidemiol Community Health. 1999;53:55–63. 21. Yen IH, Kaplan GA. Poverty area residence and changes in physical activity level: evidence from the Alameda County Study. Am J Public Health. 1998;88: 1709–1712. 22. Data File Documentation, National Health Interview Youth Risk Behavior Survey, 1992. Hyattsville, Md: National Center for Health Statistics; 1993. 23. Census of Population and Housing, 1990: Summary Tape File 3 on CD-ROM. Washington, DC: US Bureau of the Census; 1992. 24. Adams PF, Benson V. Current estimates from the National Health Interview Survey. Vital Health Stat 10. 1991;No. 181. 25. Brener ND, Collins JL, Kann L, Warren CW, Williams BI. Reliability of the Youth Risk Behavior Survey questionnaire. Am J Epidemiol. 1995;141:575–580. 26. South SJ, Crowder KD. Escaping distressed neighborhoods: individual, community and metropolitan influences. Am J Sociol. 1997;102:1040–1084. 27. Krieger N. Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based methodology. Am J Public Health. 1992;82:703–710. 28. Field AE, Colditz GA, Fox MK, et al. Comparison of 4 questionnaires for assessment of fruit and vegetable intake. Am J Public Health. 1998;88:1216–1218. 29. Massey JT, Moore TF, Parsons VL, Tadros W. Design and estimation for the National Health Interview Survey, 1985–1994. Vital Health Stat 2. 1989;No. 110. 30. Skinner CN, Holt D, Smith TMF, eds. Analysis of Complex Surveys. New York, NY: John Wiley & Sons Inc; 1989. 31. SUDAAN: Software for the Statistical Analysis of Correlated Data. Research Triangle Park, NC: Research Triangle Institute; 1997. 32. Soobader M. Effects of Income Inequality on Morbidity [dissertation]. Ann Arbor: University of Michigan; 1998. 33. Goldstein H. Multilevel Statistical Models. London, England: Arnold Publishers; 1999. 34. LeClere FB, Rogers RG, Peters KD. Ethnicity and mortality in the United States: individual and community correlates. Soc Forces. 1997;76:169–198. 35. Cubbin C, LeClere FB, Smith GS. Socioeconomic status and injury mortality: individual and neighbour-

hood determinants. J Epidemiol Community Health. 2000;54:517–524. 36. Soobader MJ, LeClere FB. Aggregation and the measurement of income inequality: effects on morbidity. Soc Sci Med. 1999;48:733–744. 37. Robert SA. Community-level socioeconomic status effects on adult health. J Health Soc Behav. 1998;39: 18–37. 38. Hox JJ. Applied Multilevel Analysis. Amsterdam, the Netherlands: TT-Publikaties; 1995. 39. Bryk AS, Raudenbush SW. Hierarchical Linear Models: Applications and Data Analysis Methods. Newbury Park, Calif: Sage Publications; 1992. 40. Duncan GJ, Connel JP, Klebanov PK. Conceptual and methodological issues in estimating causal effects of neighborhoods and family conditions on individual development. In: Brooks-Gunn J, Duncan GJ, Aber JL, eds. Neighborhood Poverty: Context and Consequences for Children. New York, NY: Russell Sage Foundation; 1997:219–250.

52. Geronimus AT, Bound J. Use of census-based aggregate variables to proxy for socioeconomic group: evidence from national samples. Am J Epidemiol. 1998; 148:475–486. 53. Tienda M. Poor people and poor places: deciphering neighborhood effects on poverty outcomes. In: Huber J, ed. Macro-Micro Linkages in Sociology. Newbury Park, Calif: Sage Publications; 1991. 54. Sarkin JA, Nichols JF, Sallis JF, Calfas KJ. Selfreport measures and scoring protocols affect prevalence estimates of meeting physical activity guidelines. Med Sci Sports Exerc. 2000;32:149–156. 55. Hajat A, Lucas JB, Kington R. Health outcomes among Hispanic subgroups: United States, 1992–1995. Adv Data Vital Health Stat. February 25, 2000;310.

41. Sallis JF, Patric K, Long BJ. Overview of the International Consensus Conference on Physical Activity guidelines for adolescents. Pediatr Exerc Sci. 1994;6: 299–301. 42. Kaufman JS, Cooper RS, McGee DL. Socioeconomic status and health in blacks and whites: the problem of residual confounding and the resiliency of race. Epidemiology. 1997;8:621–628. 43. 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. 1993;51: 276–284. 44. Troutt DD. The Thin Red Line: How the Poor Still Pay More. San Francisco, Calif: West Coast Regional Office, Consumers Union; 1993. 45. Sampson RJ. Family management and child development: insights from social disorganization theory. In: McCord J, ed. Facts, Frameworks and Forecasts. New Brunswick, NJ: Transaction Publishers; 1992:63–93. 46. Sampson RJ, Raudenbush SW, Earls F. Neighborhoods and violent crime: a multilevel study of collective efficacy. Science. 1997;277:918–924. 47. Block G, Norris JC, Mandel RM, DiSogra C. Sources of energy and six nutrients in diets of lowincome Hispanic-American women and their children: quantitative data from HHANES, 1982–1984. J Am Diet Assoc. 1995;95:195–208. 48. Dixon LB, Sundquist J, Winkleby M. Differences in energy, nutrient, and food intakes in a US sample of Mexican-American women and men: findings from the third National Health and Nutrition Examination Survey, 1988–1994. Am J Epidemiol. 2000;152: 548–557. 49. Cordain L, Gotschall RW, Eaton SB. Physical activity, energy expenditure and fitness: an evolutionary perspective. Int J Sports Med. 1998;19:328–335. 50. Goran MI, Gower BA, Nagy TR, Johnson R. Developmental changes in energy expenditure and physical activity in children: evidence for a decline in physical activity in girls before puberty. Pediatrics. 1998; 101:887–891. 51. Zaslow MJ, Takanishi R. Priorities for research on adolescent development. Am Psychol. 1993;48: 185–192.

