International Journal of Obesity (2007) 31, 968–977 & 2007 Nature Publishing Group All rights reserved 0307-0565/07 $30.00 www.nature.com/ijo

ORIGINAL ARTICLE Perceived and observed neighborhood indicators of obesity among urban adults TK Boehmer1, CM Hoehner1, AD Deshpande1, LK Brennan Ramirez1,2 and RC Brownson1 1 Prevention Research Center and Department of Community Health, School of Public Health, Saint Louis University, St Louis, MO, USA and 2Transtria, LLC, St Louis, MO, USA

Objective: The global obesity epidemic has been partially attributed to modern environments that encourage inactivity and overeating, yet few studies have examined specific features of the physical neighborhood environment that influence obesity. Using two different measurement methods, this study sought to identify and compare perceived and observed neighborhood indicators of obesity and a high-risk profile of being obese and inactive. Design: Cross-sectional telephone surveys (perceived) and street-scale environmental audits (observed) were conducted concurrently in two diverse US cities to assess recreational facility access, land use, transportation infrastructure and aesthetics. Subjects: A total of 1032 randomly selected urban residents (20% obese, 32% black, 65% female). Analysis: Bivariate and multivariate logistic regression analyses were conducted to estimate the association (adjusted prevalence odds ratio (aOR)) between the primary outcome (obese vs normal weight) and perceived and observed environmental indicators, controlling for demographic variables. Results: Being obese was significantly associated with perceived indicators of no nearby nonresidential destinations (aOR ¼ 2.2), absence of sidewalks (aOR ¼ 2.2), unpleasant community (aOR ¼ 3.1) and lack of interesting sites (aOR ¼ 4.8) and observed indicators of poor sidewalk quality (aOR ¼ 2.1), physical disorder (aOR ¼ 4.0) and presence of garbage (aOR ¼ 3.7). Perceived and observed indicators of land use and aesthetics were the most robust neighborhood correlates of obesity in multivariate analyses. Conclusions: The findings contribute substantially to the growing evidence base of community-level correlates of obesity and suggest salient environmental and policy intervention strategies that may reduce population-level obesity prevalence. Continued use of both measurement methods is recommended to clarify inconsistent associations across perceived and observed indicators within the same domain. International Journal of Obesity (2007) 31, 968–977; doi:10.1038/sj.ijo.0803531; published online 16 January 2007 Keywords: environment; measurement; neighborhood; physical activity; urban

Introduction The escalating prevalence of obesity is well recognized as a global health priority.1,2 The etiology of obesity involves complex interactions between multiple genetic, metabolic, behavioral and environmental factors.3–5 The social–ecological theory of health behavior proposes that the physical and social environments influence obesity through their effect on diet and physical activity behaviors.6,7 Modern obesogenic environments, characteristic of developed countries, promote inactivity and over-eating on a population level.7,8 Prior research indicates that attributes of urban neighborhoods (e.g., access to recreational facilities,

Correspondence: Dr T Boehmer, 724 S. Washington Street, Denver, CO 80209, USA. E-mail: [email protected] Received 13 March 2006; revised 4 September 2006; accepted 16 October 2006; published online 16 January 2007

aesthetics and mixed land use) are associated with recreational and/or transportation activity;9–12 however, the identification of environmental correlates of obesity remains under-explored. Although European studies have shown a consistent correlation between neighborhood deprivation and obesity prevalence,13–15 the relationship between obesity and specific urban design16–18 and neighborhood19–21 features has only recently been examined, primarily in the US and Australia. This study furthers existing obesity research in two important areas – assessment of the neighborhood environment and exploration of a high-risk profile. First, the majority of previous studies have utilized either perceived (i.e. self-report telephone surveys)19,20 or observed (i.e. environmental audits, existing geographic databases)16–18 methods to assess obesogenic environments (only one study has used both methods)21, and information on the correspondence between perceived and observed methods is sparse.22 Thus, it is unknown whether one measurement

Neighborhood indicators of obesity in urban adults TK Boehmer et al

969 method is more effective than the other for assessing the neighborhood environment and its association with obesity. Second, recent research indicates that high levels of physical activity (or cardiorespiratory fitness) may attenuate the negative health burden of obesity.23–26 As yet, environmental risk factors unique to a high-risk (obese and inactive) population have not been thoroughly explored.20 Further investigation of obesogenic neighborhoods may inform the development of effective health-promoting environmental and policy interventions with the prospect of reducing morbidity, mortality and medical costs associated with obesity. Using telephone survey and environmental audit data from diverse US urban settings, the current study sought to identify perceived and observed neighborhood characteristics associated with obesity and determine which measurement method provides the most robust neighborhood indicators of obesity.

Methods Study design A telephone survey and environmental audit were conducted concurrently among higher and lower income areas Table 1

of two US cities: Savannah, Georgia (representing a highwalkable city) and St Louis, Missouri (representing a low-walkable city). Higher and lower income study areas were matched on number of households, percentage of the population below poverty in 1999 and geographic size (Table 1).12 This study was reviewed and approved by the Saint Louis University Institutional Review Board.

Data collection The telephone survey collected information on respondents’ demographic characteristics, physical activity behavior, height and weight and perceptions of the neighborhood environment. Environmental items were derived from existing questionnaires with established test–retest reliability.11,18,27–30 Physical activity behavior was measured by using the international physical activity questionnaire.31 Surveys were conducted between February and June 2003 using a modification of the behavioral risk factor surveillance system sampling scheme and computer-assisted telephone interviewing techniques. A total of 1073 adults residing within Savannah (n ¼ 600) and St Louis (n ¼ 473) participated in the telephone survey (response rate ¼ 45%).

Study area and participant characteristics of the total sample and by study area Total

St Louis Missouri

P-value a

Savannah Georgia

Lower income

Higher income

Lower income

Higher income

11 1154 4.5 29.8

3 282 1.3 59.0

1 193 1.0 4.5

5 330 1.1 51.3

2 349 1.1 7.8

F F F F

1032 65.2 32.3

211 78.2 95.3

241 65.2 0.4

256 62.5 46.1

324 59.0 4.0

F o0.001 o0.001

Age group % (years) 18–29 30–39 40–49 50–59 60–69 70+

20.7 18.4 19.4 15.5 11.9 14.1

17.1 15.6 21.8 20.4 13.3 11.9

11.2 19.9 23.2 13.3 9.1 23.2

41.8 19.1 10.9 10.9 10.6 6.6

13.6 18.5 21.6 17.6 14.2 14.5

o0.001

Educational attainment % College graduate Some college High school graduate Less than high school

42.3 24.1 21.3 12.2

11.4 18.0 37.4 33.2

44.0 26.1 25.3 4.6

33.2 30.9 20.7 15.2

68.5 21.3 8.3 1.9

o0.001

Weight classification % Obese Overweight Normal weight Underweight

20.4 34.1 43.2 2.3

40.3 36.5 22.3 1.0

15.8 32.8 50.2 1.2

20.7 27.7 47.7 3.9

10.5 38.6 48.2 2.8

o0.001

Study area characteristics Number of census tracts Number of street segments Area (square miles) Percentage population below poverty b Participant characteristics Number of respondents Female % Non-Hispanic black %

a 2

w statistic comparing distribution of participant characteristics across four study areas. bWeighted average by number of households per census tract.

International Journal of Obesity

Neighborhood indicators of obesity in urban adults TK Boehmer et al

970 A reliable and comprehensive audit instrument was developed from available tools to systematically measure the neighborhood environment through direct observation.32 Global positioning system technology was used to record location and attribute information for each street segment (i.e. distance between two consecutive intersections) in Savannah (n ¼ 679) and St Louis (n ¼ 475) between March and May 2003. Details regarding the audit instrument, data collection methods and reliability have been described previously.12,32 Agreement between survey and audit indicators has also been evaluated in this study population.22 Geographic information systems software (ESRI ArcView 8.3, Redlands, CA, USA, 2002) was used to link the survey and audit data. Street-segment data from the audit was summarized within a 400-m radius (approximately 5-min walk) surrounding each respondent’s residence.

Measures Dependent variables. Two dichotomous dependent variables were examined in this study: obese vs normal weight and obese/inactive vs normal weight/active. Body mass index (BMI, kg/m2) was calculated from self-reported weight and height and categorized as normal weight (BMI ¼ 18.5–24.9), overweight (BMI ¼ 25.0–29.9) and obese (BMIX30.0) according to international standards.3 Respondents were classified as ‘active’ if they met current physical activity recommendations through transportation or recreational activities (X30 cumulative minutes of walking or moderate activity 5 days per week or X20 continuous minutes of vigorous activity 3 days per week).33 Those who did not achieve recommendations were classified as ‘inactive’. The second dependent variable combined BMI and physical activity to identify the highest risk (obese and inactive) and lowest risk (normal weight and active) subgroups.24,26 Overweight was not included in the main analysis because, compared with obesity, the health effects are unclear34 and there is a greater potential to misclassify lean, muscular persons (especially men) into this group.35 For completeness, however, additional analyses were performed to examine overweight vs normal weight, and the findings are discussed in brief. Environmental indicators. Perceived and observed characteristics of the neighborhood environment were classified into four domains: recreational facilities, land use, transportation and aesthetics. The items and response options for survey and audit measures are provided in Tables 2 and 3. The telephone survey consisted of ordinal and open-ended items. Most environmental items had a four-level response scale; adjacent response categories were combined if a category contained fewer than 5% of respondents. Openended items assessed how many minutes it took to walk from home to the nearest of five types of recreational facilities (e.g., park, trail, recreation center) and 13 types of nonresidential destinations (e.g., supermarket, restaurant, post office). The International Journal of Obesity

perceived number of recreational facility types and destination types within a 5-min walk from home (B400 m) were categorized into quartiles. Two additional perceived measures were created: distance (i.e., walking time) to the nearest recreational facility and no nonresidential destinations within a 10-min walk. Aggregation of the street-segment audit data within each respondent’s 400-m buffer resulted in continuous measures (i.e., counts, proportions and averages) that were categorized into quartiles based on the distribution in the total sample. Two count variables were assessed – total number of recreational facilities and nonresidential destinations within the buffer. Ordinal items were aggregated as the proportion of street segments within the buffer with a specific attribute (e.g., percent of segments with a bus stop). The physical disorder and street safety summary measures were averaged across all segments within the buffer. The physical disorder score was calculated as the weighted sum of eight audit items: beer/liquor bottles or cans, cigarette/cigar butts or packages, condoms, drug-related paraphernalia, garbage/ litter/broken glass, abandoned cars, graffiti and broken windows.36 The street safety score was calculated as the unweighted sum of seven audit items: number of traffic lanes, connectivity, street design characteristics to reduce volume or speed, traffic calming devices, aggressive drivers (reverse coded), crossing aids and street lighting.

Statistical analysis Analyses were conducted separately for both dependent variables (obese vs normal weight and obese/inactive vs normal weight/active). The association between each perceived and observed neighborhood indicator and obesity was estimated using logistic regression to calculate adjusted prevalence odds ratios (aOR) and 95% confidence intervals (CI). The referent group for all environmental indicators was selected such that positive associations (i.e., aOR41.0) were expected. The regression models were adjusted for demographic characteristics (age, gender and education) associated with obesity.19,21 Categorical age and education variables were dummy-coded. The extended Mantel– Haenszel correlation statistic was used to test for linear trends across ordinal categories and quartiles.37 To examine effect modification by gender and annual household income (o$25 000 and X$25 000), stratified analyses were performed; aORs were considered heterogeneous if the corresponding interaction term was significant at a ¼ 0.10. Two multivariate models were created to identify the most robust indicators of obesity after controlling for other environmental variables and demographic characteristics. Six perceived and six observed indicators were entered into the initial multivariate model: one each from the recreational facilities, land use and aesthetics domains, and three from the transportation domain (i.e., sidewalks, public transit and traffic safety). Using a manual backward elimination procedure, environmental indicators were removed if they did not

Neighborhood indicators of obesity in urban adults TK Boehmer et al

971 Table 2

Distribution of perceived environmental indicators in the total sample and their association with obesity-related outcomes Total sample n ¼ 1032

Obese vs normal weight (n ¼ 656)

%a

aOR b

Obese/inactive vs normal weight/active (n ¼ 348)

95% CI

aOR b

95% CI

There are many places to be physically active in my community, not including streets for walking or jogging Strongly agree 24.1 1.0 F Agree 49.6 1.0 0.6–1.6 Disagree 18.6 1.3 0.7–2.3 Strongly disagree 7.2 1.8* 0.9–3.6

1.0 1.1 1.9 5.0*

F 0.6–2.0 0.8–4.2 1.5–15.9

There is equipment available for physical activity in my community Strongly agree 18.2 Agree 48.7 Disagree 22.5 Strongly disagree 8.0

1.0 0.8 1.3 1.3

F 0.5–1.4 0.7–2.3 0.6–2.6

1.0 0.6 1.2 2.8

F 0.3–1.2 0.5–2.7 0.9–9.4

Number of recreational facilities within 5-min walk from home 3–5 18.8 2 22.6 1 26.1 0 24.8

1.0 0.6 0.9 0.9

F 0.3–1.1 0.5–1.5 0.5–1.5

1.0 0.5 1.2 1.5

F 0.2–1.2 0.5–2.5 0.7–3.5

Distance from home to nearest recreational facility 1–2 min walk 3–5 min walk 6–10 min walk 410 min walk

1.0 1.2 1.2 1.4

F 0.8–1.9 0.7–2.0 0.7–2.5

1.0 1.1 1.7 1.7

F 0.6–2.0 0.7–3.8 0.7–4.0

There are many destinations to go within easy walking distance from my home Strongly agree 41.5 Agree 45.0 Disagree/strongly disagree 13.2

1.0 1.4 1.5

F 0.96–2.2 0.8–2.6

1.0 1.2 1.0

F 0.7–2.1 0.5–2.2

Number of nonresidential destinations within a 5-min walk from home 5–11 22.8 3–4 21.6 1–2 28.4 0 25.0

1.0 1.2 0.7 1.2

F 0.7–2.1 0.4–1.3 0.7–2.1

1.0 1.0 0.9 1.2

F 0.5–2.2 0.4–1.9 0.6–2.8

No destinations within a 10-min walk from home

15.7

2.2

1.3–3.5

2.4

1.1–5.0

There are sidewalks on most of the streets in my community Strongly agree 60.7 Agree 32.9 Disagree/strongly disagree 6.4

1.0 1.0 2.2

F 0.7–1.6 1.1–4.3

1.0 0.7 0.9

F 0.4–1.2 0.3–2.2

It is easy to walk to a bus stop, train or subway station from my home Strongly agree 53.7 Agree 39.8 Disagree/strongly disagree 6.0

1.0 1.2 1.8

F 0.8–1.8 0.9–3.8

1.0 1.0 1.6

F 0.6–1.7 0.5–5.2

F 0.6–1.8 0.5–1.8 0.7–3.2

1.0 1.1 0.9 0.7

F 0.5–2.4 0.4–2.2 0.2–2.2

F 0.96–2.1

1.0 2.4

F 1.3–4.3

Recreational facilities

30.2 35.2 17.5 13.4

Land use

Transportation

How safe from traffic do you feel while walking or riding your bike in your neighborhood? Extremely safe 14.5 1.0 Quite safe 47.2 1.1 Slightly safe 27.8 1.0 Not at all safe 9.9 1.5 Aesthetics How would you rate your community as a place to be physically active? Very pleasant 49.2 Somewhat pleasant 36.0

1.0 1.4

International Journal of Obesity

Neighborhood indicators of obesity in urban adults TK Boehmer et al

972

Table 2 (continued) Total sample n ¼ 1032

Obese vs normal weight (n ¼ 656)

Obese/inactive vs normal weight/active (n ¼ 348)

%a

aOR b

95% CI

aOR b

95% CI

7.4 7.0

2.7 3.1*

1.3–5.6 1.5–6.5

3.0 4.9*

1.01–8.9 1.6–14.9

There are many interesting things to look at while walking in my neighborhood Strongly agree 36.6 Agree 40.1 Disagree 14.9 Strongly disagree 8.0

1.0 1.7 2.3 4.8*

F 1.1–2.7 1.3–4.0 2.4–9.8

1.0 1.9 4.7 6.7*

F 1.02–3.4 1.9–11.7 2.2–19.9

My neighborhood is generally free from garbage, litter or broken glass Strongly agree 34.7 Agree 43.0 Disagree 14.6 Strongly disagree 7.5

1.0 1.2 1.8 1.4

F 0.8–1.8 1.00–3.1 0.7–2.9

1.0 0.9 1.2 1.2

F 0.5–1.7 0.5–2.8 0.4–3.7

My neighborhood is well maintained Strongly agree Agree Disagree Strongly disagree

1.0 1.4 1.5 2.1

F 0.9–2.2 0.8–2.8 0.98–4.4

1.0 1.1 1.1 1.2

F 0.6–1.9 0.5–2.8 0.4–4.1

Not very pleasant Not at all pleasant

38.4 42.3 12.1 6.7

Abbreviations: aOR, adjusted odds ratio; CI, confidence interval. a1032 is denominator for all percentages. Sum does not equal 100% owing to missing data. b Adjusted for gender, age (18–29, 30–39, 40–49, 50–59, 60–69, 70+years) and education (less than high school, high school graduate, some college, college graduate). *Indicates significant linear trend (Po0.05).

attain statistical significance at a ¼ 0.10 or exhibit moderate strength of association between extreme values (aOR41.5). Three phases of modeling resulted in a parsimonious multivariate model for each outcome: (1) perceived indicators only, (2) observed indicators only and (3) combination of indicators retained from phases 1 and 2.

Results The final sample included 1032 respondents after excluding respondents with invalid addresses (n ¼ 5) or missing data (n ¼ 36). The distribution of gender, race, age, educational attainment and weight status is provided for the total sample and by study area (Table 1). Statistically significant differences in these variables were observed across the four study areas. The analytic sample size was 210 obese and 446 normal weight adults for the primary outcome and 143 obese/inactive and 205 normal weight/active adults for the secondary outcome. The distribution of perceived and observed environmental indicators in the total sample is provided in Tables 2 and 3.

Perceived indicators Few perceived measures of recreational facilities were associated with obesity (Table 2). Obese/inactive persons were more likely than normal weight/active persons to strongly disagree to items assessing ‘many places’ and ‘equipment available’ for physical activity; however, the latter association was not statistically significant. Stratified International Journal of Obesity

analyses revealed that the effect of perceived recreational facility access differed by gender in the obese vs normal weight comparison (lack of equipment was stronger among women) and by income level in the obese/inactive vs normal weight/active comparison (lack of equipment and many places were stronger within lower-income strata). Regarding land use, obese and obese/inactive persons were twice as likely to report having no nonresidential destinations within a 10–min walk than their normal weight counterparts. In the transportation domain, obese persons were 2.2 times more likely to disagree that there were sidewalks present on most streets, but no association was present in the obese/inactive comparison. Obesity was not associated with traffic safety. Reduced access to public transit was reported slightly more often among obese and obese/ inactive persons, although this finding was not statistically significant. Between-strata differences were not observed by gender or income within the land use and transportation domains, with one exception (lack of sidewalks was stronger among higher-income individuals in the obese/inactive analysis). Two measures of aesthetics that incorporated physical activity behavior (i.e., ‘rate your community’ and ‘interesting things’) were strongly associated with being obese and obese/inactive (aOR43.0) and exhibited dose–response relationships. The associations between obesity and general perceptions of neighborhood maintenance and garbage were smaller in magnitude and not statistically significant. In the stratified analyses, larger effect sizes for perceived neighborhood aesthetics were observed among women and higherincome individuals.

Neighborhood indicators of obesity in urban adults TK Boehmer et al

973 Table 3

Distribution of observed environmental indicators in the total sample and their association with obesity-related outcomes Total sample n ¼ 1032

Obese vs normal weight (n ¼ 656)

Obese/inactive vs normal weight/Active (n ¼ 348)

%

aORa

95% CI

aOR a

95% CI

19.8 27.2 25.0 28.0

1.0 0.7 0.7 0.5*

F 0.4–1.2 0.4–1.1 0.3–0.9

1.0 0.9 0.7 0.6

F 0.4–2.0 0.3–1.4 0.3–1.2

23.6 22.6 27.7 26.2

1.0 1.2 1.1 0.8

F 0.7–2.1 0.7–1.8 0.5–1.4

1.0 1.8 1.7 1.4

F 0.8–3.8 0.6–5.1 0.6–3.0

Segments with sidewalks with no or a little unevenness (%) 83–100 25.9 64–82 23.1 50–63 24.4 0–49 26.6

1.0 1.4 1.3 2.1*

F 0.8–2.3 0.7–2.2 1.3–3.6

1.0 1.3 0.7 1.5

F 0.6–2.6 0.3–1.5 0.7–3.1

Segments with a bus or other transit stop (%) 25–53 21.0 19–24 24.8 14–18 25.9 0–13 28.3

1.0 1.0 1.3 0.8

F 0.6–1.7 0.8–2.1 0.5–1.4

1.0 2.0 1.3 0.9

F 0.9–4.4 0.6–2.8 0.4–1.9

Average street safety scoreb 18.1–19.2 (most safe) 17.9–18.1 17.6–17.9 16.1–17.6 (least safe)

24.3 25.4 24.5 25.8

1.0 0.9 0.7 0.9

F 0.6–1.5 0.4–1.1 0.5–1.4

1.0 1.4 0.6 0.8

F 0.7–2.9 0.3–1.3 0.4–1.5

25.5 24.9 25.0 24.6

1.0 1.6 2.7 4.0*

F 0.9–2.7 1.6–4.8 2.3–7.0

1.0 1.9 1.9 3.3*

F 0.9–4.0 0.9–4.3 1.5–7.4

Segments with some or a lot of garbage, litter or broken glass (%) 0–5 22.4 6–15 27.7 15–49 24.8 50 or more 25.1

1.0 1.0 1.9 3.7*

F 0.6–1.8 1.1–3.3 2.1–6.6

1.0 0.8 1.5 3.0*

F 0.4–1.8 0.7–3.2 1.3–6.9

Segments with some or a lot of attractive features (%) 25–50 23.2 17–24 25.7 8–16 25.8 0–7 25.4

1.0 0.6 0.7 0.8

F 0.4–1.04 0.4–1.2 0.5–1.3

1.0 0.6 0.8 1.3

F 0.3–1.2 0.4–1.8 0.6–2.7

Segments with some or a lot of comfort features (%) 26–60 24.9 13–25 23.7 3–12 26.0 0–2 25.4

1.0 0.6 0.8 0.8

F 0.4–1.04 0.5–1.3 0.5–1.3

1.0 0.9 1.4 1.5

F 0.4–1.9 0.7–2.8 0.7–3.1

Recreational facilities Number of recreational facilities 6–17 4–5 2–3 0–1 Land use Number of nonresidential destinations 42–131 23–42 11–22 0–10 Transportation

Aesthetics Average physical disorder summary scorec 0.0–1.2 (least disorder) 1.2–2.8 2.8–12.8 12.8–22.6 (most disorder)

Abbreviations: aOR, adjusted odds ratio; CI, confidence interval. aAdjusted for gender, age (18–29, 30–39, 40–49, 50–59, 60–69, 70+years) and education (less than high school, high school graduate, some college, college graduate). bUnweighted sum of seven audit items where none ¼ 1, a little ¼ 2, some ¼ 3 and a lot ¼ 4. c Weighted sum of eight audit items where none ¼ 0, a few ¼ 2, some ¼ 5 and a lot ¼ 9. *Indicates significant linear trend (Po0.05).

International Journal of Obesity

Neighborhood indicators of obesity in urban adults TK Boehmer et al

974 Observed indicators Neither the number of recreational facilities nor the number of nonresidential destinations observed within the 400-m buffer was associated with obesity in the hypothesized direction (Table 3). Contrary to expectations, having fewer recreational facilities within close proximity was related to lower odds of obesity; the stratified analysis revealed that this inverse association was present among women, but not men. Regarding transportation infrastructure, poor sidewalk condition was related to being obese (particularly within the lower-income strata), but not to being obese/inactive. No association was observed between obesity and access to public transit or street safety score; however, an inverse association between low street safety score and obesity was observed among women. Aesthetic indicators of physical disorder and presence of garbage exhibited significant linear trends across quartiles and were strongly associated with increased odds of being obese and obese/inactive (aOR ¼ 3.0–4.0 when comparing extreme values), particularly among women.

Multivariate models Both perceived and observed indicators of aesthetics (i.e., interesting things and physical disorder) and land use Table 4

(i.e., no or few nonresidential destinations within walking distance) were retained in the final multivariate models for both dependent variables as the most robust neighborhood correlates of obesity (Table 4). Perceived access to recreational facilities was also retained in the obese/inactive model. Interestingly, the observed land-use indicator of nonresidential destinations showed a dose-response association with both obesity-related outcomes after controlling for physical disorder in the multivariate models.

Discussion By simultaneously examining perceived and observed environmental measures, this study contributes substantially to the growing evidence base of community-level correlates of obesity and suggests salient environmental and policy intervention strategies that may reduce population-level obesity prevalence. This study is among the first to examine and document inconsistencies in the degree to which perceived and observed measures of the neighborhood environment are associated with obesity. Although the generalizability of findings from two US cities may be limited, international researchers may explore related

Final multivariate models of perceived and observed environmental indicators with obesity-related outcomes Obese vs normal weight (n ¼ 639) n

aOR

a

95% CI

Obese/inactive vs normal weight/active (n ¼ 339) n

aORa

95% CI

77 174 65 23

1.0 0.8 1.0 2.4

F 0.4–1.7 0.4–2.5 0.6–8.8

Perceived (survey) indicators Many places to be physically active Strongly agree Agree Disagree Strongly disagree

Not retained

No destinations within a 10-min walk from home

104

1.8

1.1–3.1

54

1.8

0.8–4.0

Many interesting things to look at Strongly agree Agree Disagree Strongly disagree

235 255 101 48

1.0 1.6 1.7 2.6

F 1.04–2.6 0.96–3.2 1.2–5.6

131 135 46 27

1.0 2.0 3.4 3.2

F 1.1–3.7 1.3–9.0 0.9–10.9

Number of nonresidential destinations 42–131 23–42 11–22 0–10

166 143 165 165

1.0 1.2 1.5 1.7

F 0.7–2.1 0.8–2.6 0.9–3.2

89 74 89 87

1.0 1.3 1.9 2.0

F 0.6–3.0 0.8–4.4 0.8–5.1

Average physical disorder summary scoreb 0.0–1.2 (least disorder) 1.2–2.8 2.8–12.8 12.8–22.6 (most disorder)

162 154 160 163

1.0 1.4 3.0 3.6

F 0.8–2.5 1.6–5.7 1.8–7.2

76 82 93 88

1.0 1.7 2.1 2.8

F 0.8–3.9 0.8–5.3 1.02–7.7

Observed (audit) indicators

Abbreviations: aOR, adjusted odds ratio; CI, confidence interval. aAdjusted for gender, age (18–29, 30–39, 40–49, 50–59, 60–69, 70+years) and education (less than high school, high school graduate, some college, college graduate) and all other environmental indicators shown in Table 4. bWeighted sum of eight audit items where none ¼ 0, a few ¼ 2, some ¼ 5 and a lot ¼ 9.

International Journal of Obesity

Neighborhood indicators of obesity in urban adults TK Boehmer et al

975 indicators and adapt the methodology described herein to investigate various urban environments. Contrary to previous studies in Missouri, US and Perth, Australia,19,21 the current study found that perceived indicators of recreational facility access were not associated with obesity, and the observed number of recreational facilities was inversely associated with obesity compared to a priori expectations. Post hoc analyses determined that the inverse association between recreational facilities and obesity could not be explained by other observed neighborhood characteristics (i.e., number of nonresidential destinations, integration of land uses and amount of physical disorder). Similar findings were reported previously from this study population, in which there was little to no association between observed recreational facility availability and meeting physical activity recommendations.12 Although it appears that perceptions of recreational facility access are more strongly associated with physical activity behavior and obesity,9,19,21 future measurement of actual access should consider the type, cost and condition of nearby facilities, not simply the quantity.12 Regarding land-use patterns, the current study found that fewer nonresidential destinations (i.e., less land-use mix) within a 400-m buffer moderately increased the odds of obesity after controlling for physical disorder in the multivariate model. That is, the ‘true’ association between obesity and land use was masked by the positive correlation between the number of nonresidential destinations and physical disorder (r ¼ 0.53). Using a larger (1-km) buffer, Frank et al.17 reported a stronger association between land-use mixture and obesity than was observed in the current study. Furthermore, perceiving no destinations within a 10-min walk was a robust correlate of obesity in the multivariate model, but perceiving no destinations within a 5-min walk was not. These findings suggest that larger neighborhood boundaries may more accurately reflect the influence of land use on obesity. With respect to transportation infrastructure, the current study corroborates previous findings of an association between obesity and reduced access to sidewalks (both perceived and observed);19,21 but neither perceived sidewalk access nor observed sidewalk quality were retained in the multivariate models. Within the selected study areas, availability of public transit (i.e., number of bus stops) was not associated with obesity; however, street-level audit data does not adequately detect potentially important indicators of public transit utilization (e.g., routes, fares and operation times). Little empirical research exists regarding specific neighborhood aesthetic qualities and obesity.19 Both perceived and observed aesthetic indicators were strong, robust correlates of obesity in the multivariate model. The aesthetic indicators examined in this study reflect urban design qualities of imageability (i.e., qualities that make a place distinct, recognizable and memorable) and tidiness (i.e., condition and cleanliness of a place) shown to be associated with

walkability.38 This study did not include a comprehensive evaluation of urban design qualities, making it difficult to identify specifically which features make some streets more aesthetically pleasing and walkable (and perhaps less obesogenic) than others. Post hoc analyses indicated that the association between obesity and aesthetic indicators (e.g., pleasantness, physical disorder, garbage) was almost entirely explained by poverty level when study area was included as a control variable in the logistic regression models. In fact, the physical disorder summary score was a nearly perfect indicator of study area poverty status in this study; that is, 100% of respondents in the lower income areas were above the median physical disorder score (indicating greater disorder) and 92% of respondents in the higher income areas were below the median (indicating less disorder). This correlation between physical disorder and poverty was also one of the chief explanations for the strong inverse association between physical disorder and transportation-related activity that was observed previously, in that people from higher poverty areas with more physical disorder were more likely to walk and bicycle for transportation.12 Thus, it appears that insufficient physical activity may not be the only mediating factor in the relationship between poverty and obesity, and engaging in transportation-based activity alone may be insufficient for minimizing obesity. In general, findings from the obese/inactive vs normal weight/active comparison mirrored those from the primary analysis with few exceptions. Stronger associations emerged for perceived recreational facilities and aesthetic measures including ‘physical activity’ or ‘walking’ in the item wording, suggesting potential reporting bias based on physical activity level. Slightly weaker associations were detected for perceived sidewalks and traffic safety and observed sidewalk quality, physical disorder and garbage in the secondary analysis. Indicators from the land-use and aesthetic domains were retained in the multivariate analysis, and perceived and observed indicators from these domains demonstrated similar strength of association in their relationship with obesity. Thus, both measurement methods provided robust neighborhood indicators of obesity despite previous findings of poor to fair agreement between the survey and audit measures.22 Future investigations should utilize both types of neighborhood assessments to help determine whether one methodology is more valid. Additional analyses were performed to examine overweight vs normal weight adults, as was carried out in one previous study.21 Elevated (and significant) ORs were observed for perceived and observed measures of aesthetics, but they were weaker in magnitude compared with obese vs normal weight. There was no relationship between overweight and environmental indicators (perceived or observed) from the recreational facilities, land use and transportation domains. Future research should examine whether there is a gradient of environmental influence on overweight and obesity or unique correlates for each outcome. International Journal of Obesity

Neighborhood indicators of obesity in urban adults TK Boehmer et al

976 This study is subject to at least five limitations. First, crosssectional data prohibit conclusions regarding a causal relationship between the neighborhood environment and obesity. In particular, the degree of bias owing to residential preference or self-selection is unknown. Second, sub-analyses with reduced sample size (i.e., secondary outcome and stratified analysis) lacked sufficient power to detect weak associations. Third, misclassification of subjects based on outcome status may have occurred owing to reliance on self-reported weight and physical activity data. Fourth, respondents who lived near the study area boundaries had incomplete buffers owing to nonaudited street segments. Lastly, the small number of study areas restricted generalizability and prohibited the use of multilevel modeling to examine the contextual effect of the neighborhood environment while controlling for individual-level risk factors.

Conclusion As proposed in the social–ecological framework, physical activity and dietary behaviors mediate the relationship between the physical environment and obesity.6,7 This study identified neighborhood characteristics that either directly influence obesity or, more likely, act indirectly through physical activity. Additional research is needed to explore direct and indirect effects and evaluate the joint impact of the food and physical activity environments on energy balance. An ecological perspective also emphasizes the importance of multilevel interventions that integrate micro- and macrolevel approaches.39 Combined with previous research, this study suggests that environmental and policy strategies that promote mixed land use and improve aesthetic quality may advance the development of active living communities and reduce the prevalence of obesity on a population level.40 Although several studies have associated neighborhood deprivation with higher obesity prevalence,13–15 the underlying mechanisms through which neighborhood disadvantage affects health remain unclear.41 This study was unable to separate the physical environment from the social context because obesogenic neighborhood features (e.g., sidewalk quality, physical disorder and land use) coexisted with each other and with neighborhood poverty level. To understand the relationship between physical and social determinants of health and their relative contribution to obesity, future studies should evaluate a larger number of study areas representing diverse social, economic and environmental conditions.

Acknowledgements We are grateful for the assistance of Rebeka Cook, Montenia Anderson, Cheryl Kelly, Kathy McMullen, Omar Brown, International Journal of Obesity

Dwayne Ellis, Glen Grant, Timothy Hill and Albert Spears in data collection, and of Michael Elliott in data management. We also thank C Tracy Orleans and Marla Hollander of the Robert Wood Johnson Foundation for their support throughout this project. This study was funded through The Robert Wood Johnson Foundation contract no. 051603, including support from the Centers for Disease Control and Prevention contract U48/DP000060-01 (Prevention Research Centers Program).

References 1 US Department of Health and Human Services. The surgeon general’s call to action to prevent and decrease overweight and obesity. US Department of Health and Human Services, Public Health Service, Office of the Surgeon General: Rockville, MD, 2001. 2 World Health Organization. Obesity: Preventing and managing the global epidemic Report of a WHO consultation World Health Organ Tech Rep Ser 2000; 894: 1–253. 3 National Institutes of Health. Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults. The evidence report: National Heart, Lung, Blood Institute. NIH Publication No. 98-4083; 1998. 4 Hill JO, Peters JC. Environmental contributions to the obesity epidemic. Science 1998; 280: 1371–1374. 5 Poston WS, Foreyt JP. Obesity is an environmental issue. Atherosclerosis 1999; 146: 201–209. 6 Egger G, Swinburn B. An ‘ecological’ approach to the obesity pandemic. BMJ 1997; 315: 477–480. 7 Swinburn B, Egger G, Raza F. Dissecting obesogenic environments: the development and application of a framework for identifying and prioritizing environmental interventions for obesity. Prev Med 1999; 29: 563–570. 8 Franklin BA. The downside of our technological revolution? an obesity-conducive environment. Am J Cardiol 2001; 87: 1093–1095. 9 Humpel N, Owen N, Leslie E. Environmental factors associated with adults’ participation in physical activity: a review. Am J Prev Med 2002; 22: 188–199. 10 Saelens BE, Sallis JF, Frank LD. Environmental correlates of walking and cycling: findings from the transportation, urban design and planning literatures. Ann Behav Med 2003; 25: 80–91. 11 Brownson RC, Baker EA, Housemann RA, Brennan LK, Bacak SJ. Environmental and policy determinants of physical activity in the United States. Am J Public Health 2001; 91: 1995–2003. 12 Hoehner CM, Brennan Ramirez LK, Elliott MB, Handy SL, Brownson RC. Perceived and objective environmental measures and physical activity among urban adults. Am J Prev Med 2005; 28: 105–116. 13 Ellaway A, Anderson A, Macintyre S. Does area of residence affect body size and shape? Int J Obes Relat Metab Disord 1997; 21: 304–308. 14 van Lenthe FJ, Mackenbach JP. Neighbourhood deprivation and overweight: the GLOBE study. Int J Obes Relat Metab Disord 2002; 26: 234–240. 15 Sundquist J, Malmstrom M, Johansson SE. Cardiovascular risk factors and the neighbourhood environment: a multilevel analysis. Int J Epidemiol 1999; 28: 841–845. 16 Ewing R, Schmid T, Killingsworth R, Zlot A, Raudenbush S. Relationship between urban sprawl and physical activity, obesity, and morbidity. Am J Health Promot 2003; 18: 47–57. 17 Frank LD, Andresen MA, Schmid TL. Obesity relationships with community design, physical activity, and time spent in cars. Am J Prev Med 2004; 27: 87–96.

Neighborhood indicators of obesity in urban adults TK Boehmer et al

977 18 Saelens BE, Sallis JF, Black JB, Chen D. Neighborhood-based differences in physical activity: an environment scale evaluation. Am J Public Health 2003; 93: 1552–1558. 19 Catlin TK, Simoes EJ, Brownson RC. Environmental and policy factors associated with overweight among adults in Missouri. Am J Health Promot 2003; 17: 249–258. 20 Boehmer TK, Lovegreen SL, Haire-Joshu D, Brownson RC. What constitutes an obesogenic environment in rural communities? Am J Health Promot 2006; 20: 411–421. 21 Giles-Corti B, Macintyre S, Clarkson JP, Pikora T, Donovan RJ. Environmental and lifestyle factors associated with overweight and obesity in Perth, Australia. Am J Health Promot 2003; 18: 93–102. 22 Boehmer TK, Hoehner CM, Wyrwich KW, Brennan Ramirez LK, Brownson RC. Correspondence between perceived and observed measures of neighborhood environmental supports for physical activity. J Physical Activity Health 2006; 3: 22–36. 23 Blair SN, Brodney S. Effects of physical inactivity and obesity on morbidity and mortality: current evidence and research issues. Med Sci Sports Exerc 1999; 31: S646–S662. 24 Lee CD, Blair SN, Jackson AS. Cardiorespiratory fitness, body composition, and all-cause and cardiovascular disease mortality in men. Am J Clin Nutr 1999; 69: 373–380. 25 Stevens J, Cai J, Evenson KR, Thomas R. Fitness and fatness as predictors of mortality from all causes and from cardiovascular disease in men and women in the Lipid Research Clinics Study. Am J Epidemiol 2002; 156: 832–841. 26 Farrell SW, Braun L, Barlow CE, Cheng YJ, Blair SN. The relation of body mass index, cardiorespiratory fitness, and all-cause mortality in women. Obes Res 2002; 10: 417–423. 27 Kirtland K, Porter D, Addy C, Neet MJ, Williams JE, Sharpe PA et al. Environmental measures of physical activity supports: perception versus reality. Am J Prev Med 2003; 24: 323–331. 28 Brownson RC, Eyler AA, King AC, Shyu Y-L, Brown DR, Homan SM. Reliability of information on physical activity and other chronic disease risk factors among US women aged 40 years or older. Am J Epidemiol 1999; 149: 379–391. 29 Brennan LK, Baker EA, Haire-Joshu D, Brownson RC. Linking perceptions of the community to behavior: are protective social factors associated with physical activity? Health Educ Behav 2003; 30: 740–755.

30 Brownson RC, Chang JJ, Eyler AA, Ainsworth BE, Kirtland KA, Saelens BE et al. Measuring the environment for friendliness toward physical activity: a comparison of the reliability of three questionnaires. Am J Public Health 2004; 94: 473–483. 31 Craig CL, Marshall AL, Sjostrom M, Bauman AE, Booth ML, Ainsworth BE et al. International physical activity questionnaire: 12-country reliability and validity. Med Sci Sports Exerc 2003; 35: 1381–1395. 32 Brownson RC, Hoehner CM, Brennan LK, Cook RA, Elliott MB, McMullen KM. Reliability of two instruments for auditing the environment for physical activity. J Physical Activity Health 2004; 1: 189–207. 33 Pate RR, Pratt M, Blair SN, Haskell WL, Macera CA, Bouchard C et al. Physical activity and public health: a recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA 1995; 273: 402–407. 34 Flegal KM, Graubard BI, Williamson DF, Gail MH. Excess deaths associated with underweight, overweight, and obesity. JAMA 2005; 293: 1861–1867. 35 Wellens RI, Roche AF, Khamis HJ, Jackson AS, Pollock ML, Siervogel RM. Relationships between the Body Mass Index and body composition. Obes Res 1996; 4: 35–44. 36 Raudenbush SW, Sampson RJ. Ecometrics: toward a science of assessing ecological settings, with application to the systematic social observation of neighborhoods. Sociol Methodol 1999; 29: 1–41. 37 Mantel N. Chi-square tests with one degree of freedom: extensions of the Mantel–Haenszel procedure. J Am Stat Assoc 1963; 58: 690–700. 38 Ewing R, Handy S, Brownson RC, Clemente O, Winston E. Identifying and measuring urban design qualities related to walkability. J Physical Activity Health 2006; 3: S223–S240. 39 McLeroy KR, Bibeau D, Steckler A, Glanz K. An ecological perspective on health promotion programs. Health Educ Q 1988; 15: 351–377. 40 Killingsworth RE. Health promoting community design: a new paradigm to promote healthy and active communities. Am J Health Promot 2003; 17: 169–170. 41 Macintyre S, Ellaway A, Cummins S. Place effects on health: how can we conceptualise, operationalise and measure them? Soc Sci Med 2002; 55: 125–139.

International Journal of Obesity