Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Am J Public Health. Author manuscript; available in PMC 2011 December 7.
Published in final edited form as:
PMCID: PMC3097303

Association of Walkability with Obesity in Baltimore City, Maryland



To investigate the association between walkability and obesity stratified by neighborhood race and socioeconomic status (SES) among adults residing in Baltimore City.


We conducted a cross-sectional study among 3,493 participants from the Healthy Aging in Neighborhoods of Diversity across the Life-Span (HANDLS) study. The Pedestrian Environment Data Scan (PEDS) was implemented to measure neighborhood walkability in 34 neighborhoods of varying racial and socioeconomic composition in which participants resided. Walkability was determined using confirmatory factor analysis. Prevalence ratios were calculated for the association between walkability and obesity using multilevel modeling.


Among individuals living in predominately white and high-SES neighborhoods, residing in highly walkable neighborhoods was associated with a lower prevalence of obesity compared to individuals living in poorly walkable neighborhoods after adjusting for individual-level demographic variables (PR=0.58, p=<0.001; PR=0.80, p=0.004, respectively); prevalence ratios were similar after controlling for potential mediators. The association between walkability and obesity among individuals living in low-SES neighborhoods was not significant after accounting for main mode of transportation (PR=0.85, p=0.060).


Among some neighborhoods, high walkability was associated with less obesity.

Keywords: Walkability, Obesity, Neighborhood, Socioeconomic Status, Racial Composition


The United States currently faces an epidemic of obesity with roughly 34% of adults being obese (1). This condition increases the risk for many chronic diseases including cardiovascular disease, diabetes and some cancers (13). Furthermore, non-Hispanic blacks and less educated individuals are more likely to be obese compared to non-Hispanic whites and more educated individuals (2, 3).

Given the high prevalence of obesity, recent research has focused on the role the built environment plays in influencing individual physical activity, modes of transportation and health outcomes. Despite racial and socioeconomic disparities, few studies have addressed the association between the built environment and obesity across neighborhoods of varying racial and socioeconomic composition. Rather, associations have been documented among varying populations and geographic locations without regard for contextual neighborhood factors.

Numerous features of the built environment have been associated with physical activity, a protective health behavior for obesity, including residential density, land-use mix, urban sprawl, intersection density, walkability, park availability and accessibility to physical activity-related resources (613). One study found that higher SES neighborhoods had an increased likelihood of having one or more physical activity facilities; more facilities were also associated with an increased likelihood of achieving moderate-vigorous physical activity (4). Land-use mix, nearby destinations, and the presence of sidewalks have been associated with less obesity (1418) but little research has investigated differences stratified by neighborhood characteristics. Furthermore, few studies have developed measurement models for walkability; the use of a composite neighborhood walkability score would reduce the likelihood of finding associations by chance alone (i.e., Type I error) (5, 6).

To investigate the association between neighborhood walkability and obesity, an environmental audit that measures the micro-scale features of the pedestrian environment was implemented in several Baltimore City neighborhoods (7). It was hypothesized that individuals living in neighborhoods with higher walkability would have a lower prevalence of obesity than individuals living in neighborhoods with lower walkability. A secondary hypothesis was that these associations would differ by neighborhood race and socioeconomic composition.


Parent Study

The Healthy Aging in Neighborhoods of Diversity Across the Life Span (HANDLS) study is a multidisciplinary, prospective epidemiologic study set in Baltimore City and examines the influence and interaction of race and socioeconomic status (SES) on the development of cardiovascular health disparities among urban-dwelling minority and lower SES groups. The detail of the study is presented elsewhere (8). The HANDLS study design was stratified across four factors: age, sex, race, and SES. Baseline recruitment included 3,493 black and white adults aged 30–64 of high- and low-SES living in 12 neighborhoods across Baltimore City. Each neighborhood consisted of 2–5 census tracts. Data collection was implemented in two stages: (1) an in-home household survey and (2) a physical examination and medical history in mobile research vehicles (MRV). Survey and medical information is confidential and approved by the Medstar Research Institute.

Study Population of the Neighborhood Ancillary Study

HANDLS participants were selected from 12 pre-defined Baltimore City neighborhoods that were likely to meet the age, race, sex and SES design specifications. Recruitment and sampling contractors produced household listings to identify residential dwellings in each neighborhood. The contractors performed doorstep interviews, identified eligible persons in each household, selected one of two eligible persons per household and invited the eligible candidates to participate in HANDLS. Inclusion criteria for participants included the ability to give informed consent, age 30–64, the ability to perform at least five measures and able to present a valid picture identification. Exclusion criteria included pregnancy, being within six months of active cancer treatment and multi-ethnic individuals who did not identify strongly with either the black or white race.

Individual-level Household Interview Measures

Individual-level demographic measures from the HANDLS in-home questionnaire included self-reported age, sex, race, education, household income and general health status. Low-SES was defined as having a household income below 125% the poverty threshold. General health status was measured using the Short Form Health Survey (SF-12) (21). Perceived crime was evaluated with three 5-point Likert scale questions on how common serious crime, drug and gang activity was in their neighborhood. Participants also reported on the main mode of transportation used for traveling outside of their neighborhood from the following options: car, someone else’s car, public transportation, walking.

Neighborhood Census Measures

Neighborhood race and SES were determined using data from the 2000 U.S Census. Neighborhoods were classified as predominately black or white if ≥60% of the residents were black or white, respectively (9). Since only 3 tracts failed to meet these criteria and the racial composition included few (<2%) non-blacks or non-whites, these racially mixed tracts were classified by the racial majority. Neighborhood SES was determined by the percentage of individuals in the tract living in poverty. Census tracts with ≥25% of residents below the poverty threshold were categorized as low-SES and <25% living below the poverty threshold as high-SES. These cut-points were determined based on median values for percent poverty in the census tract.

Neighborhood Walkability

The Pedestrian Environment Data Scan (PEDS) was implemented to evaluate the pedestrian walking environment for exercise and transportation (7). The audit captures micro-scale features that are frequently apparent to pedestrians but not easily captured in publicly available data and has been conceptualized into four sections: uses and design, pedestrian facilities, road attributes and the walking/cycling environment. The PEDS audit has demonstrated moderate to high reliability for most items and was comparable in reliability to other environmental audits (kappa statistics >0.70) (7).

To implement the PEDS audit in neighborhoods of HANDLS participants, pairs of trained raters systematically assessed a 20% random sample of street segments. Prior literature indicates that a 20% sample is an appropriate selection size for measuring physical features in a neighborhood (10). In addition, roughly 30% of street segments in HANDLS neighborhoods were alley streets. Therefore, a 5% random sample of alley streets was taken from each neighborhood. These segments were sampled separately to increase the likelihood of capturing segments most often traveled by pedestrians; alleys were determined less ideal for walking, physical activity and for safety reasons. Street segments were selected using GIS technology and centerline files from the U.S. Census. Segments inaccessible to pedestrians, such as limited-access highways and ramps, were not included for sample selection. Auditors worked in pairs and discrepancies were reconciled on-site. A 5% reliability sample was conducted for each neighborhood where pairs of raters assessed the same segments for quality control purposes; Kappa statistics were high, ranging from 0.75 to 0.99 for most PEDS items.

Health Behaviors and Outcomes

Medical staff measured height and weight using standard measurement tools during the MRV participant visit. Obesity was defined as a body mass index (BMI) equal to or greater than 30 kg/m2 (11). Physical activity was self-reported using the Houston Physical Activity Scale for a subset of HANDLS participants (n=717) (12).

Statistical Analysis

Confirmatory Factor Analysis

To construct a walkability score with data from the PEDS audit, confirmatory factor analysis (CFA) was conducted in MPlus (13). A one-factor model for both categorical and continuous dependent variables was chosen with a weighted least-squares mean and variance estimator. After reviewing Spearman and Pearson correlation matrices for PEDS items, seven items were included in the initial model (type of intersection, obstructions in the sidewalk, connections to other sidewalks and crosswalks, stop signs, absence of traffic control devices, crosswalks and absence of amenities). These were chosen based on both empirical (inter-item correlations ~≥0.40) and theoretical evidence from previous literature (17, 26). For each iteration, the adequacy of the model fit was evaluated using the following statistics to assess the degree of fit between the estimated and observed variance/covariance matrix: chi-square test, the relative likelihood ratio (X2/df), the comparative fit index (CFI), weighted root mean square residual (WRMR) and the Tucker-Lewis Index (TFI). Based on the final model, factors scores for walkability were determined for each HANDLS neighborhood.

Descriptive Analysis

Participant characteristics were presented stratified by tertiles of neighborhood walkability. The percentage of obese individuals for each walkability tertile was calculated using one-way analysis of variance and stratified by neighborhood race and SES.

Regression Analysis

Prevalence ratios were estimated using multilevel (random-effects) log-binomial models with a random intercept for each neighborhood. The main exposure variable was walkability, a latent construct comprised of the PEDS variables identified by CFA. The main dependent variable was obesity.

All regression models were adjusted for individual-level variables: age, sex, race, education, poverty status and self-reported health and stratified by neighborhood race and SES. To investigate possible mediation pathways, the perception of crime, physical activity and main mode of transportation were assessed in independent regression models; adjustment for all three potential mediators in the same model was also investigated. All analyses were conducted using STATA; the gllamm procedure (14). Participants with missing data for primary outcomes were excluded from analyses and evaluated for exclusion bias.


Confirmatory Factor Analysis

The final model included four indicators: connections to other sidewalks and crosswalks, the presence of stop signs, obstructions in the sidewalk, and designated crosswalks (Figure 1). All estimates were significant (p<0.05) with loadings ranging in magnitude from 0.480 to 0.895. The Chi-square value was 2.31 with 2 degrees of freedom (p=0.316). This yielded a relative likelihood ratio (X2/df) of 1.16 where values less than 3 indicate good fit. In addition, the CLI and TFI were high (0.979 and 0.969, respectively) and the weighted root mean square residual (WRMR) was less than 1 (WRMR=0.278). Walkability factor scores for HANDLS neighborhoods ranged from −0.809 to 0.752 with a mean score of zero. Overall, the CFA results supported the hypothesized model for walkability.

Figure 1
Standardized estimates (SE) for indicators of walkability

Individual-level Characteristics by Neighborhood Walkability

Significantly more blacks resided in medium walkability neighborhoods (86%) with the majority of whites residing in low walkability neighborhoods (65%) (p<0.001) (Table 1). Individuals above the poverty threshold were significantly more likely to reside in low walkability neighborhoods (p<0.001). Reporting the use of a car as a main mode of transportation was significantly more frequent among those that lived in low walkability neighborhoods (p<0.001). Of the total participant population, 43% were obese and 29% were overweight. A higher percentage of obese participants resided in low walkability neighborhoods (45%) compared to high walkability neighborhoods (38%) (p=0.004). Significantly more individuals in predominately black neighborhoods resided in medium walkability neighborhoods while more individuals in low-SES neighborhoods resided in high walkability neighborhoods (p<0.001). There was no difference in age, sex, health insurance, education or self-reported health by neighborhood walkability (p>0.05).

Table 1
Characteristics (%) of HANLDS study participants stratified by neighborhood walkability (n=3493)

Obesity by Neighborhood Walkability

Among individuals residing in predominately white neighborhoods, fewer obese participants lived in high walkability neighborhoods (p<0.001) (Table 2). A similar significant association was shown for individuals living in high-SES neighborhoods (p=0.001). There was no significant association between obesity and neighborhood walkability for individuals residing in predominately black and low-SES neighborhoods (p>0.05).

Table 2
Body mass index (BMI, kg/m2)1 of HANDLS participants by neighborhood walkability (n=2616)

Neighborhood Walkability and the Association with Obesity

Overall, there was no significant association between neighborhood walkability and obesity after adjustment for demographic characteristics (Table 3). Among individuals living in predominately white neighborhoods, residing in a high walkability neighborhood (highest tertile of walkability) was associated with a significantly lower prevalence of obesity compared to individuals living in neighborhoods with poor walkability (lowest tertile) (PR=0.58, p<0.001). A similar association for obesity was found among individuals residing in high- and low-SES neighborhoods (PR=0.80, p=0.004; PR=0.83, p=0.046, respectively). There was no significant association among individuals residing in predominately black neighborhoods.

Table 3
Adjusted associations between walkability and obesity (PR [95% CI])

For individuals residing in low-SES neighborhoods, the association between walkability and obesity became insignificant after additionally adjusting for main mode of transportation (PR=0.85, p=0.060); in a similar approach, independently adjusting for the perception of crime did not alter significance findings for the association between walkability and obesity in these neighborhoods. However, controlling for perceived crime and/or main mode of transportation did not significantly attenuate prevalence ratio estimates and associations remained significant for individuals residing in predominately white or high-SES neighborhoods.

Among individuals residing in low-SES neighborhoods, the association between walkability and obesity was significantly attenuated after controlling for physical activity (PR=1.06, p=0.753) in this small subset of participants (n=281). Conversely, controlling for physical activity did not significantly alter prevalence ratio estimates among individuals residing in predominately white or high-SES neighborhoods.


Previous literature suggests that several attributes of walkability are associated with physical activity and obesity (618) but few studies have examined these associations by neighborhood race and SES (4). The main findings from the current study indicated that among individuals residing in predominately white or high-SES neighborhoods, a highly walkable neighborhood was associated with lower obesity compared to individuals living in poorly walkable neighborhoods after controlling for demographic variables and investigating possible intermediate variables.

There are three key explanations for these findings. First, national data has shown that whites report more leisure-time physical activity compared to blacks (15). Thus, individuals living in predominately white neighborhoods may be more likely to observe or socialize with active neighbors and, consequently, be more likely to engage in physical activity in an effort to maintain or lose weight. Prior research has indicated that observing the exercise habits of peers and neighbors may be beneficial for improving individual physical activity behaviors (29–31). In these neighborhoods, a walkable neighborhood environment may promote and increase the likelihood of activity and, subsequently, lower obesity. In contrast, there was no association between neighborhood walkability and obesity among individuals residing in predominately black neighborhoods. This may be due to few individuals inclined to engage in activity, regardless of the environment. One reason for less activity may be concern for neighborhood safety (29, 32, 33). A study conducted in Los Angeles and Louisiana determined that blacks more often perceived their neighborhood as unsafe and that this neighborhood perception was most strongly associated with less frequent utilitarian walking (16). Among HANDLS participants, the perception of crime was more often reported among individuals residing in highly walkable, predominately black neighborhoods, which suggests that crime may negate any effect of a walkable neighborhood.

Second, individuals living in high-SES neighborhoods may utilize cars more often for daily transportation than individuals living in low-SES neighborhoods. Indeed, 64% of individuals residing in high-SES neighborhoods reported using a car as their main mode of transportation compared to 36% of individuals in low-SES neighborhoods. Therefore, high-SES neighborhoods that are conducive to walking for transportation and physical activity may facilitate more activity and, subsequently, lower obesity. The measures used to capture this activity may not have been sensitive enough. In contrast, individuals living in low-SES neighborhoods may walk for transportation out of necessity. Indeed, there was no significant association between walkability and obesity after controlling for mode of transportation. Fifty-one percent of individuals in low-SES neighborhoods reported either walking or using public transportation most often and those who reported walking or using public transportation had significantly lower BMI compared to car users.

Third, there may be other population-level factors that influence obesity but were not accounted for in these analyses. For example, variation between neighborhoods in the availability of healthy food may impact obesity status in this population. Previous work conducted in Baltimore found that predominately black and lower-income neighborhoods and the supermarkets located within had significantly lower healthy food availability compared to predominately white and higher-income neighborhoods (17). In addition, lower availability of healthy foods was associated with a lower quality dietary pattern; the association was insignificant after adjusting for race (18). Given increases in total energy consumption (19) and the low prevalence of physical activity among adults (20), it is important to understand how neighborhood walkability and healthy food availability interact and influence obesity in neighborhoods of varying characteristics.

Although few studies have examined associations between walkability and obesity stratified by neighborhood characteristics, the literature has been consistent in that, even after controlling for individual-level SES, living in an economically deprived neighborhood increases the likelihood of being obese or having a high BMI (35–37). Similar associations for neighborhood race have been documented (21), although the literature has been less consistent. Two studies found no association between neighborhood race and obesity (37, 38) while one study found that neighborhood racial isolation was significantly associated with obesity among black residents only (22). This research suggests that the effects of neighborhood-level race and SES may be at least partially mediated through differential access to health-promoting or health-constraining environments and resources (23). Indeed, previous literature indicates that physical activity resources are less likely to be located in lower-SES and minority neighborhoods (4). In addition, one study determined that residing in high- versus low-walkability neighborhoods increased weekly physical activity but found no difference by neighborhood income (6). Despite these previous and current findings, some caution in interpretation should be noted. Adjusting for individual-level demographics may, in fact, be an over-adjustment and result in overestimation if these demographic variables are antecedent to the exposure and not true confounders.

Sampling is paramount when conducting observational studies intended to determine the effects of neighborhood walkability. An almost universal problem for these types of studies is overcoming neighborhood self-selection where individuals choose or are limited to neighborhoods with certain attributes (e.g. poor walkability) based on individual characteristics (e.g. low-SES) that are likely related to the outcome (24). Rather than using regression methods to “control for” the effects of SES, a few recent studies have attempted to overcome these effects by sampling high- and low-walkability neighborhoods with the caveat that these neighborhoods included a homogenous population with respect to SES and geographic location (42, 43). Although measures of walkability differed in these studies, both found positive associations for physical activity and/or utilitarian walking. The authors of the latter study note, though, that walkability was not meaningfully related to overall mean miles walked or increased physical activity (25). To assess the effect of neighborhood income, a more recent study recruited participants from high- vs. low-walkability and high- vs. low-SES neighborhoods (6). Again, participants were more active if they lived in high-walkability neighborhoods but, in contrast to the current results, there was no difference by neighborhood SES. Although the parent study for the current analysis was not designed to sample neighborhoods based on level of walkability, the results do suggest that the effect of neighborhood walkability on obesity differs by neighborhood characteristics.

Major strengths of this study were objective measures of the pedestrian environment and individual-level outcome measures. The PEDS audit measured the walking environment for transportation and physical activity at the micro-scale, thus, captured information that is not available through national databases; confirmatory factor analysis was used to reduce random measurement error in each PEDS item. Additionally, obesity was objectively measured which is the preferred measurement method for large epidemiologic studies. Furthermore, the stratified sampling design allowed for associations to be compared by neighborhood characteristics.

Nevertheless, this study had some limitations. First, the study was cross-sectional which limits the ability to make causal statements about observed associations. Second, census tract boundaries were used to approximate neighborhoods, which creates the potential for measurement error in environmental attributes located in a participant’s neighborhood. Nevertheless, the use of census tracts is common in the field since these boundaries are created to represent a homogeneous area (26). Third, the length of exposure to certain neighborhood characteristics is unknown, thus, associations may not reflect the walkability characteristics measured in this study. However, it is unlikely that mobile individuals move to drastically resource-different neighborhoods due to financial limitations and social preferences (45). Fourth, physical activity was self-reported for only on a small number of HANDLS participants. Nevertheless, since the physical activity measure did not capture utilitarian walking, an important factor in this line of research, a separate transportation question was used as a proxy for information on walking habits. Finally, the use of multilevel level models makes conclusions on possible intermediate variables difficult given that the variance structures of the neighborhood- and individual-level variables are different; caution should be taken when interpreting potential mediating pathways.

The literature base on walkability and obesity is still in its infancy. Although earlier work is promising, there are methodological issues that should be challenged in future work. First, the environmental determinants of obesity are numerous and few studies have incorporated comprehensive models to account for both energy expenditure and energy intake (i.e., environmental supports for physical activity and healthy dietary intake). Second, most previous studies were cross-sectional which severely limits the ability to imply causal associations; experimental or longitudinal studies are needed. Third, formative research should be used to establish the most appropriate neighborhood spatial scale for varying demographic and geographic populations. Fourth, few studies have explored walkability beyond the scope of neighborhoods (e.g., workplace). Finally, improvements in the conceptualization of walkability are warranted. Numerous attributes of walkability have been associated with obesity; future work should further define walkability for varying populations and geographic areas.

In conclusion, neighborhood walkability may partially explain racial and socioeconomic disparities in obesity. A thorough understanding of the underlying mechanisms in which these associations operate is justified. At the very least, individual physical activity recommendations and weight-management guidelines should recognize neighborhood walkability as an important enabler or inhibitor to meeting these guidelines.

Reference List

1. Mokdad AH, Bowman BA, Ford ES, Vinicor F, Marks JS, Koplan JP. The continuing epidemics of obesity and diabetes in the United States. JAMA. 2001;286:1195–200. [PubMed]
2. Centers for Disease Control and Prevention. National Diabetes Surveillence System. Atlanta, GA: 2006.
3. Centers for Disease Control and Prevention. Trends in Overweight and Obesity. Atlanta, GA: 2006.
4. Gordon-Larsen P, Nelson MC, Page P, Popkin BM. Inequality in the built environment underlies key health disparities in physical activity and obesity. Pediatrics. 2006;117:417–424. [PubMed]
5. Frank LD, Sallis JF, Saelens BE, et al. The Development of a Walkability Index: Application To the Neighborhood Quality of Life Study. Br J Sports Med. 2009 [PubMed]
6. Sallis JF, Saelens BE, Frank LD, et al. Neighborhood built environment and income: Examining multiple health outcomes. Soc Sci Med. 2009 [PMC free article] [PubMed]
7. Clifton K, Livi Smith AD, Rodriguez D. The development and testing of an audit for the pedestrian environment. Landscape and Urban Planning. 2006 In Press.
8. National Institute on Aging. Healthy Aging in Neighborhoods of Diversity Across the Life-Span. Baltimore: Department of Health and Human Services; 2004.
9. Morland K, Wing S, Diez RA, Poole C. Neighborhood characteristics associated with the location of food stores and food service places. Am J Prev Med. 2002;22:23–29. [PubMed]
10. Boarnet MG, Day K, Alfonzo M, Forsyth A, Oakes M. The Irvine-Minnesota inventory to measure built environments: reliability tests. Am J Prev Med. 2006;30:153–159. [PubMed]
11. Yanovski S. Overweight, Obesity and Health Risk: National Task Force on the Prevention and Treatment of Obesity. Arch Intern Med. 2000;160:898–904. [PubMed]
12. Jurca R, Jackson AS, Lamonte MJ, et al. Assessing cardiorespiratory fitness without performing exercise testing. Am J Prev Med. 2005;29:185–193. [PubMed]
13. Muthen LK, Muthen BO, editors. Mplus. Mplus User’s Guide. Los Angeles: Muthen & Muthen; 2008.
14. Stata Statistical Software 9.0. College Station, Texas: 2005.
15. Centers for Disease Control and Prevention. Prevalence of Physical Activity, Including Lifestyle Activities Among Adults- United States, 2000–2001. Atlanta, GA: Morbidity and Mortality Weekly; 2003. pp. 764–769. [PubMed]
16. Scott MM, Dubowitz T, Cohen DA. Regional differences in walking frequency and BMI: What role does the built environment play for Blacks and Whites? Health Place. 2009;15:882–7. [PMC free article] [PubMed]
17. Franco M, Diez Roux AV, Glass TA, Caballero B, Brancati FL. Neighborhood characteristics and availability of healthy foods in Baltimore. Am J Prev Med. 2008;35:561–7. [PMC free article] [PubMed]
18. Franco M, Diez-Roux AV, Nettleton JA, et al. Availability of healthy foods and dietary patterns: the Multi-Ethnic Study of Atherosclerosis. Am J Clin Nutr. 2009 [PubMed]
19. Centers for Disease Control and Prevention. Trends in Intake of Energy and Macronutrients- United States, 1971–2000, Morbidity and Mortality Weekly Report. Atlanta, GA: Department of Health and Human Services; 2004. pp. 80–82. [PubMed]
20. Crespo CJ, Keteyian SJ, Heath GW, Sempos CT. Leisure-time physical activity among US adults. Results from the Third National Health and Nutrition Examination Survey. Arch Intern Med. 1996;156:93–98. [PubMed]
21. Boardman JD, Saint Onge JM, Rogers RG, Denney JT. Race differentials in obesity: the impact of place. J Health Soc Behav. 2005;46:229–43. [PMC free article] [PubMed]
22. Chang VW. Racial residential segregation and weight status among US adults. Soc Sci Med. 2006;63:1289–303. [PubMed]
23. Sallis JF, King AC, Sirard JR, Albright CL. Perceived environmental predictors of physical activity over 6 months in adults: activity counseling trial. Health Psychol. 2007;26:701–9. [PubMed]
24. Oakes JM. The (mis)estimation of neighborhood effects: causal inference for a practicable social epidemiology. Soc Sci Med. 2004;58:1929–52. [PubMed]
25. Oakes JM, Forsyth A, Schmitz KH. The effects of neighborhood density and street connectivity on walking behavior: the Twin Cities walking study. Epidemiol Perspect Innov. 2007;4:16. [PMC free article] [PubMed]
26. 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. [PMC free article] [PubMed]