History

F A J

B

M

R

C

I

W o b o i e l u

i f

F I L P L

D i k

w

A ©

ast Food, Race/Ethnicity, and Income Geographic Analysis

ason P. Block, MD, MPH, Richard A. Scribner, MD, MPH, Karen B. DeSalvo, MD, MPH, MSc

ackground: Environmental factors may contribute to the increasing prevalence of obesity, especially in black and low-income populations. In this paper, the geographic distribution of fast food restaurants is examined relative to neighborhood sociodemographics.

ethods: Using geographic information system software, all fast-food restaurants within the city limits of New Orleans, Louisiana, in 2001 were mapped. Buffers around census tracts were generated to simulate 1-mile and 0.5-mile “shopping areas” around and including each tract, and fast food restaurant density (number of restaurants per square mile) was calculated for each area. Using multiple regression, the geographic association between fast food restaurant density and black and low-income neighborhoods was assessed, while controlling for environmental confounders that might also influence the placement of restaurants (commercial activity, presence of major highways, and median home values).

esults: In 156 census tracts, a total of 155 fast food restaurants were identified. In the regression analysis that included the environmental confounders, fast-food restaurant density in shopping areas with 1-mile buffers was independently correlated with median household income and percent of black residents in the census tract. Similar results were found for shopping areas with 0.5-mile buffers. Predominantly black neighborhoods have 2.4 fast-food restaurants per square mile compared to 1.5 restaurants in predominantly white neighborhoods.

onclusions: The link between fast food restaurants and black and low-income neighborhoods may contribute to the understanding of environmental causes of the obesity epidemic in these populations. (Am J Prev Med 2004;27(3):211–217) © 2004 American Journal of Preventive Medicine

o i i r y

h i T f t r t l t a t t m t b b s

ntroduction

hile obesity has a range of causes from ge- netic to environmental, the environment is a key factor in the rapid development of the

besity epidemic.1– 4 Increased food consumption may e the most important of recent changes leading to an besogenic environment.5 Despite stable physical activ-

ty patterns during the last 20 years,6,7 Americans are ating more,8 portion sizes have increased substantial- y,9 and inexpensive, high-calorie food is now biquitous. The growth of the fast-food industry has been an

mportant environmental inducement for increased ood consumption. In the last 20 years, the percentage

rom the Tulane University School of Medicine, Department of nternal Medicine (Block, DeSalvo), New Orleans, Louisiana, and ouisiana State University Health Sciences Center, Department of ublic Health and Preventive Medicine (Scribner), New Orleans, ouisiana Address correspondence to: Karen DeSalvo, MD, MPH, MSc, epartment of Internal Medicine, Tulane University School of Med-

cine, 1430 Tulane Avenue, SL 16, New Orleans LA 70112. E-mail: [email protected]. The full text of this article is available via AJPM Online at

dww.ajpm-online.net.

m J Prev Med 2004;27(3) 2004 American Journal of Preventive Medicine • Published by

f calories attributable to fast-food consumption has ncreased from 3% to 12% of total calories consumed n the United States.10 U.S. spending on fast food has isen from $6 billion to $110 billion over the last 30 ears.11

Fast food is notably high in fat content,12 and studies ave found associations between fast food intake and

ncreased body mass index (BMI) and weight gain.13,14

hese same studies reported increased consumption of ast food among nonwhite and low-income popula- ions. Despite these relationships between income, ace/ethnicity, obesity, and fast food, limited research o date has examined such associations on an ecologic evel. Morland et al.15 examined the relationship be- ween fast-food restaurants, race/ethnicity, and wealth s an ancillary analysis in a large ongoing study based in he United States, and discovered no consistent rela- ionship between wealth, measured with census tract

edian home values, and fast-food restaurants. Addi- ionally, they found no difference between the num- ers of fast-food restaurants in black and white neigh- orhoods. Reidpath et al.16 found diverging results in a tudy addressing fast-food restaurant density and me-

ian individual income in Melbourne, Australia. Resi-

2110749-3797/04/$–see front matter Elsevier Inc. doi:10.1016/j.amepre.2004.06.007

d t l s i p

M D

R t d a c O f t f A r a t

r S i a w

G

U r a c s f

p M 2 c s

( ( a T s t i “ t

L

B f v c T d n O s h d m

c o b g o g l h v r g f l s r e 0

V

U W a b m a c p w v

T a

N

B B C C D K M P P P R S T W T a

a

2

ents of the lowest-income neighborhoods had 2.5 imes more exposure to fast-food restaurants than those iving in the most affluent neighborhoods. The current tudy was an assessment of whether black and low- ncome neighborhoods have increased geographic ex- osure to fast food restaurants.

ethods efinition and Identification

esearchers defined fast-food restaurants as chain restaurants hat have two or more of the following characteristics: expe- ited food service, takeout business, limited or no wait staff, nd payment tendered prior to receiving food. The national hain restaurants included had at least two restaurants in rleans Parish (parish is the unique Louisiana designation

or a county; the boundaries of Orleans Parish approximate he City of New Orleans), and tend to be recognized as ast-food restaurants in the media and by the general public. dditionally, one local fast-food chain that has five restau-

ants in Orleans Parish was included (Table 1). These criteria llowed for inclusion of the fast-food restaurants that cap- ured the largest portion of the fast-food market.

Between August and October 2001, researchers identified estaurants by examining the log books for the Orleans Parish anitation Department, which inspects all chain food outlets n the parish, by reviewing the local Yellow Pages phone book, nd by accessing restaurant locator engines on fast food chain ebsites.

eocoding and Census Tract Inclusion Criteria

sing geographic information system software, all fast food estaurants were geocoded in Orleans Parish.17 Geocoding is

technique now widely used in public health research to reate electronic mapping of exposure to physical structures uch as toxic waste plants, alcohol outlets, or in this study,

able 1. Fast-food restaurant chains included in the nalysis

ame

Number of restaurants in Orleans Parisha

ud’s Broiler 5 urger King 16 hick-fil-A 3 hurch’s Chicken 11 omino’s Pizza 11 entucky Fried Chicken (KFC) 7 cDonald’s 18

apa John’s 4 izza Hut 9 opeyes Chicken and Biscuits 23 ally’s Hamburgers 9 ubway 23 aco Bell 6 endy’s Old Fashioned Hamburgers 10 otal 155

Parish is the unique Louisiana designation for a county; the bound- ries of Orleans Parish approximate the City of New Orleans.

ast-food restaurants. The geocoded restaurants were im- t

12 American Journal of Preventive Medicine, Volume 27, Num

orted onto a census tract map for Orleans Parish using apInfo, version 6.2 (Matchware Technologies Inc., Troy NY,

000). Previous small-area research guided the selection of ensus tracts as the model of a neighborhood in this tudy.17,18

The analysis was restricted to those census tracts with 1) �500 people, (2) �2000 people per square mile, and 3) �200 alcohol outlets per 1000 people. Researchers used lcohol outlet density as a proxy for commercial activity. hese restrictions ensured that neighborhoods analyzed are

imilar (i.e., urban and residential). Despite these restric- ions, fast-food restaurants in the excluded tracts are included n the analysis when these restaurants were located within the shopping area” (described below) of a neighboring census ract that met the inclusion criteria.

evels of Analysis

ecause of interest in the association between environmental actors and neighborhood demographics, researchers used ariables on two geographic levels. The first level was the ensus tract, where the population variables were measured. hese variables—the percentage of black residents and me- ian household income— describe the demographics of the eighborhoods. The second level was the “shopping area.” n this level were the environmental variables, which de-

cribe geographic exposures to those living in the neighbor- oods: fast-food restaurant density (FFRD), alcohol outlet ensity, presence of interstate or major state highways, and edian home value as a proxy for property values. We created “shopping areas” by producing buffers around

ensus tracts. These shopping areas, which included the area f the census tract and the area between the buffer and the order of the tract, provide a more realistic representation of eographic exposure than census tracts alone because people ften have to travel outside of their census tract to purchase oods. For example, many of the fast-food restaurants were ocated just beyond the border of a particular tract and would ave been easily accessible to and patronized by many indi- iduals living within that census tract. However, these restau- ants would have been excluded in the calculation of geo- raphic exposure for that tract unless buffers were used. In act, 62% of the census tracts have no fast-food restaurants ocated directly within their borders. However, only 2% of the hopping areas with 1-mile buffers have zero fast food restau- ants. For a sensitivity analysis, fast-food restaurant density was xamined separately by creating buffers that were 1 mile and .5 mile from the census tract borders.

ariables

sing multiple regression in SPSS (Graduate Pack 10.0 for indows, SPSS Inc., Chicago IL, 1999), the geographic

ssociation between FFRD and black and low-income neigh- orhoods was assessed after controlling for other key environ- ental variables: alcohol outlet density, presence of highways,

nd median home value. These variables were included as ovariates in the model because they might influence the lacement of fast-food restaurants. All variables in the analysis ere log transformed, except for the dichotomous highway ariable, to adjust for skew and to allow for elasticity calcula-

ions. Elasticity calculations show that for a given percentage

ber 3

c c

n F 1

l a m a (

r s v l

m a A c s a s s i c t r t m b a

i T l m m

R

I e t c e u c m F

R D

O c n t

p a

1 m 0 c w t d

B

S c 1 r p m w ( p t F d

R

R s ( m 2 m r B c h v c r t a h t c r t

s h F v

1 i h

hange in the independent variable, the dependent variable hanges by a certain percentage.

The dependent variable—FFRD—was calculated as the umber of restaurants per square mile in the shopping area. FRDs were evaluated separately for shopping areas with -mile and 0.5-mile buffers. The primary predictor variables were the following popu-

ation variables: neighborhood percentage of black residents nd median household income. We used census tract esti- ates for 1999, which were based on the 1990 census with

djustments made by a commerical vendor of census data Claritas Trend Map, San Diego CA, 1999).

The environmental variables controlled for in the analysis equired geocoding, and were collected at the level of the hopping area. Despite the classification of median home alue as an environmental variable, it was only available at the evel of the census tract.

We calculated the alcohol outlet density in the same anner as the FFRD for each shopping area. Locations of

lcohol outlets were available in 1999 from the Louisiana lcohol Policy Needs Assessment Database.17 The database haracterizes alcohol outlets as on-sale (sites where alcohol is old for consumption on the premises, such as restaurants nd bars) and off-sale (sites where alcohol is sold for con- umption away from the premises, such as liquor or grocery tores). A summary alcohol outlet variable was created includ- ng both on-sale and off-sale outlets as a proxy variable for ommerical activity. Controlling for commerical activity in his study was necessary because fast-food chains may place estaurants in highly commercial areas due to zoning restric- ions. Alcohol outlets are an ideal proxy measure for com-

ercial activity in Louisiana because these outlets include ars, restaurants, liquor stores, grocery stores, drug stores, nd convenience stores.

The highway variable accounted for the presence of an nterstate highway or state highway in each shopping area. he presence of highways may dictate fast food restaurant

ocation. Median home values also may influence the place- ent of fast food restaurants because to control costs, chains ay locate on land with lower property values.

egression Analysis

n the regression analysis, the researchers expected that the nvironmental covariates would explain a large percentage of he variance in FFRD. Therefore, a base regression model was onstructed with FFRD as the dependent variable and the nvironmental variables as the predictor variables. The pop- lation variables—median household income and the per- entage of black residents—were sequentially added to the odel to determine their effect on explained variance in

FRD. All variables were part of the final model.

esults escriptive

f the 184 census tracts, a total of 156 met the inclusion riteria. Table 1 provides a list of fast food chains and umbers of restaurants in Orleans Parish. The census

ract map of Orleans Parish in Figure 1 shows both the a

lacement of fast food restaurants as well as excluded nd included census tracts.

The mean FFRD for shopping areas defined with a -mile buffer was 2.48 fast food restaurants per square ile; the mean FFRD for shopping areas defined with a

.5-mile buffer was 2.54 restaurants per square mile. In ensus tracts, the mean percentage of black residents as 60.6%. The mean household income at the census

ract level was $25,450. Table 2 contains all relevant escriptive information.

ivariate Analysis

pearman’s rank correlation coefficients were signifi- ant when comparing FFRD in the shopping areas with -mile buffers to the neighborhood percent of black esidents and median household income (r �0.160, �0.046 for percent black; r ��0.275, p ��0.001 for edian household income). Similarly, correlations ere significant in shopping areas with 0.5-mile buffers r �0.180, p ��0.024 for percent black; r ��0.266, ��0.001 for median household income). Correla-

ion coefficients also were significant when comparing FRDs to the presence of highways and alcohol outlet ensities (all p values �0.001).

egression

egression statistics are shown in Table 3. Using the hopping areas with 1-mile buffers, the base model Model 1: alcohol outlet density, presence of highways, edian home value as predictor variables) explained

5.0% of the variance in FFRD. For Models 2 and 3, edian household income and the percentage of black

esidents were added to the base model, respectively. oth variables were significant predictors of FFRD after ontrolling for the base model variables. Median house- old income explained an additional 3.3% of the ariance in FFRD above that of the base model (F test hange�7.87; p �0.006), and the percentage of black esidents explained 19.1% of the variance above that of he base model (F test change�52.7; p �0.001). When ll variables were included together in Model 4, median ousehold income was no longer significant. However,

he percentage of black residents remained a signifi- ant predictor of FFRD. Adding the percentage of black esidents to Model 2 explained an additional 16% of he variance (F test change�44.2; p �0.001).

The sensitivity analysis found similar results for the hopping areas with 0.5-mile buffers. However, median ousehold income was not a significant predictor of FRD in Model 2 after controlling for the base model ariables.

The regression equation for shopping areas with -mile buffers demonstrates that for every 10% increase n fast food restaurant density, neighborhood median ousehold income decreased by 4.8% and the percent-

ge of black residents increased by 3.7%. The regres-

Am J Prev Med 2004;27(3) 213

s f n n d c b m 6 m s w p

D

F p a h b t

F

T d

V

P

E

E

a

o

2

ion equation was solved to determine how many more ast food restaurants are located in predominately black eighborhoods compared to predominately white eighborhoods. Neighborhoods with 80% black resi- ents had 2.4 fast food restaurants per square mile ompared to 1.5 restaurants per square mile in neigh- orhoods with 20% black residents. In this study, the ean size of a shopping area with a 1-mile buffer was

.2 square miles, with a range of 4.2 to 15.0 square iles. Therefore, for an average-sized neighborhood

hopping area, predominantly black neighborhoods ere exposed to six more fast food restaurants than redominantly white neighborhoods.

iscussion

ast-food restaurants are geographically associated with redominately black and low-income neighborhoods fter controlling for commercial activity, presence of ighways, and median home values. The percentage of lack residents is a more powerful predictor of FFRD

igure 1. Census tract map of Orleans Parish, Louisiana.

han median household income. Predominantly black S

14 American Journal of Preventive Medicine, Volume 27, Num

able 2. Population and environmental variable escriptives

ariable Mean SD

opulation Percentage of black

residents 60.6% 32.9%

Median household income $ 28,282 $17,211 nvironmental: shopping areas with 1-mile buffer Fast-food restaurant density

(restaurants/square mile)

2.48 1.6

Alcohol outlet density (outlets/square mile)

30.8 25.7

Percentage of census tracts with highway(s) in shopping area

81.4% 39.0%

Median home valuea $101,224 $61,947 nvironmental: shopping areas with 0.5-mile buffer Fast-food restaurant density

(restaurants/square mile) 2.54 2.0

Alcohol outlet density (outlets/square mile)

33.3 36.1

Percentage of census tracts with highway(s) in shopping area

66.7% 47.3%

Median home valuea $101,224 $61,947

Median home value is an environmental variable but was available nly at the level of the census tract.

D, standard deviation.

ber 3

n f p T u c s m a

t 0 i c a b f t A d l A s 0 m

G

M c i f t a t e f

l v t T d r t b e s h F u e

b c c f f a L t d O t r a a n n n c “ g

T

1

2

3

4

S

eighborhoods (i.e., 80% black) have one additional ast-food restaurant per square mile compared with redominantly white neighbohoods (i.e., 80% white). hese findings suggest that black and low-income pop- lations have more convenient access to fast food. More onvenient access likely leads to the increased con- umption of fast food in these populations,13,14 and ay help to explain the increased prevalence of obesity

mong black and low-income populations. Researchers chose to evaluate geographic associa-

ions with FFRD in shopping areas with 1-mile and .5-mile buffers because of an uncertainty of how far ndividuals were willing to routinely travel outside their ensus tract to purchase food. The use of shopping reas defined by 1-mile buffers seems more justified ased on reports regarding McDonald’s strategy for ranchise locations. McDonald’s has established a res- aurant within a 3- to 4-minute trip for the average merican.19 Under the assumption that an individual rives 25 miles per hour, a McDonald’s should be

ocated within approximately 1.5 miles of the average merican’s home. This distance is more consistent with

hopping areas with 1-mile buffers than those with .5-mile buffers, thereby potentially explaining the ore powerful results for the 1-mile buffer analysis.

eographic Associations

orland et al.15 reported contrasting results from the urrent study, but their study diverged from this study n several ways. First, they did not adjust their analysis or other environmental factors that might influence he placement of fast food restaurants. In the bivariate nalysis, a significant (although weak) relationship be- ween FFRD and the percentage of black residents xisted, which increased substantially after controlling

able 3. Regression models for shopping areas with 1-mile b

Model Variables included Coe (�)

Alcohol outlet density 0 Highway 0 Median home value �0 Median household income �0 Alcohol outlet density 0 Highway 0 Median home value 0 Percentage of black residents 0 Alcohol outlet density 0 Highway 0 Median home value 0 Median household income �0 Percentage of black residents 0 Alcohol outlet density 0 Highway 0 Median home value 0

E, standard error.

or environmental confounders, including alcohol out- h

et density, presence of highways, and median home alues. Second, they did not utilize shopping areas as he area of geographic exposure for a neighborhood. his method is important because many census tracts o not have any fast food restaurants; however, people esiding in these tracts are still geographically exposed o restaurants that are nearby but not within the tract oundaries. By creating shopping areas, geographic xposure is more effectively modeled. Third, the mea- ure of wealth in the current study was median house- old income. No consistent relationships between FRD and median home value (the measure of wealth sed by Morland et al.15) were found in this study ither. Other geographic research has shown associations

etween neighborhood demographics and exposure to onsumer goods that contribute to negative health onsequences. As previously discussed, Reidpath et al.16

ound similar results to this study when comparing fast ood restaurant density to median household income mong neighborhoods in Melbourne, Australia. aVeist and Wallace,20 as well as Scribner et al.,21 found

hat liquor stores are more commonly located in pre- ominantly black and low-income neighborhoods. ther studies have found links between higher densi-

ies of alcohol outlets and increased rates of alcohol- elated outcomes, such as motor vehicle crashes22 and ssaultive violence.23 For food availability, Morland et l.15 found that wealthy and predominantly white eighborhoods have more supermarkets and fewer eighborhood grocery stores than poor and predomi- antly black neighborhoods, an important finding be- ause research indicates that supermarkets have more heart-healthy” foods when compared to neighborhood rocery stores and convenience stores.24 Still others

nt SE

Two- sided p values

Adjusted r2

of model

0.065 0.006 0.250 0.071 �0.001 0.128 0.352 0.173 0.006 0.283 0.065 0.025 0.070 �0.001 0.214 0.089 0.051 �0.001 0.441 0.057 �0.001 0.061 �0.001 0.152 �0.001 0.158 0.221 0.442 0.053 �0.001 0.059 �0.001 0.062 �0.001 0.199 �0.001

uffer

fficie

.183

.407

.120

.485

.147

.375

.367

.368

.256

.364

.631

.195

.350

.238

.353

.790

ave demonstrated a positive association between in-

Am J Prev Med 2004;27(3) 215

c g

F

R o d c c e

o i m d c g t 3 c F f i 1

t v w p a

F

T b W c t p b c t f o “ w i r m m

i f w c e

a h

a o o n t m F s l c r p

L

D r d D r r c s

e r c h o r t s r a s “ p a c t f s t m r

b h d d o m a

2

ome and the availability of “healthful products” in rocery stores at the community (city or county) level.25

ast Food and Obesogenic Environment

esearchers have implicated environmental influences n body weight as the primary contributor to the evelopment of the obesity epidemic.3,4,26 The in- reased availability and consumption of food is a major omponent of an increasingly obesogenic nvironment. Despite a decrease in the fat content (as a percentage

f total calories) of the average American’s diet, Amer- cans are consuming more calories. The U.S. Depart-

ent of Agriculture reported an increase in the average aily food energy intake from 1854 calories to 2002 alories between 1977–1978 and 1994 –1996.8 The rowth of “dining out” has significantly contributed to his rise.10 In 1995, “away-from-home” foods provided 4% of total caloric intake and 38% of total fat intake ompared to 18% for both categories in 1977–1978. ast food is a major component of the away-from-home ood category, accounting for 12% of total caloric ntake for Americans in 1995 compared to only 3% in 977–1978. Fast-food consumption is also related to obesity, and

his relationship is strongest among low-income indi- iduals.13,14 All of this supporting evidence, coupled ith the results of this study, suggests that fast food may lay a role in the obesity epidemic among low-income nd black communities.

ood Availability and Diet

hese results suggest that black and low-income neigh- orhoods have increased exposure to fast food. hether increased availability of fast food promotes

onsumption is not the subject of this study. However, heoretically, more convenient access to fast food cou- led with the decreased availability of healthy food in lack and low-income neighborhoods may increase onsumption of unhealthy foods. In keeping with this heoretical construct, Cheadle et al.25,27 reported that ood availability in grocery stores was linked to the diet f residents in the nearby areas. They found that more healthful products” in grocery stores were associated ith increased consumption of “healthful products” by

ndividuals living near those stores. Another study18

eported that black Americans consume one third ore fruits and vegetables for every additional super- arket found in their census tract. Evidence also suggests that low-income and nonwhite

ndividuals do consume more fast food and unhealthy ood. French et al.13 noted that low income and non- hite ethnicity were associated with increased fast food onsumption. According to a British study, lower socio-

comic groups had diets with less vegetables and fruit, b

16 American Journal of Preventive Medicine, Volume 27, Num

nd more meat products, fats, and sugars compared to igher socioeconomic groups.28

One explanation for these findings is that restaurants nd stores adapt their selection to the food preferences f individuals living nearby. Therefore, they may not ffer healthy food options in black and low-income eighborhoods because their market research indicates

hat demand for such products is weak in those com- unities. However, the opposite might also be true.

ood preferences could partly be dictated by available election in a neighborhood, especially because of the ower access to transportation in black and low-income ommunities.15,29 Likewise, because of limited financial esources, black and low-income populations may sim- ly seek out the most calories for the lowest price.

imitations and Future Research

espite an exhaustive effort to identify all fast-food estaurants in Orleans Parish by searching telephone irectories, websites, and the Orleans Parish Sanitation epartment records, we may have missed some restau-

ants. However, it is unlikely that we under-counted estaurants disproportionately based on demographic haracteristics of neighborhoods (i.e., nondifferential election bias).

This study’s definition of fast-food restaurants also xcludes some similar restaurants. Many local restau- ants may have expedited food service but are not onnected to a chain, and some national chains may ave only one restaurant in Orleans Parish. In an ngoing study of fast-food restaurants and full-service estaurants, we have discovered that the fast-food res- aurants included in this study comprise 67% of all imilar restaurants in Orleans Parish (including restau- ants that do not fit this study’s inclusion criteria such s single-site, fast-food restaurants, and chain, full- ervice restaurants identified as serving fried chicken, po-boys,” sandwiches, fries, burgers, hot dogs, shakes, izza). These excluded restaurants may be located in reas of the parish that have different demographic haracteristics than what was discovered for those res- aurants included in this study. However, major fast- ood chains with a significant presence in the area hould serve the great majority of fast food meals and, herefore, are most relevant to this analysis. Further-

ore, no data exist to suggest that these other restau- ants serve different areas of the parish.

Despite this study’s recognition of an association etween FFRD and black and low-income neighbor- oods, the directionality of the relationship cannot be etermined by this study. For example, neighborhood emographics could be temporally shaped by the type f restaurants in the area (making a neighborhood ore or less desirable) or other local features associ-

ted with these restaurants. Likewise, restaurants could

e established within neighborhoods that demograph-

ber 3

i s t d n m d

g l f e t fi

t s i l

C

F b s t l

R

1

1 1

1

1

1

1

1

1

1 2

2

2

2

2

2

2

2

2

2

cally fit a restaurant’s target audience. Future research hould examine the association between fast food res- aurants and neighorhood characteristics in a longitu- inal manner. Only by tracking the demographics of eighorhoods over time and identifying the establish- ent date of fast food restaurants can researchers

etermine a temporal relationship. This study’s focus on one parish/county limits the

eneralizability of results. Orleans Parish has a very arge poor and black population. The link between ast-food restaurants and these neighborhoods may xist due to unique characteristics of this region. Fu- ure research should attempt to duplicate this study’s ndings in diverse regions. Studies should also examine the geographic associa-

ion between neighborhood fast-food restaurant den- ity and obesity rates at both the neighborhood and ndividual levels. The ability to geocode and use multi- evel designs now make this type of study possible.17

onclusions

ast-food restaurants are more commonly located in lack and low-income neighborhoods. This link may uggest environmental exposure to fast food as a con- ributor to the high prevalence of obesity in black and ow-income populations.

eferences 1. Comuzzie A, Allison D. The search for human obesity genes. Science

1998;280:1374 –7. 2. Bouchard C. Current understanding of the etiology of obesity: genetic and

nongenetic factors. Am J Clin Nutr 1991;53:1561S–5S. 3. Hill J, Peters J. Environmental contributions to the obesity epidemic.

Science 1998;280:1371– 4.

What This Study Adds . . .

The connection between fast-food restaurants and black and low-income neighborhoods may contribute to the understanding of environmen- tal causes of the obesity epidemic in these popu- lations.

This study is the first to document that predom- inantly black neighborhoods have higher densi- ties of fast-food restaurants compared to largely white neighborhoods.

Additionally, this study contributes to the body of literature that has linked high fast-food restau- rant density to low-income neighborhoods.

4. Koplan J, Dietz W. Caloric imbalance and public health policy. JAMA 1999;282:1579 – 82.

5. Swinburn B, Egger G, Raza F. Dissecting obesogenic environments: the development and application of a framework for identifying and prioritiz- ing environmental interventions for obesity. Prev Med 1999;29:563–70.

6. Mortality Statistics Branch, National Center for Health Statistics. Physical activity trends—United States, 1990 –1998. MMWR Morb Mortal Wkly Rep 2001;50:166 –9.

7. Centers for Disease Control and Prevention. Physical activity and health: a report of the Surgeon General. Atlanta GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, 1996.

8. Tippett K, Cleveland L. How current diets stack up— comparison with dietary guidelines. In: Frazao E, ed. America’s eating habits: changes and consequences. Washington DC: U.S. Department of Agriculture, 1999: 59 – 63 (Information Bulletin A1B-750).

9. Young L, Nestle M. The contribution of expanding portion sizes to the US obesity epidemic. Am J Public Health 2002;92:246 –9.

0. Lin B, Frazao E. Away-from-home foods increasingly important to quality of American diet. Washington DC: U.S. Department of Agriculture, 1999 (Agriculture Information Bulletin 749).

1. Schlosser E. Fast food nation. New York: Houghton Mifflin, 2001. 2. Massachusetts Medical Society Committee on Nutrition. Fast-food fare.

N Engl J Med 1989;321:752– 6. 3. French S, Harnack L, Jeffery R. Fast food restaurant use among women in

the Pound of Prevention study: dietary, behavioral and demographic correlates. Int J Obes 2000;24:1353–9.

4. Jeffery R, French S. Epidemic obesity in the United States: are fast foods and television viewing contributing? Am J Public Health 1998;88:277– 80.

5. Morland K, Wing S, Diez Roux A, Poole C. Neighborhood characteristics associated with the location of food stores and food service places. Am J Prev Med 2002;22:23–9.

6. Reidpath D, Burns C, Garrand J, Mahoney M, Townsend M. An ecological study of the relationship between social and environmental determinants of obesity. Health Place 2002;8:141–5.

7. Scribner R, Cohen D, Fisher W. Evidence of a structural effect for alcohol outlet density: a multilevel analysis. Alcohol Clin Exp Res 2000;24:188 –95.

8. Morland K, Wing S, Diez Roux A. The contextual effect of the local food environment on residents’ diets: the atherosclerosis risk in communities study. Am J Public Health 2000;92:1761–7.

9. Lubow A. Steal this burger. New York Times, 19 April 1998, p. 38. 0. LaVeist T, Wallace J. Health risk and inequitable distribution of liquor

stores in African American neighborhood. Soc Sci Med 2000;51:613–7. 1. Scribner R, Cohen D, Kaplan S, Allen S. Alcohol availability and homicide

in New Orleans: conceptual considerations for small area analysis of the effect of alcohol outlet density. J Stud Alcohol 1999;66:310 – 6.

2. Scribner R, MacKinnon D, Dwyer D. Alcohol outlet density and motor vehicle crashes in Los Angeles County cities. J Stud Alcohol 1994;55:447–53.

3. Scribner R, MacKinnon D, Dwyer J. The risk of assaultive violence and alcohol availability in Los Angeles County. Am J Public Health 1995;85:335– 40.

4. Sallis J, Nader R, Atkins J. San Diego surveyed for heart healthy foods and exercise facilities. Public Health Rep 1986;101:216 – 8.

5. Cheadle A, Psaty B, Curry S, et al. Community-level comparisons between the grocery store environment and individual dietary practices. Prev Med 1991;20:250 – 61.

6. Poston W, Foreyt J. Obesity is an environmental issue. Atherosclerosis 1999;146:201–9.

7. Cheadle A, Psaty B, Curry S, et al. Can measures of grocery store environment be used to track community-level dietary changes? Prev Med 1993;22:361–72.

8. James W, Nelson M, Ralph A, Leather S. The contribution of nutrition to inequalities in health. BMJ 1997;314:1545–9.

9. Turrell G. Structural, material, and economic influences of the food purchasing choices of socioeconomic groups. Aust N Z J Public Health 1996;20:11–7.

Am J Prev Med 2004;27(3) 217