RESEARCH AND PRACTICE

Clustering of Fast-Food Restaurants Around Schools:A Novel Application of Spatial Statistics to the Studyof Food Environments

S, Bryn Austin, ScD, Steven J. Melly, MS. Brisa N. Sanchez. ScM. Aarti Patel. BA. Stephen Buka, ScD, and Steven L. Gortmaker, PhD

Over the past 3 decades, fast-food retail salesin the United States have soared 900"/o. from$16.1 billion in 1975 to a projected $153.1billion in 2004,' Tlie number of fast-foodrestaurants in the country now exceeds280000. hi this same period, Americanshave become increasingly dependent onrestaurants and fast-food chains for theirmeals, ^ with almost half of US food spendinggoing toward food eaten away from home.''The fast-food industry maikets heavily tochildren and adolescents, who make up animportant part of the industry's consumerbasa. *' Among youths aged 12 to 18 years,the percentage of tutal energ}' intake con-sumed from fast-food and other restaurantshas increased from 6.51 /0 in 1977-1978 to19-3% in 1994-1996.^ On a Epical day, al-most a third of children and adolescents eatfast food/ Portion sizes and the correspon-ding caloric content of foods served at fast-food restaurants have also increased apprecia-bly over the past several decades.^ On dayswhen youths eat fast loud, compai'ed withdays when they do not, they consume moretotal calories, more fat. more added .sugars.more sugar-sweetened drinks, and fewerfruits and vegetables/ ' Among adults, eatingin fast-food restaurants is cross-sectionally as-sociated with higher body mass index'" andlongitudinally associated wiih weigbt gain" 'and development of insulin resistance,'' bothimportant risk factors for diabetes,"

Consumers' ease of access to fast food is apriority for the industry, as business plannershave long been aware of the potential tomaximize sales by selecting restaurant sitesthat ensure close proximity to the consumerbase." '*• ln re.staurant site selection, businessplanners consider neighborhood demograph-ics and the presence of other businesses, com-munity organizations, and aspects of theneighborhood roadways that may draw in po-

Objectives. We examined the concentration of fast food restaurants in areasproximal to schools to characterize school neighborhood food environments.

Methods. We used geocoded databases of restaurant and school addresses toexamine locational patterns of fast-food restaurants and kindergartens and pri-mary and secondary schools in Chicago. We used the bivariate K function sta-tistical method to quantify the degree of clustering (spatial dependence) of fast-food restaurants around school locations.

Results. The median distance from any school in Chicago to the nearest fast-foodrestaurant was 0.52 km, a distance that an adult can walk in little more than 5 min-utes, and 78% of schools had at least 1 fast-food restaurant within 800 m. Fast-foodrestaurants were statistically significantly clustered in areas within a short walk-ing distance from schools, with an estimated 3 to 4 times as many fast-food restau-rants within 1.5 km from schools than would be expected if the restaurants weredistributed throughout the city in a way unrelated to school locations.

Conclusions. Fast-food restaurants are concentrated within a short walkingdistance from schools, exposing children to poor-quality food environments intheir school neighborhoods. (-Am JPub//cHea/f/?. 2005;95:1575-1581, doi:10.2105/AJPH.2004.056341)

tential consumers.'^ '** In recent years, publichealth researchers have begun to developmethods to characterize neighborhood pat-tems in location and density of different typesof food establishments, such as fast-foodrestaurants or grocery stores. ' •^" Given tliefast-food industry's targeted marketing tochildren and adolescents and the importanceof young people in the industry's consumerbase, we used spatial statistical methods to ex-amine whether fast-food restaurants may con-centrate in areas proximal to kindergartensand primary and secondary' schools wherethey would be highly accessible to students,

METHODS

Fast-Food Restaurant DatabaseIn 2002, we compiled a comprehensive

list of fast-food restaurants in Chicago alongwith their street addresses. To create thecomprehensive list, we used as our primarysource Technomic Inc, a food industry mar-ket research company that publishes a list of

the leading limited service chain brands(2002 Technomic Top 100),^^ as ranked bytotal US sales, which includes the major fast-food chains.^' Consistent with food industrystandards. Technomic defines limited servicechain restaurants as eating places where cus-tomers order items and pay before eatingand where food can be eaten on the prem-ises or taken out,^' To ensure that our data-base was both comprehensive and appropri-ate for Chicago, we gathered information onfast-food restaurants in the city from severalother sources. In addition to the Technomiclist, we also referred to a privately run Website. Fast Food Source (http://www.fastfQod-source.com), to identify fast-food restaurantsin Chicago. Two final sources, Centerstage(http://centerstage.net} and Citysearch(http;//chicago,citysearch.com), which pro-vide comprehensive online guides to Chi-cago's restaurants, were used to identify fast-food restaurants that might have beenmissed by the prior sources. Finally, we ob-tained full addresses for identified restau-

September 2005, Vol 95, No. 9 I Amertcan Journal of Public Health Austin et al. \ Peer Reviewed ! Research and Practice I 1575

RESEARCH AND PRACTICE

rants with the online Yahoo! Yellow Pages(http ://www,yahoo.com} and Superpages(http://www.superpagcs.com). A privatecompany that was shown in prior research toprovide highly accurate geocoding for ad-dress databases geocoded the restdting data-hase of 624 last-food sites to assign longi-tude and latitude coordinates and censustract identification codes,* " All records weresuccessfully geocoded, with 98"/i) (n^610)matched to the street address, the best cen-sus code accuracy rating. Eleven sites weredetermined to he located outside Chicagodty limits and were excluded, resulting in613 fa.st-food restaurant sites within Chicagoincluded in the analyses.

Restaurants in the database included 79restaurant chain brands, which offered menuitems such as hatnburgers and french fries.fried chicken, pizza, submarine sandwiches,and doughnuts. The top 10 restaurant chainshy numher of sites were McDonald's (99locations; 16"/o of total). Subway (90; 15%),Dunkin- Donuts (72; 12%). KFC (50; 8"/a),Burger King (41; 7%), Popeye's Chickenand Biscuits (38; 6%), Wendy's (22; 4%),Church's Chicken (19; 3%}. Taco Bell (19;3%), and WTiite Castle (16; 2%). These 10chains constituted 76% of our database.

School DatabaseIn 2002, we used school listings from the

National Center for Education Statistics(http ;//nces.ed,gov) and Chicago I'ublicSchools to create a comprehensive list of Chi-cago puhlic and private kindergarten and pri-mary and .secondaiy school names aitd ad-dresses. A combined database with 1478listings was geocoded, with 91*'/o (n— 1351)of the records matched to the street address.It was not possible to match 6% (n^85) ofthe schools, .so they were excluded from theanalyses. An additional 101 records were de-tennined to be duplicates or located outsideChicago and were excluded. Analyses in-cltided 1292 schools within Chicago.

Statistical AnalysesUsing several methods, we examined loca-

tional patterns of fast-food restaurants andschools in Chicago. First, we calculated themean and median distance from schools tothe nearest fast-food restaurant. Second, we

used the mapping and analytic capabilitiesof a geographic infonnation system rtuiningArcGlS 8.3 software^' to create 400-m-raditisand 800-m-ntdius buffers around each schooland then calculated the number of fast-foodrestaurants located within buffers of each size.The 400-m buffer bas been used as a primaryunit of aggregation in research on walking andthe environment, on the hasis of an estimatethat on average an adult can walk 400 m in5 minutes.' "* We also created 800-m-radiusbuffers to characterize a larger section of theschool neighborhood food environment.

TTiird, we used the bivariate K functionmethod to quantify the degree of clustering(spatial dependence) of fast-food restaurantsaround school locations at distances rangingfi-om 0 to 1.5 km from school addresses.Cuthbeil and Anderson"' provide a detaileddescription of the hivariate K function nietiiodand its application in analyses of commercialand residential land use, A strength of theK function method is that it provides a statisti-cal test of clustering. The bivariate K functionmethod estimates the spatial dependence be-tween 2 different types of locations in a geo-graphic region, identifying distances withinwhich spatial dependence is evident and test-ing the statistical significance of the obsei vedclustering relative to expected distribution ofpoints if there were no spatial dependence,""''We used the Splancs program ** ^ of theR statistical package'" for the K functioncalculations.

We analyzed data from the city as a wholeand then conducted 3 additional stratifiedanalyses. Fii"st, we examined spatial depen-dence within downtown Chicago (69 schoolsand 138 fast-food restaurants) and outsidedowntown (1223 schools and 475 fast-foodrestaurants). We defined downtown Chicagoas a 3-km-raditis circle centered at approxi-mately the intersection of Dearborn and Mon-roe streets. We modified data from the IllinoisDepaitment of Natural Resources to definethe city boundaries as a dosed, contiguotispolygon.'^ Second, we examined sfiatial de-pendence within 3 strata defined as high{314 schools), medium (479 schools), andlow (498 schools) level of commercialization.Level of commercialization was hased on1995 Nortliern Illinois Planning Commissiondata, the most recent data availahle at the

time of our study.•"' Commercial land in-cluded shopping malls; business parks; offices;hotels; government administration and ser-vices; prisons; and medical, institutional, in-dusti-ial, and transportation facilities. We cal-culated the percentage of commercial land in1990 census tracts and divided the tracts intotertiles, categoiizing them as high (>30.3%),moderate (14,6"/i) to 30.2"/i)), and low(< 14.6%}, We then identified the schoolsthat fell in each of these teriiles. Thii-d, weexamined spatial dependence within 3 strataof median annual household income, basedon 2000 US Census data, which we groupedinto tertiles, defined as high- (441 schools;>$43 700), moderate- (423; $30 300 to$43 700), and low-income (428; <$30 300)regions." We assigned each school a medianhousehold income value hy calculating anarea-weighted average of block groups thatintersect the school buffer.

It is important to note that our study data-bases were assembled as censuses, ratherthan samples, of schools and fast-food restau-rants in Chicago; therefore, it was not neces-sary or appropriate to carry out statistical testsof strata differences in meaas and frequenciespresented in Tables 1 and 2. In contrast, statis-tical tests were appropriate for the K functionanalyses, which examine differences betweenobserved and expected clustering.

RESULTS

Figure 1 shows the location of fast-foodrestaurants and schools incltided in analysesand indicates schools that have at least 1 fast-food restaurant within 400 or 800 m. Table 1presents the numher of schools and fast-food restaurants and the distance from anyschool to the nearest fast-food restaurant inChicago overall and within each stratum ofneighborhood characteristics. In Chicago asa whole, the mean and median distancefrom any school to the nearest fast-foodrestaurant were 0.60 and 0.52 km, respec-tively (Table 1},

Table 2 shows the means, medians, andranges for the number of fast-food restaurantsin 400-m-radius and 800-m-radius buffersaround schools. In Chicago as a whole, thenumber of fast-food restauranls within 400 mof a school ranged Irom 0 to 33, whereas the

1576 I Research and Practice | Peer Reviewed | Austin et at. American Journal of Pubijc Health \ September 2005, Vol 95, No. 9

RESEARCH AND PRACTICE

Schoolso No Restaurants within 800 m

• At least 1 restaurant within 800 m

• At least 1 restaurant within 400 m

^^ Fast food restaurant

^ H Downtown Kilometers

FIGURE 1-Locations of schools and fast-food restaurants in Chicago.

number within 800 m of a school rangedfrom 0 to 85. Thirty-live percent (452/1292)of schools in Chicago as a whole had at least1 fast-food restaurant within 400 m. whereasnearly 80% (1010/1292) of schools had atleast 1 fast-food restaurant within 800 m.approximately a 10-minute walk. Two schoolshad the same address as the nearest fast-foodrestaurant. In the downtown region, therewere twice as many fast-food restaurants asthere were schools, and 94% (65/69) ofschools had at least 1 fast-food restaurantwithin 800 m. For schools within regionsstratified by commercialization and median

jinnual household income, the percentage ofschools with at least 1 fast-food restaurantwithin 400 m ranged from 27'Vii to 40%and within 800 meters ranged from 76%to 80'yo (Table 2).

Figure 2a displays the results of bivariateK fimction analyses for Chicago as a whole.The value of the K function (shown on they-axis) indicates the average number of fast-food restaurants within a specified distancefrom a school (shown on the x-axis), dividedby the overall density' of fast-food restaurantsin Chicago. The solid line represents the ob-served value of the K function, and the dotted

line represents the expected value of theK Hinction under the assumption of no spatialdependence between fast-lbod restaurantsand schools. The dashed lines represent theupper and lower bounds of the 95*'/() confi-dence interval around the expected-valuedotted line. Confidence intervals are basedon 100 simulations with Ihe data under theassumption of no spatial dependence. Theobserved value of the K function in Figure 2afalls above the 95"/ii confidence interval be-tween distajices of roughly 0.25 km and1.5 km on the x-axis. indicating tliat a signifi-cantly gi-eater number of fast-food restaurantsare located within a short distance fromschools than would be expected if there wereno spatial dependence. Figure 2b di.splays aplot of the relative clustering of fast-foodrestaurants around schools within the corre-sponding distance on the x-axis ranging fromzero to 1.5 km. We estimate that there arebetween 3 and 4 times as many fast-foodrestaurants within an area up to 1.5 kmfrom schools than would be expected if therestaurants were distributed throughout thecity in a way that was unrelated to the loca-tion of schools.

Results of stratified analyses revealed statis-tically significant clustering of fast-foodrestaurants within 1.5 km of schools locatedwithin ai'eas of the city outside downtown(not shown). The relative clustering estimateindicated that fast-food restaurants wereabout 2.5 times more clustered aroundschools in areas outside downtown thanwould be expected under the null hyjjothesis.Although fast-food restaurants were muchmore numerous in the downtown area thanoutside downtown, tliey did not appear tocluster significantly arouiid schools more thanwould be expected given the high density offast-food restaurants in downtown Chicago.

Fast-food restaurants were found to clus-ter significantly around schools in both thebigh- and moderate-commercialization re-gions of the city (not shown). The relativeclustering estimates indicated that in thehigh-commercialization areas, there wereapproximately 6 times more fast-food restau-rants within 1.5 km of schools, and in themoderate-commeirialization areas, tlierewere approximately 3 times more fast-foodrestaurants witbin 1.5 km of schools than

Septembef 2005. Vol 95. No. 9 : American Journal of Public Health Austin et al. \ Peer Reviewed | Research and Practice | 1577

RESEARCH AND PRACTICE

TABLE 1-Number of Schools and Fast-Food Restaurants and Distance in Kilometers From

Any Schooi to Ciosest Fast-Food Restaurant, Overall and by Neighborhood Characteristic

Strata: Chicago, 2002

Chicago overall

Downtown

Outside downtown

High commerciaiization

Medium commercialization

Low commercialization

High median household income

Moderate median household income

(S30300-$43700)

Low median household income

(<$30300)

Schools

No,

1292

69

1223

314

479

498

441

423

428

Restaurants,

No.

613

138

475

305

177

131

282

296

27

Mean

0,60

0.32

0.62

0.61

0.56

0.64

0.61

0,64

0.60

Distance,

SD

±0.45

±0,22

±0.46

±0.59

±0.40

±0.39

±0,44

±0.52

±0,44

lim

Median

0.52

0,27

0,53

0.49

0.47

0.61

0.53

0,55

0.51

Range

0-3.69

0-0.83

0-3,69

0-3,69

0-3.25

0-3,20

0-3.20

0-3,69

0-3,61

deludes 8 restaurants in census tracts with no information on median annual household income.

v/ould be expected. We did not find evidenceof du-stering of fast-food restaurants aroundschools in low-commercialization areas.

Fast-food restaurants were fonnd to clus-ter significantly around schools in regions ofthe city in the highest income teitile, wheremedian annual household incomes were$43 700 or greater (not shown). The relative

clustering estimate indicated that in theseareas, there were approximately 5 to 7 timesmore fast-food restaurants within 1.5 km ofschools than would he expected. We did notfind evidence of clustering of fast-foodrestaurants around schools in areas wherethe median annual household income wasless than $43 700.

TABLE 2~Number of Fast-Food Restaurants Within 400 and 800 m of Any School,

Overali and by Neighborhood Characteristic Strata: Chicago, 2002

DISCUSSION

Fast-food consumption has increased dra-matically over the past several decades andmay be an important contrihulor to the risein the prevalence of obesity in children andadolescents. In our study of fast food in Chi-cago, we found thai although fast-food restau-rants are located throughout tlie city, they areclustered in areas within a short walking dis-tance from schools. We estimate that tliereai'e 3 to 4 times as many fast-food restaurantswithin 1.5 km from schools than would be ex-pected if the restaurants were located aroundthe city in a way unrelated to schools. Themedian distance from any school to the near-est fast-food restaurant was 0.5 km, indicatingthat in half the city's schools, students need towalk little more than 5 minutes to reach afast-food restaurant. Nearly 80% of schoolsin Chicago had at least 1 fast-food restaurantwithin 800 m. In the downtown area, themedian distance from any school to the near-est fast-food restaurant was just 270 m. Wealso found evidence of clustering of last-foodrestaurants around schools in areas of the cityoutside downtown and within areas with highand medium levels of commercialization andareas where mediaji annual household in-comes were $43 700 or greater.

Restaurants Within 400 m of a School, No. Restaurants Within 800 m of a School, No.

Mean SD

Schools With at

5th 95th Least 1 Restaurant

Range Percentile Median Percentiie Within 400 m, No. {% Mean SD

Schools With at

5th 95th Least 1 Restaurant

Range Percentile Median Percentile Within 800 m, No. (%)

Chicago overall

Downtown

Outside downtown

High commercialization

Medium commerciaiization

Low commercialization

High median household

income [> $43 700)'

Moderate median househoid

income ($30 300-43 700)

Low median household income

(<S30300)

0.9

6.60.5

1,4

0,9

0.4

1.5

0.5

0.6

±2,9

±10.1

±1.0

±4.7

±2.5

±0.9

±4.5

±1.4

±1,0

0-33

0-33

0-7

0-33

0-26

0-7

0-33

0-17

0-7

0

0

0

0

00

0

0

0

0

1

0

0

0

0

0

0

0

3

29

3

4

3

2

4

2

3

452 (34,9%)

46 (66,7%)

406 (33.2%)

124 (39.5%)

191 (39.9%)

136 (27.3%)

168 (38.1%)

124 (29,3%)

160 (37,4%)

^Excludes 8 restaurants in census tracts with no information on median annual household income.

3,1

19.9

2.1

4.4

3.3

1.9

4.7

2.4

2.0

±7,3

±25,2

±2,1

±11.5

±7,1

±2.0

±11,5

±4,3

±1,9

0-85

0-85

0-19

0-85

0-66

0-18

0-85

0-58

0-18

0

1

0

0

00

0

0

0

2

6

2

2

2

1

2

2

2

7

70

6

11

7

6

19

6

6

1010 (78,2%)

65 (94,2%)

945 (77.3%)

248 (79.0%!

382 (79.7%)

379 (76,1%)

337 (76,4%)

330 (78.0%)

343 (80.1%)

1578 I Research and Practice | Peer Reviewed | Austin et al. American Journal of Public Health I September 2005, Voi 95 , No. 9

RESEARCH AND PRACTICE

Observed K functionExpected K function95% Cl under null hypothesis

(no clustering)

0.0 0-5 1.0

Distance From a School, km

1-5

10-

•ii 2-

6 -

4 -

0.0 0.5 1.0

Distance From a School, km

1.5

FIGURE 2-Clustering of fast-food restaurants in Chicago: (a) bivariate K function plot with

95% confidence intervals showing ciustering of fast-food restaurants around schools and

(b) relative concentration of fast-food restaurants around schools.

Students in all grades may have access tofast-food restaurants before and after school,and older students are also likely to have ac-cess at iLinchtime. because of widespread

open-camptis policies in Chicago high schools.Approximately 6"'o of students attending Chi-cago public schools are provided transporta-tion in school buses, whereas most walk, take

public transportation, or ride in cai-s to andfrom school (G. Mancuso, director. SchoolDemographics and Planning, Board of Educa-tion of the City of Chicago, e-mail correspon-dence, January 6, 2005), These modes oftransportation may give schoolchildren oppor-tunity to access last-food restaurants in theirschool neighborhoods.

Our methods were designed to characterizethe degree of fast-food clustering aroundschools. The specific mechanisms underlyingthe patterns ol' clustering we observed are notclear. The degree to which restaurant indus-try site selection practices, dty zoning regula-tions on commercial land use, or other factorsmay contribute to the patterns will need to beexplored in future studies. Nevertheless, theconcentration of fast-food restaui'ants aroundschools within a shoit walking distance forstudents is an important publie health con-cern in that it represents a deleterious influ-ence in the food environment that may un-dermine public health efforts to improvenutiitional behaviors in young people.

We found clusteiing of fast-food I'estaurantsin both moderate and high-commercializationareas, indicating that proximity to fast-foodvenues afiects the majority of schools in ourstudy, not just those in the most commercial-ized ai'eas. In fact we found more evidenceof clustering outside downtown than inside,perhaps because the downtown area is sosuffused with fast-food restaurants that thesevenues are present in most neighborhoods.Of note is our finding of a small number offast-food restaurants (Table 1) in the lowest-income-tertile areas (median household in-come <$30300), It is plausible that fast-foodrestaurant owners are reluctant to locate inimpoverished areas. It was in the highest-income-tertile areas (median household in-come >$43 700) where we found evidenceof fast-food clustering, suggesting that chil-dren attending schools in these areas of thecit}' rather than in low-income areas may bemost exposed to the problem of concentratedfast-food venues in school neighborhoodfood environments.

The neighborhood food environment is arelatively new coneept in public health re-search, and methods for defining, character-izing, and quantifying the food environmentare still under development. Studies have

September 2005, Vol 95, No. 9 , American Journal of Public Health Austin et al. I Peer Reviewed I Research and Practice I 1579

RESEARCH AND PRACTICE

used varied methods to characterize fastfood in neighborhood environments. For in-stance, in a study with geocoded data fromthe Cincinnati area, Burdette and Whitaker'^estimated that the average distance from achild's home to the nearest fast-food restau-rant was 0.7 miles. In a study using geo-coded 1990 US Census data from 4 states,Morland et al.'' characterized food environ-ments in terms of the mean number of fast-food restaurants per census tract, finding anestimated 2 fast-food re.staurants per tract.We ai-e not aware of studies similar to ourethat used spatial statistical methods to quan-tify the degree of clustering of fast-foodrestaurants around schools. We believe nurmethod represents a novel and potentiallyinformative and powerful technique forquantifying and statistically testing spatialpattems in the food environment and landuse in relation to important public healthconcems.

We were not able to identify any acceptedcriteria in the public health literature to de-fine fast-food restaurants. As described previ-ously, to create our fast-food database, we re-lied on infonnation from the food industrymarketing research lirm Technomic Inc andother sources providing lists of restaurants inChicago. For their analyses, Morland et al.defined fast-food restaurants as those listed inthe limited-service restaurant category of the1997 North American Industry ClassificationSystem." Their category mcludes fast-foodfranchises and other restaurants that servefood over the counter and encompasses anarray of restaurants veiy similar to that iden-tified by our Technomic source. One differ-ence is that the Technomic category limited-service restaumni includes doughnut shojjsamong iiist-ibod restaurants, whereas theNorth American Industry Classification Sys-tem does not^' In a Cincinnati-based study,Burdette et al.'' defined fast-food restaurantsas those that had franchises in multiple states,had more than 1 restaurant in Cincinnati,served meals without waiters, and providedseating for customers. They identified 8 chainre.staurant brands in Cincinnati, 7 of whichwere also included in our database lor Chi-cago. To the extent that definitions of fast-food restaurants differ across .studies, compar-isons oi' findings are limited. Our study was

restricted to schools and fast-food restaurantsin Chicago. Further research will need to ex-plore to what degree other cities and ruraland suburbati areas ditler from what we ob-served in Chicago.

Our study has several limitations. The mostrecent land use data available to us to classifylevel of commercialization were compiled in1995 by the Northem Illinois Planning Com-mission, but more recent data might have al-lowed more accurate classification. There isprecedent in tlie health research literature forusing 400-m-radius buffers for analyses ofjihysical environments in part because it is adistance easily walked by most adults in 5minutes.""' but young children may be ex-pected to take longer to walk this distance.We were not able to assess the relation be-tween density of fast-food restaurants inschool neighborhoods and student nutritionalpattems. Burdette and Whitaker'^ did notfind an association between proximity of fast-food restaurants to a child's home and thechild's weight status, although they did notassess nutnticmal behaviors. Future i-esearchwill need to examine the relation between thepi-esence of fast-food restaurants in schooland home neighborhoods and fast-food con-sumption, diet quality, and caloric intake.

A groundbreaking report on childhoodobesity issued by the Institute of Medicine in2004 calls on the food industry to voluntarilyrestrict advertising unhealthfii! foods to chil-dren." In addition, the Institute of Medicinereport recommends that schools improve thenutritional quality of foods serv'cd and re-strict sales of sodas and nonnutiitious snackfoods.'''' Other school policy changes that maywaiTant coTisideration include restricting un-healthful foods that can be brought into theschool by students and curtailing open-campasrules that allow students access to nearbyfast-food restaurants at lunchtime. Our find-uigs suggest that additional municipal or statepolicy initiatives may be needed to addressthe concentration of fast-food venues inneighborhoods sun'ounding schools.

Ashe et al.'" argued that there is legal[precedent for local governments to imposestricter controls on fast-food restaui-ant sites,similar to site selection restrictions placed onalcohol and firearm vendors, such as distancelimits from schools and playgrounds and

caps on the number of venues permitted ina neighborhood, N'ew regulation initiativessuch as these may be a way to remove nox-ious elements in the food environments thatschoolchildren arc exposed to every day.They may also help to spur the fast-food in-dustry to improve the nutritional quality ofits products to avoid zoning restinctions.Greater puhlic attention to the proximity offast-food restaurants to schools may add ur-gency to eflbrts to improve food environ-ments for schoolchildren. •

About the Authors5. Bryn /iHsrm i.s with the Division of Adolescent andYoung .'{dull Medicine al Ckildren s Hospilul. Boston,Mass, und the DqMnmenl of Society. Human Dtvelopmenl.und Health. Harvard Schtiul of Public Health. Boston. .4«mPittel is with the Division of .Adolescent and Young .XduUMedicine ai Children s Hospital. Stei'enf .Mdlif is with lheDepartment of Biostalistics und the Department ofEnm-ronmeiitul Health at the Harvard Schoal of Public Health.Boston lirisa N. Sanchez is wilh the Department of Bio-statisfics at the Harvard School uf Public Health. StephenBuka and Steven L Gortmaker are with the Dejiartmentof Society, Human Development, and Health at the Har-vard Sc'hdcil of Public Health.

Requests for reprints should be sent in S. Bryn AiL-itin.SeD, Division of Adolescent and Young .'\dult Medicine.Children's Ho.spital. .100 Longivotid .-Iw. liostcm, MA02115 (e-mail. hryn.aiLstin@chMrenH.harvard.eduj.

Tliis urticle wcvi ciccejyiedjanitury 22, 2005.

Conta'ibutorsS. B. Austin was responsible for study conception, dataanalysis and interpretation, and artirlo preparation.SJ. Melly provided gcogfaphic iiifonnadon systems expert-ise and carried out data analysis. B. N. Sanchez provided.^taastifal expertise. S.j. Melly. B.N. Sanchez,. A. Patel.S, Buka. and S. I.. Gortmaker contributed to study con-ception, data inteipretation. and cntical revision ofthe article.

Human Participant ProtectionNo protocol appnnal was needed for this study.

AcknowledgmentsI his pii)|etl was supported hy a grant fnirti thi' Centersfor Di.sease Control anci Prevention, Prevention Re-search Centers (grant LI48/CCU115807). S.B, Austin Issupported by the Leadership ["education in AdolescentHealtli project, Matemal and Child I lealth Bureau, USDepartment of Health and Human Scr\-ices (6T7I-MC00009-12-(ll), B.N. Sanehe? is supported by apredoctoral fellowship in biological saeiices from (heI Inward Hughes Medical Institute and the Knvimnmen-tal Statistics Core of the National Institute of Environ-mental Health Sciences center (grant KS00002).

The aiitliors thank LoiiLse M. Ryan, KathyMcGafiigan, Beth E. Molnar, Angie Cradock, I'hili])Tmped, Jean VVieeha, and Jennifer J, Lander fortheir contributions lo otir study.

1580 I Research and Practice I Peer Reviewed | Austin et al. American Journal of Putilic Health | September 2005, Vol 95, No. 9

RESEARCH AND PRACTICE

Note. 'l"his w()rk is solely the responsibility of (lieanthiirs and does not necessarily represent the officialviews of the Centers for Disease Control and Preven-tion or other granting institutions.

References1. Technomic Foodservice Segment Time Series: LimitedService Restaurants (1975-2005). Chicago, nt: Tech-nomic Inc; 2004.

2. Technomic Web site, October 1. 2004. Availableat; http://vvww.technomic.com. Accessed October 1,2004,

:i, Nielsen .Sj, Siega-Riz AM, Popkin BM, Trends infood localions and sources among adolescents andyoung adults, PrevMed. 2002;35:I07-113,

4, Oauson A. Share of food spending for eating outreaches 47 percent. Food Rev 1999:22:20-22.

5, Nestle M. Food Politics: How the Food Industry In-fluences Nutrition und Health. Berkeley, Calif: Universityof California Press; 2002,

6, Schlosser K. Fast Food Nution The Dark Side of theAll-American Meal New York; Houghton Mifllin; 2001,

7, Bowman S,'V Gortmaker SL, Fibbeling CA, PereiraMA, [Ardwig DS. EffccLs of fas!-food consumption onenergy intake and diet qiialitv' among children in a na-tional hdusehold sur-vey. Pediatrics. 2004;113;112-l!8,

8, YoLing LR, Nestle M. 'ITie contribution of expand-ing poition sizes lo the US obesity epidemic, Amf Pub-lic Health. 2002;92i246-249,

9, Paeratakiil S. Ferdinand DP, Champagne CM,Ryan DH, Uray GA. Past-fuod consumption among USadtilts and children; dielary ajid nutrient intake profile./Am Diet Assoc. 2()03;103:n.12-1338,

10, Jeflery RW, Erench SA, I'.pidemic obesity in IheUnited States; are fast foods and television viewingmntnhiitin^? AmJ Public Health. 1998;88;277-280,

11, Erench S.\. i Jamack I,. Jetfery RW, East food res-taurant tise among women in the Pound of Preventionstudy: dielary, behavioral and demographie correlates.Int f Obesitt/. 2OOO;24;U53-1359,

12, iVreira MA, Kartashov Al. Ebbeling CA. et al,East-fofid habiK weight gain, and insulin resistance(tbe C','\RDIA study): 15-year prospective analysis.Lancet 2 00 5 ;3 65:3 6-42.

13, Edelstein SL. Knowler WC, Bain RP, et al. Predic-tors of pn)gres,sion from impaired gltic-ose tolerance toNIDDM: an analysis of .six prospective studies, Dia-l>e!es. 1997;46;701-7I0,

14, Melaniphy JC. Restaurant and Fust Food Site Selec-tion New York: John Wiley & Sons; 19y2,

15, Solomon Kl, Katz N, Profitable Restaurant Manage-ment. Englewood Cliffs, N]; Prentice Hall; 1981,

Hi, McKenna EX, Starting and Managing a SmallDrive-m Restaumnt. Washington, DC; Smali BusinessAdministration; 1972.

17 Bunleite HL. Whitaker RC, Neighborhood play-grounds, fast food restaurants, and erinie: relationshipsto overweight in low-income preschool children, PrevMed 2004;36:57-63,

18, Ashe M. Jemigan D, Kline R, Galaz R. Land useplannmg and the control of alcohol, tobacco, firearms,and fast food restaurants, .AmJ Public Health. 2OU3;93:1404-1408,

19, Morland K, Wing S. Diez Ronx A. ftjole C, Neigh-borhood characteristics associated with the location offood stores and food service places, AmJ Prev Med.2OO2;22;23-29.

20, Heidpath DD, Bums C, Gan-ai-d J, Mahoney M,Townsend M, An ecological study of the relationshipbetween social and environmental detenninanLs ofobesity. Health Place. 2002;8;141-145,

21, Technomic Inc. Technomic lop 100. Available at;https://technomic,securelook,comtechnom ic/co vers_pdf/toplO(),pdf. Accessed July 6, 2004.

22, Krieger N, On the wrong side of the tracts? Evalu-ating Ihe accuracy of geocodirig m public health re-seareh. Am j Public Health. 2001;91;1114-l!16,

23, ArcGIS [comptJter progiram]. Version 8,3, Red-lands. Calif: Environmental Systems Research Institute;2003.

24, Pikora TJ, Bull LCL, Janiro/,ik K, Knuiman M,Giles-Corti B, Donovan RJ, Developing a reliableaudit instrument to measure the physical environ-ment for physical activity, AmJ Prev Med. 2002;23;187-194.

25 Ctithberl AL, Anderson WT, Using spatial statis-tics to examine the pattems of iLrban land cievelopmenlin Halifax-Dartmouth, Professional Geographer. 2002;54:521-532.

26, Howlingson B, Diggle F. Splancs: spatial point pat-tem analysis eode in S-PltLs. Comput Geosdence. 1993;I9:(S27-B55.

27, Splancs Package. Spatial and Space-Time PointPattem Analysis, t'ersron 2.01-9 Icomputer pn>gram|.2002, The R Eoundation for Statistical Computing,Availahle at: http://cran r-pniject org/src/eontrib/Descriptions/splanL-s,btml.

28, Venables WN, Smith DM. R Development Coreleam. An intn»duction to R. a programming environ-ment for data analysis and grapbics, version 1,6.2, TheR Eoundation fnr Statistical Comptiting. Available at:bttp://cran.r-project,org/doc/m an uals/R-intro.pdf. Ac-cessed SepL 21, 2004,

29, Illinois Department of Nattiral Resourees. Illinoisnatural resources geospadal data clearinghouse. Avail-able at: http://www,isgs,tiiucedu/nsdihome/ISGSin-dex html. Accessed Sept. 21. 2004.

30, Northeasteni Illinois Planning Commission Website. Northeastern Illinois Planning Commission's 1;24000-Scale 199(1 Land Use Inventory (48 Cate-gories! for Noriheastem Illinois, GtS Coverage, Ver-."iion 3.0 (updated using 1995 Fhotos). Available at;http;//www,nipcorg and http://www,nipc,org/test/dnir_cd,htm. Accessed May 27, 2005,

31, Census 2000. Washington, DC; US Census Bu-n?aui 2003,

32, Morland K. Wing S. Diez Rtitix A, The contextualeffect of the local food environment on residents' diets:the Atherosclerosis Risk in Communities sttidy, Amf Pub-lic Health. 2002,92:1761-1767

33, US Census Bureau. 2002 NAICS Definitions..Available at. http //www.censtis.gov/epcd/naicsO2/def/ND722211,HIM. Accessed July (i, 2004.

34, Institute of Medicine. Preventing Childhood Obesity:Health in the Balunce. Washington, DC; National Acad-emy of Sciences; 2004,

MlCKOiiiDLUUlCALEXAMINATION

OF FOODS FourthEdition

Compendium of Methods for the

MicrobiologicalExamination of FoodsEdited by Frances Pouch Downes

and Keith Ito

The fourth edition of the CoinpfiuiiiiiiidcKumentb the latest, state-of-the-art

technologies in microbiology and molecu-lar biology that play a key role—from pro-duction to consumption—in the safety offood.

This book brings together more than200 experts from business, governmentand academic to address key issues suchas qualitv' assurance, common pathogens,laboratory procedures, shipment practicesand the safety of specific foods.

This book is vital to all those involvedin I Focxi prcxresslng and manufacturing IFood testing for safety and quality IPathogen sur\'eillance I Laboratory scienceI Public health safety

1SBNO-87553-175-X200) I 704 pages I Hardcover

$87,50 APHA MembersS125.00 Nonmembers

plus shipping ond handling

Bulk Order Discounts andExamination Copies are available.

ORDER TODAY!

American Public Health AssociationPublication SalesWeb: www,apha,orgE-mail: APHA^pbd,comTel: 888-320 APHAFAX: 888-361 APHA

September 2005. Vol 95, No, 9 | American Journal of Public Health Austin et al. I Peer Reviewed I Researeh and Practice I 1581