proximity of fast-food restaurants to schools and...
Post on 03-Feb-2020
4 Views
Preview:
TRANSCRIPT
RESEARCH AND PRACTICE
Proximity of Fast-Food Restaurants to Schools andAdolescent Obesity
Brennan Davis, PhD, and Christopher Carpenter. PhD
The marketing of food to chiJdren is in thenadona] spoUi^t as rates of childhood obe-sity rise in the United States. More than 9 mil-lion US children and adolescents are obese, andjust as many are at risk of becoming obese.' Theconsequential health risks include asthma, hy-pertension, type 2 diabetes, cardiovascular dis-ease, and depression, Fast food consumption by2- to 18-year-olds increased 5-fold from 1977 to1995; by the latter year, fast food was consumedat 9% of eating occasions and comprised 12% ofdaily cdoric intake,' Almost one third of allyoutlis now eat at fast-food restaurants on anygiven day.'* One study reports that weekly eon-sumption of fast food by young adults is directlyass(xiated with a 0.2-unit increase in body massuidex (BMI).
Despite the possibility that proximity of fast-food restaurants to schools affects children'sheaith, research has not yielded consensus onthis issue. Multiple studies have found that fast-food restaurants are systematically concen-trated within a short walking distance ofschools, giving children greater access to low-quality food, but these studies do not make anexplidt connection between proximily to fast-Ibod restaurants and diet-related outcomes. *^Studies that have examined possible assodationsbetween the density of fast-food outlets andoutcomes such as food consumption and weightstatus among youths have not found a relation-sliip. •' Our study revisits these questions usingnew detailed data on youths in Califomia
METHODS
Our empirical approach related the presenceof fast-food restaurants near schools to youths'wei^t status and food consumption. We usedinfonnation from individual-level student re-sponses to the 2002-2005 Califomia HealthyKids Survey (CHKS).'" The CHKS is an anony-mous, school-based survey consisting of a coreset of questions and several topical modules thatfocus on ^edfic health-risk behaviors. TheCalifomia Department of Education requires
Objectives. We examined the relationship betv /een fast-food restaurants nearschools and obesity among middle and high school students in California,
Methods. We used geocoded data (obtained from the 2002-2005 CaliforniaHealthy Kids Survey) on over 500000 youths and multivariate regression models toestimate associations between adolescent obesity and proximity of fast-foodrestaurants to schools.
Results. We found that students with fast-food restaurants near (within onehalf mile of) their schools (1 ) consumed fewer servings of fruits and vegetables, (2)consumed nnore servings of soda, and (3} were more likely to be overweight (oddsratio [0R| = 1.06; 95% confidence interval [Cl] = 1.02,1,10) or obese (OR = 1.07; 95%CI = 1.02, 1,12) than were youths whose schools were not near fast-food restau-rants, after we controlled for student- and school-level characteristics. The resultwas unique to eating at fast-food restaurants (compared with other nearbyestablishments) and was not observed for another risky behavior (smoking).
Conclusions. Exposure to poor-quality food environments has important ef-fects on adolescent eating patterns and overweight. Policy interventions limitingthe proximity of fast-food restaurants to schools could help reduce adolescentobesity, {Am J Public Health. 2009;99:505-510. doi:10.2105/AJPH,2008,137638)
middle schools and high schools in California toadminister the CHKS for œmpliance with pro-visions of the No Child Left Behind Act of 2001(PubLNo. 107-110),and the sampling is designedto produce estimates that are representative atthe district level. Consequently, the CHKS pro-vides very large sample sizes; our data includeinformation on over a half million students.
Our primary outcome of interest was BMI(defined as wei^t in kilograms divided byhei^t in meters squared). We also consideredhinary outcomes for overweight and obesity.The obesity measurements of those youngerthan 19 years were based on percentiles by ageand gender reference grou[), according to theBMI-for-age percentiles chart published hy theCenters for Disease Control and Prevention(CDC)," A child at or above the 85th perci n-tile of BMI distribution by age and gender wasconsidered overwei^t A child at or above the95th percentile was considered obese (andoverweight).
We also created an indicator variable fordrinking any soda in the last 24 hours. Similarvariables were created as indicators for theconsumption of any vegetables, juice, fruit, andMed potato foods. We also measured the
number of servings of each food type that ayouth reported consuming in the past 24hours.
To measure the proximity of fast-food out-lets to schools, we used (1) a database oflatitude-longitude coordinates and otherschool mformation for middle and high schoolsfrom the California Department of Education,'^(2) a database of restaurants in Califomia in2003 with latitude-longitude coordinates fromMicrosoft Streets and Trips (Microsoft Corpora-tion, Redinond, WA), and (3) a list of restaurantbrands classified as "top limited-service restau-rants" by Technomic Inc, a food industry con-sulting firm.''
Using these data, we created the followingindicator; a youth who attended a school lo-cated within one half mile of at least 1 restau-rant whose brand was on the list of top limited-service restaurants was considered near a fast-food restaurant. Previous research has alsoused the one half mile measure of proximity. 'A pereon can walk this distance in 10 minutes.We also created the variable "near other res-taurant" to indicate that the student's school wasnear a restaurant not on Technomic's list of toplimited-service restaurants. Most restaurants in
March 2009, Vol 99, No, 3 | American Joumal of Public Health Davis and Carpenter \ Peer Reviewed | Research and Practice | 505
RESEARCH AND PRACTICE
Ulis latter category were probably nonchain,limited-service restaurants or smaller-chain.limited-service restaurants whose total US saleswere not high enou^ to be on the list of toplimited-service restaurants but th^ probablycatered to youths in a manner similar to thai of(he larger fast-food cbain restaurants. Given ourinability to precisely identify the type of theseotlier restaurants, bowever, we bave focused ontbe results for being near a fast-food restaurant
We estimated standard multivariate regres-sion models that linked adolescent obesity out-comes to variables measuring the proximity offast-food establishments to tbe student's school.The dependent variables were BMI, over-weigbt, various food consumption outcomes,and obesity. For the BMI outcome, we esti-mated ordinary least squares regressionmodels; for the dicbotomous overweight andobesity outcomes, we used logistic regressionand present adjusted odds ratios. Tbe inde-pendent variables we controlled for includedindicator variahles for the following: femalegender, age category (<12,13,14,15,16, or>17years), grade {<7,8, 9,10,11, or 12), and race/ethnicity (White, Asian, Black, Hawaiian, His*panic American Indian, multiple race, or otber).
We also controlled for various measures ofthe youth's physical activity and exercise regi-men (number of days in the previous week therespondent engaged in vigorous physical acdv-ify and number of days of musde-strengtheningactivities), which should be directly assodatedwith weight status outcomes. In addition, wecontrolled for scbool and other contextualcba]-acteristics, including indicators for schooltype (high scbool vs middle school), the pro-poridon of students eligible for free or reduced-price meals, school enrollment, indicators forschool location types (separate indicators forlarge, medium, and small urban locations;large, medium, ajid small suburban locations;town location; and rural location), and countyindicators. In all models, we also indudedcontrols for the survey wave (2002-2003,2003-2004, or 2004-2005).
All analyses were performed vAth Stata 10.0software (StataCorp LP, College Station, TX),which takes the complex sampling design anddifferentia] respondent wei^ts into accountwhen calculating standard errors. We allowedfor the arbitrary correlation of errors acrossstudents within a scbool by correcting our
standard errors at the school level with Stata'sCLUSTER con-ection. In addition to the base-line models with the hill set of controls alreadydescribed, we also investigated effects for spe-cific demograpbic subgroups such as racial andetbnic mmorities.
Being near a fast-food restaurant was cbosento be consistent witb previous research, • Toexamine tbe sensitivity of our results to otberplausible ways of measuring proximi^, we alsoperformed additional tests. For example, wetested (he assodation between students' BMI andproximify of tbe nearest fast-food restaurât totheir schools with the following mutually exdu-sive categories of proximity: (1) within onequarter of a mile (400 m), (2) between onequarter and one half mue, and (3) between onehalf mile and three quariers of a mile. We alsoconsidered, as an alternative measure of prox-imity, the distance between the youth's schooland tbe dosest fast-food restaurant amongyouths wbose school was within 3 miles of a fast-food establishment (if pi-oximity to a fast-foodrestaurât affects wei^t status, we would expectlower BMI measures at scbools farther fhjm fast-food restaurants). Finally, we examined thenumber of fast-food restaurants witbin a halt-mile radius of the youth's school.
Our measures of food c:onsumption out-comes in the CHKS came from questions thatasked about tbe student's reported intake of 5food types: vegetables, fruit, juice, soda, midfiled potatoes. For each food type, we esti-mated logit models of tbe likelihood of con-suming that food type on the day before theinterview, and we also estimated the number ofservings of that food type on the day before tbeinterview by using a negative binomial model.For these food consumption outcomes, wecontrolled for al! of the student and schoolcharacteristics, as well as the school locationvariables and county indicators. For the logitmodels, we present the adjusted odds ratios forbeing near a fast-food restaurant, and for thenegative binomial models, we repori the mar-ginal effect estimated at the sample means.
As a set of additional tests for our mainweight status models, we examined the sensi-tivity of being near a fast-food restaurant toadditional controls for otber types of estab-lishments such as gas stations, motels, andgrocery stores. We identified the presence ofthese establisbments using Microsoft Streets
and Trips in the same way that we identifiedthe proximity of restaurants to schools. Weestimated models of weight status outcomesthat were identical to tiie baseline modelsexcept that being near a gas station, near amotel, and near a grocery store were added asadditional controls. Finally, we estimated sim-ilarly spedfied models for a placebo outcome—past-month tobacco consumption—that shouldnot be directly affected by proximity to a fast-food restaurant in the same way as weightstatus. Spedfically, we considered an indicatorof any past-montb dgarette smoking as theoutcome of interest (in a logit model in whichadjusted odds ratios are presented), and wecontrolled for the detailed student and schoolcharacteristics, as well as the variables for beingnear a gas station, near a motel, and near agrocery store.
RESULTS
Table 1 shows tbe basic CHKS descriptivestatistics, Tbe average BMI for students in thesample was 21.7 kg/m^, which the CDC con-siders a healthy weight for boys and girls agedat least 12,5 yeai-s." About 27.7"A¡ of our samplewas overwei^t and 12"/« was obese (obesechildren are also considered to be ovei-wei^t).Subtly over half of the students were girls, and theradal/ethnic composition was largely White andHispanic About 30%ofoursample was in middlescbool. Over one third (38"/ii) of the studentsattended schools in lai^e subuHsan areas. Overhalf of ell students (55"/(i) attended schools near(i.e., within one balf mile of) a fast-food restaurant.
Table 2 presents our main results, showingthat youths wbo attended schools located nearfast-food restaurants were heavier tlian wereother students with similar observable cbarac-teristics wbo attended scbools not locked neai"fast-food restaurants. Models predicting youtbs'overweigbt (model 1) and obesity (model 2)show that a youtb bad 1,06 times tbe odds ofbeing overweight (95"/(] confidence interval[CI] = 1.02,1.10) and 1.07 times the odds ofbeing obese (95% CI-1.02.1,12) if the youth'sschool was near a fast-food establishment; bothestimates were statistically significant. In model3. attending a school within one hall mile of afast-food establishment was assodated with aO.lO-unit increase Ü1 BMl (95%CI=0.03 kg/m^,0.16 kg/m' ) compared with youths whose
506 I Research and Practice I Peer Reviewed | Davis and Carpenter American Journal of Public Health | March 2009, No. 3
RESEARCH AND PRACTICE
TABLE 1-Descriptive SUtistics of Key
Variables: California Healthy Kids
Survey, 2002-2005
% Of Mean (SO)
Outcomes
BMI 21.66 (3.96)
Weight
Oveweight 28
Obesity 12
No. of servings in past
24 h
Vegetable 1.69 (1.52)
Fnitt 1.79 (1.60)
Juice 1.71 (1.64)
Soda 1.60 (1.62)
Fried potato 1.21 (1.39)
Any serving in past
24 hours
Vegetable 75
Fniit 74
Juice 70
Soda 68
fried potato 62
Primary predictors
% of establishments near
school
Fast-food restaurant 55
Other restaurant 25
Gas station 51
Motel 31
Grocery store 53
Indindual-level covariates
Gender
Boy 47
Girl 53
Grade
<7th 30
8th 2
9th 34
10th 2
nth 30
12th 1
Age, y
<12 21
13 11
14 24
15 12
16 23>17 9
Contitiued
TABLE l-Continued
Race/ethnicily
White
Asian
Black
Hawaiian
Hispanic
Ainehcan Indian
Multiple
Otiier
Physical activity, no. days
out of past 7
Exercise, no, days out of
past 7
31
10
4
2
31
1
14
7
3.21 (2.35)
4.03 (2.20)
School-level covariates
School type
High school
Middle xhool
Students eligible for
(ree/reduced price meals
School year
2002-2003
2003-2004
2004-2005
School enrollment
School location type
Large urban
Midsize urban
Small urban
Large suburban
Midsize suburban
Small suburban
Town
Rural
68
32
34
21
44
35
1864.13 (883.08)
15
10
11
38
7
4
8
9
Note. Oata are weighted to be representative at thedistrict level througi use of sample weights providedby the California Department of Education.'"
schools were not near a fast-food restaurantafter we controlled for detailed observablecharacteristics. Given a mean height and weightof 5 feet 3 inches and 110 pounds for youthsaged 14 yeare (the mean age of this population),a 0.10-unit Increase in BMI translates to 0.56 Ib.Models 1 through 3 also give estimates for beingnear other restaurants. As expected, we found asmaller relationship between this indicator anda youth's weight status, and the estimates werestatistically significant. Across ail outcomes, ourmodels can explain 5% to 10% ofthe variationin a youth's wei^t status.
To preserve space, we present results onlyfor the BMI outcome (the analyses for over-weight and obesity, which produced similarresults, are available upon request). Model 4 inTable 2 shows that for fast-food restaurantswithin one quarter mile (400 m) of a schooland between one quarter and one half mile of aschool, estimates are similar in magnitude tothe effects seen in model 3 and are statisticallysignificant The third measure, within onehalf mile to three quarters of a mile of a school,was not significant. Model 5 shows thatreplacing the indicator for being near a fast-food restaurant with the measure of distance tothe nearest fast-food establishment was con-sistent with the original result: there is a directrelationship between the proximity of fast-foodrestaurant to a school and a students' BMI.Finally, in mode! 6, there is no statisticallysignificant relationship between the number(4 vs 3) of fast-food restaurants within onehalf mile of a school and a students' BMI,suggesting that the deasity of fast-food restau-rants neai- schools may not be relevant toyouths' obesity.
We examined reported consumption ofvegetables, fruit juice, soda, and fried potatoes.Youths attending schools located near a fast-food restaurant had significantly lower odds ofreporting that they consumed vegetables orjuice on the day prior to the survey than didother youths (Table 3), and they also reportedconsuming significantly fewer servings ofvegetables, fruits, and juice than did studentsat schools that were not located near a fast-foodrestaurant after we controlled for detailedobservable characteristics. Table 3 showsresults for the arguably less-healthy foodtypes—soda and fried potatoes. We found thatattending a school near a fast-food restaurantwas associated with significantly higher oddsof reporting soda consumption on the daybefore the survey, after we controlled fordetailed observable characteristics. We did notfind differences in fried potato consumptionassodated with the proximity of a fast-foodoutlet although when we restricted our atten-tion to limited-service restaurants that Tech-nomic classified as "burger" establishments, wefound a significantly higher likelihood ofreporting fined potato consumption (odds ratio[OR]-1.02: 95«/oCI = l.OO, 1.04; data notshown; available upon request).'"*
March 2009. Vol 99, No. 3 | American Journal of Public Health Davis and Carpenter \ Peer Reviewed | Research and Practice I 507
RESEARCH AND PRACTICE
TABLE 2-Association Between a School's Proximity to a Fast-Food Restaurant and Overweight,Obesity, and Body Mass Index (BMi) Among Its Students {N = 529367}: CaliforniaHealthy Kids Survey, 2002-2005
Indicator
Fast-foot) restaurant within 0.5 miles of schooi
(among ttie top LSR estat)lishments)
Other restaurant within 0.5 miles of school
(not among the top LSR establishments)
fast-food restaurant 0-0.25 miles fmm school
Fast-food restaurant 0.25-0.5 miles from school
Fast-food restaurant 0.5-0.75 miles from school
Distance to nearest fast-fooú restaurant
No. of nearliy fast-toot) restaurants
ff
Model 1: Overweight,
AOfi (95% CI)
1.06**M1.02,1.10)
1.04'* (1.01, 1,08)
0.05
Model 2: Obese,
AOfi (95% CI)
1.07*** (1.02, 1.12)
1.04' (Î.0, 1.09)
0.06
Model 3: BMI.
b (95% CI)
0 . 1 0 " * (0-03, 0.16)
0.08*' (0.01. 0.14)
0.10
Mode! 4: BMI.
b (95% CI)
0.12*** (0.04, 0.20)
0 . 1 4 " * (0.06,0.23)
0,06 (-0.04, 0.16)
0.10
Motlel 5: BMI,
b (95% CI)
-0,03*** (-0.05,-0.01)
0.10
Model 6: BMI,
b (95% CI)
0.00 (O.OO, o,oo;
0.10
Me. Cl - confidence inlen/al; AOR=adjusted ot)t)s ratio; LSR - limited-sen/ice restaurants. We estimated logit motJels for overweight (model 1) and obese (model 2) youths, and for these models we
present AORs. In model 1, obese youths were also considered to be ovemeighL We used ordinary least squares for the BMI outcome in models 3 throu^ 6. CIs were adjusted for clustering at the
school level. In addition to the variables shown, 3ll models also included controls for the following student characteristics: a female indicator, grade indicator, age indicator, race/ethnicity
indicators, and physical exercise indicators. All models also included indicator variables for school location type, including large urban, midsize urhan, small urban, large suburt)an, midsize
suburban, smali suburtian, town, and rural, A full set of parameter estimates is available from the author upon request,
*P<,10; " P < . 0 5 ; * * *P<,01.
Table 4 shows the observed relationshipbetween proxiniity of fast-food restaurants andadolesœnt wei^t status alter control for nearbygas stations, motels, and groceiy stores. Wefound no relationship between the presence ofany of tbese types of businesses near the stu-dent's school and youths' BMI, likelihood ofbeing overwei^t or likelihocjd of being obese.Moreover, even with these establishments addedto the models, the relation.ship between prox-imity to fast-food restaurant end wei^t statusremained. The estimates indicate that attendinga sdiool located near a fast-food restaurant wasassodated with a statistically significant 0.13-unitincrease (95% CI=0.05, 0.20) in BMI after wecontrolled for the pr^ence of nearby gas sta-tions, motels, and grocery stores, in addition tothe standard control variables.
The results from our test of the placebo out-come of past-monlh cigarette smoking are shownin Table 4. There was a much smaller estimatedassociation between being near a fast-food res-taurant and dgarette smoking compared withthe associated relationships for the overweightand obese indicators, and the relevant estimatewas not statistically significant
We also investigated whether the estimatedrelationship between proximity of fast-food
restauraunts and wei^t status differed by students'or schools' observable demogr^hic characteris-tics. We found that among Black students (but noother radal/ethnic minorities), the associations
between being near a fast-food restaurant andBMI ib-0,20; 95% 0-0.04,0.36) were largerthan were baseline assodations representing allstudents. We also found that assodations
TABLE 3-Logit and Negative Binomiai Models of Association Between a School's Proximityto a Fast-Food Restaurant and Nutritional Intake Measures Among Its Students(N = 529367): California Healthy Kids Survey, 2002-2005 _
Nutritional Intake Measure
Any vegetables yesterday
No. of vegetable setvings yesterday
Any fr\ilt seriiings yesterday
No. of fruit servings yesterday
Any juice yesterday
No. of juice servings yesterday
Any soda yesterday
No. of soda servings yestertJay
Any fried potato servings yesterday
No. of fried potato servings yesterday
Negative Binomial Model, b (95% Cl)
-0.02*" 1-0.03, 0.00)
-0.02** (-0.04, 0.00)
-0.02*** (-0.03, 0.00)
0.02 (-0.01, 0.04)
0.00 (-0.02, 0.02)
Logit Model, AOR (95% Cl)
0.97* (0.93, 1.00)
0.97 (0.93, 1.02)
0.97* (0,94, 1.00)
1.05'* (1.00, l .U )
1.01 (0.98, 1.05)
«^
0.04
0.06
0.04
0.08
0.02
0.05
0.02
Ü.Ü6
0.02
Ü.04
Note. Cl=confidence interval; AOR - adjusted odds ratio. Because parameter estimates from negative binomial models are
not directly interpretable, we report the associateti marginal effects from being near a fast-food restaurant. CIs were adjusted
for clustering at the school level. In addition to the variables shown, all models also included controls for the following
student characteristics: a female indicator, grade indicators, age indicators, race/ethnicity indicators, antl physical exercise
indicators. All models also included indicator variables for school location type, inlcuding lar^ urban, midsize urban, smali
urban, iarge suburban, midsize suburtan, small suburban, town, and rural.
•P<.10. * 'P<.05. * ' * P < . 0 1 .
508 I Research and Practrce | Peer Reviewed | Davis and Carpenter American Joumal of Public Health | March 2009, Vol 99 , No. 3
RESEARCH AND PRACTICE
TABLE 4-Assoclation Between a School's Proximity to Other Types of Establishments and
Weight Status of Students, With Student Smoking Added as a Placebo: California Healthy
Kids Survey, 2002-2005
Indicator II. bOverweight,
AOR (95% CI)Obese,
AOR (95% CI)Smoker,
AOR (95% CI)
School near fast-food restaurant 0.13** ' (0 .05, 0.20) 1.08*'* (1-03,1.13) l . U * * * (1.04,1.18) 1.04(0.97,1.11)
School rear gas station
School near motel
School near grocery
-0.03 (-0.08, 0.03)
0.01 (-0.04, 0.06)
-0.04 (-0.09, 0.01)
0.10
0.99 (0.97, 1.02)
0.99 (0.97, 1,02)
0.98 (0.95, 1.01)
0.08
0.98 (0.94, 1.01)
0.99 (0.96, 1.03)
0.97 (0.94, 1.01)
0.08
0.99 (0.94, 1.04)
1.03 (0.97, 1.08)
1.00 (0.96, 1.05)
0.05
Me. AOR-adlusted odds ratio; Cl=confidence ¡iterval. We estimated models using ordinary least squares or logit: forthe logit models, we present the adjusted odds ratio. CIs were adjusted for clustering at the school level. In addition to thevariables shown, all models also included controls for the following student characteristics: a female indicator, gradeindicators, age indicators, race/ethnicity indicators, and ph^ical exercise indicators. All models also included indicatorvariables for school location type, inlcuding targe urban, midsize urtian, small urbar, large suburban, midsize suburban, smallsuburban, town, and rural.
between proximify of a fast-ibod restauraunt andwei^t status for students at urban schools (b=0.16;95%a-0.06,0.25) were larger than werebaseline assodations representing all students.
DISCUSSION
We found that students in California wereheavier and more likely to be overweight orobese if their school was located within one halfmile of a fast-food restaurant, after control forstudent demographic characteristics, schoolcharacteristics, and detailed controls for thetype of community in which the school waslocated. This main finding still held when otherdefinitions of nearby (i.e., other than within 0-5mile) were used and was not observed for othertypes of btisiness establishments also com-monly found near schools. Finally, we shov/that nearness to fast food was unrelated tosmoking. We also addressed the concern thatfast-food proximity may simply be a proxy forother unobserved characteristics about loca-tions that are independently correlated withweight status, such as the degree of economicdevelopment around a school. Overall, ourpatterns are consistent with the idea that fastfood near schools affects students' eatinghabits, overweight, and obesity.
Limitations
There were a few limitations to our study.BMI has been criticized as a measure of obesity
in part because the hei^t and wei^t used tocalculate it are self-reported. However, re-search with self-reported measures of BMfcombined with actual measures of BMi hasshown the 2 to be highly correlated.''*
The CHKS is compulsory for all Californiamiddle and high schools; however, as with anyschool-based survey, students were not in-duded if (1) the parents did not provide consentfor them to take the survey, (2) they wereabsent on the day the survey was administered,or (3) they had dropped out of school by theday of the survey. Although it is unlikely thatdata from missing respondents would afFect thefindings, concerns remain about generalizabil-ity and external validity. Student absence fromschool may be caused by illness, which couldbe more prevalent among overweight youths. Ifso, our results may understate the i-elationshipbetween a school's proximity to fast-food res-taurants and weight status. We also found thatIhe results for the yorangest students in thesurvey were consistent with the overall results,reducing concerns that student dropout mighthave biased our estimates.
One of our measures for unhealthful con-sumption was soda intake, which did not ac-count for whether the soda was sugar based ordiet soda Some may argue that soda con-sumption is an invalid measure of unhealthftildietaiy intake, because it may indude diet sodaconsumption; however, such potential mea-surement error would tend to mask (rather
than enhance) the association between fast-food consumption and childhood overwei^tA measure of solely sugar based soda con-sumption would only strengthen our results. Inaddition, recent research has shown that dietsoda consumption may be assodated witliobesity because the sweet taste encourages theconsumption of other high-calorie foods.'^
Other dimensions of the school environmentthat we did not observe could be important.For example, it would be useful to knowwhether students were allowed to leave schoolfor lunch, because our observed relationshipshould be sti-onger for those youths. In addi-tion, we controlled throughout for an extensiveset of variables related to sodoeconomic status,ethnidty, gender, and age; sodoeconomic sta-tus, however, was controlled at the school levelbut not at the individual level.
It is also undear how well our results gener-alize beyond California Children living in theSouth, for example, are more likely than arechildren in the West to be obese." Future workwith large samples of students in other stateswould be usefiil. Finally, we do not know thecausal directions of the associations betweenproximity to fast-food restaurants and overweightamong youths. If fast-food restaurants are suffi-dently savvy about locating near youths who willconsume their products, there might be a positiveassociation between proximity of fast-food res-taurants and adolescent overweight even if suchproximity did not directly cause the weight statusoutcome. We present only assodations betweennearby fast-lbod restaurants and adolescent obe-sily from data that are cross-secüonal.
Conci usions
Despite these limitations, our results can beused to inform current debates over schooleating polides. Our i-esults suggest that it mightbe useful to consider polides such as providingadolescents with alternatives to fast-food res-taurants. In addition, more research is neededon how public policy might target demograph-ically identifiable subgroups for interventionsrelated to proximity to fast-food restaurants.
A more drastic public policy measure wouldbe for local governments to restrict comnierdalpemiits for fast-food restaurants within walkingdistance of a school. ^ Policymakers could alsoconsider i estridions on the menus of restaurantsthat already exist within those zones, especially
March 2009. Vol 99, No. 3 [ American Journal of Public Health Davis and Carpenter \ Peer Reviewed | Research ano Practice I 509
RESEARCH AND PRACTICE
during lunch times and immediately before andafter school. Alternatively, offidals could con-sider ways to enœur^e vendors of hcalthiulfood to locate near schools.
Regaidless of v/hich option policymakerschoose, the need for intervention is dear. TheUnited States spends 12.7% of its gross do-mestic product on health care, and obesity isone ofthe most costly medical conditions.'^The sheer m^nitude of the problem of child-hood obesify demands attention, •
About the AuthorsAt Ihe lime of ihe siudy. Brennan Davis was with Ihe Schmlof Business and Management. Azusa Pacific Vniverfiity.Azusa. CA. Christopher Carpenter is with the Paul MerageSchool of Business, Universitif of California, ¡rvtne.
Requests ßir reprints should be sen! to Breniian Davis.PhlX Hankamer School of Business. Baylor Universitf).Orte Bear Place 98007. Waco. Tx 76798 (e-mail:markeîing_department@batflor.edu}.
This article was accepted fune 15. 2008.
ContributorsB. Davis originated the study and completed the analysis.C. Carpenter supervised the study. Both authors helpedto conceptualize the ideas, interpret findings, and writeand review drañs of the artide.
AcknowledgmentsWe are grateful to Ihe Paul Merage School of Business forgenerous financial support to purchase data.
We thank Mary Cilly for helpful comments, GregAustin for answering {}uesdons aboui the CHKS data, andTrade Etheredge for providing and answering questionsabout the Technomic data
Human Participant Protection.No protocol approval was needed for this study.
References1. McGinnis JM, Gootman JA, Ki'aak VI. Food Market-ing to Children and Youlh: Threat or Opportunity? Wash-ington DC: National Academies Press; 2006.
2. Cawiey J. Markets and childhood obesity policy.Future Child. 2006:16:69-88.
3. Wiecha JL, Peterson KE. Ludwig DS, Kim J, Sohol A,Gortmaker SL, When children eat what (hey watdi:impact of television viewing on dieiary intake in youth.Arch Pediatr Adolesc Med 2006; 160:436-442.
4. Bowman SA. Gortmaker SL, Ebbeling CA, PerdraMA, Ludwig DS. Effects of fast-food consumption onenergy intake and diet qualitj' among children in a na-donal household sludy. Pediatrics. 2004;l 13:112-118.
5. Duffey KJ, Goi-don-Larsen P, Jacobs DR. WilliamsOD, Popkin BM. Differentia] assodations of fast food andrestaurant food consumption with .1-y change in bodymass index: the Coronary Artery Risk Development in
r. 2007;85:201-208.
schools: a nove! application of spatial statistics to the.sEudy of food environments. .XmJ Fuhlic Health. 2005;95:1575-1581.
7. Zenk SN, Powell LM. United States secondaryschools and food ouüets. Health Place. 2008;14:336-346.
8. Powell LM, Auid C, Chaloupka FJ, O'MaUey P[,Johnston FD. Access to fast food and food prices: rela-tionship with fmit and vegetahle consumption and over-weight among adolescents. Adv Health Econ Health ServRes. 2006;17:23-48.
9. Sturm R, Datar .\. Body mass index in elementaryschool children, metropolitan area food prices and foodoutlet density. Public Health. 2OO5;l 19:1059-1068.
10. WestEd, Califomia Healthy Kids Survey, Availahleat: http://www.wested.org. Accessed February 22,2008.
11. Centers for Disease Control and Prevention. Growthcharts. Avail^le at; http://www.cdc.gov/growthcharts.Accessed Fehniaiy 22, 2008.
12. Califomia Dept of Education Web site. Available at:htlp://www.cde.cagov. Accessed February 22, 2008.
13. Technomic Web site. Availahle at; http://www.tedmomic.com. Accessed February 22, 2008.
14. Deitz WH, Robinson T. Use of body mass index(BMi) as a measure of overweight in children and ado-lescents, f Pediatr. 1998; 132:191 - 1 9 3 .
1Ö. Swithers Sfï. Davidson TL. A role for sweet taste:calorie predictive relations in energy regulation by rats,
i. 2008:122:1 f i l -173.
16. Mair JS, Pierce MW, Teret SP. The Use of Zoning toRestrict Fast Food Outlets: A Potential Strategy to CombatObesity. Baltimore, MD: Center for Law and the Public'sHealth, Johns Hopkins University; 2005.
17. Dellande S, Gilly MC, Graham JL. Gaining compli-ance and losing wei^t: the rote of the service provider inhealth care services,/ Ma;*. 2004;68;7ö-91,
CommunicatingPubUc HealthInformation EffectivelyA Guide for PractitionersEdited by David E. Nelson. MD. MPH: Ross CBrowmon. PhD: Patrick L Remington, MD, MPH;and Claudia Parvanta. PhD
As the first of its kind,this book provides ^a comprehensive \approach to help pub- ;lie health practitioners :in both the public andprivate sector to im-prove their ability tocommunicate withdifferent audiences.
Covering all the various modes of com-munication, each chapter provides practi-cal, real-world recommendations andexamples of how to communicate publichealth information to nonscientific audi-ences more effectively. The knowledge andskills gleaned from this book will assistwith planning and executing simple andcomplex communication activities com-monly done by public health practitioners.
ORDER TODAY!ISBN 0-87553027-3240 pages, softcovef, 2002$28.50 APHA Members (plus s&h)$40.00 Nonmembers (plus s&h)
American Public Health AssociationPUBLICATION SALESWEB: www.apha.orgE-MAIL: APHA@pbd.comTEL: 888-320-APHA FAX: 888-361-APHA
6. Austin SB, Meliy SJ, Sanchez BN, Patel A. Buka S,Gortmaker SL. Qustering of fast-food restaurants around
510 I Research and Practice I Peer Reviewed I Davis ano Carpenter American Joumal of Public Health | March 2009. Vol 99, No, 3
top related