Health & PlaceHow our built environments impact children’s health

Environmental health

Environmental health

Environmental health

43% decrease in Medicaid claims for acute asthma

equivalent to asking what the mode split would be ifthere had been no change in the spatial distribution ofschools and students across survey years. The differencebetween the distance-standardized and the age–race-standardized decline in ATS represents the effect ofchanging distance between home and school on youthtravel behavior.

Logit Model

To understand the relative influence of individual,household, and trip characteristics across the studyperiod, binary logit models predicting whether a childwalked or biked to school were constructed usingindividual trip records from the 1977 through 2001surveys. The Swait–Louviere test22–24 showed that theparameter vectors, !, were different across years("2!226, p"0.01) even allowing scale factors and thealternative-specific constant to differ by year. This sug-gests that modeling each survey year separately pro-vides the best fit to the data. Wald tests are used toanalyze whether the coefficients vary across the surveyyears to assess whether the relative influence of factorshas changed. Logit analyses were conducted in Stata 9.2(Stata Corp., College Station, TX) using the logitcommand with robust standard errors using appropri-ate weighting factors. Sample sizes are reduced becauseindividuals with missing data for household incomeand vehicle ownership have not been included in theanalysis.

Results

Analysis of the NPTS data shows that walking andbiking were the most common means of getting toschool in 1969, accounting for 40.7% (95% CI!37.9–43.5) of all trips (Figure 1). By 2001, active commuting

to school had declined by 27.8% to 12.9% (95%CI!11.8–13.9) of school trips. Nearly the entire de-cline in ATS occurred between 1969 and 1983 with thesharpest change between 1969 and 1977. The decreasein walking and biking is mirrored by a rise in driving toschool. For example, 55.0% (95% CI!53.6–56.5) ofstudents reached school by private vehicle in 2001compared with 17.1% (95% CI!14.9–19.3) in 1969.Use of school buses and public transit declined duringthe study period but not as sharply as active modes.

Variation in Active Transportation Rates

Elementary students, who had the highest rates of ATSin most years, experienced the steepest decrease,34.2%, in walking and biking between 1969 and 2001(Table 1). Approximately three quarters of the overalldecline among young students occurred between 1969and 1977. Although walk and bike rates have continuedto slip for elementary students since 1977, high schoolstudents experienced the largest decline (14.9%) be-tween 1977 and 2001 of any age group.

Boys have higher rates of ATS in each year, but thedecline in walking to school has affected both gen-ders equally (z!0.74, p!0.459). Minority studentsare twice as likely to walk to school as whites across allsurvey years, likely reflecting their lower level ofautomobile ownership across all survey years (datanot shown). Although the decline in walking between1977 and 2001 is higher for minority students, thereis no statistically significant difference in the declinebetween white and minority students (z!1.54,p!0.123).

Walking accounts for more than 90% of ATS.Although walk rates have fallen sharply at all agelevels, biking had a statistically significant declinebetween 1977 and 2001 only at the high school level.For those who walk, travel times have remainedrelatively constant during the study period rangingfrom a low of 10.0 minutes (95% CI!9.2–10.8) in1990 to a high of 12.7 minutes (95% CI!11.2–14.3)in 2001. Bike trip times range from a low of 8.6minutes (95% CI!6.8 –10.5) in 1990 to a high of 13.2(95% CI!10.8 –15.6) in 2001. Walk and bike traveltimes increase slightly with the age of the students,but the differences between grade levels are notstatistically significant.

Longitudinal Effects of Distance to School

Across all years, ATS varies sharply with distance toschool. In 1969, 85.9% (95% CI!82.3–89.4) of stu-dents living less than 1 mile from school walked orbiked compared with 1.6% (95% CI!0.4–2.8) forstudents living 3 or more miles from school (Figure 2).By 2001, the pattern remained the same. However, theproportion that chose to walk for trips of less than 1

Figure 1. Standardizeda mode shares for trips to school.aStandardized to 2001 age and race distribution. Error barsrepresent the 95% confidence interval.

June 2007 Am J Prev Med 2007;32(6) 511

McDonald Am J Prev Med 2007

Perc

ent (

%)

Year

Trends in U.S. Children’s Travel to School

equivalent to asking what the mode split would be ifthere had been no change in the spatial distribution ofschools and students across survey years. The differencebetween the distance-standardized and the age–race-standardized decline in ATS represents the effect ofchanging distance between home and school on youthtravel behavior.

Logit Model

To understand the relative influence of individual,household, and trip characteristics across the studyperiod, binary logit models predicting whether a childwalked or biked to school were constructed usingindividual trip records from the 1977 through 2001surveys. The Swait–Louviere test22–24 showed that theparameter vectors, !, were different across years("2!226, p"0.01) even allowing scale factors and thealternative-specific constant to differ by year. This sug-gests that modeling each survey year separately pro-vides the best fit to the data. Wald tests are used toanalyze whether the coefficients vary across the surveyyears to assess whether the relative influence of factorshas changed. Logit analyses were conducted in Stata 9.2(Stata Corp., College Station, TX) using the logitcommand with robust standard errors using appropri-ate weighting factors. Sample sizes are reduced becauseindividuals with missing data for household incomeand vehicle ownership have not been included in theanalysis.

Results

Analysis of the NPTS data shows that walking andbiking were the most common means of getting toschool in 1969, accounting for 40.7% (95% CI!37.9–43.5) of all trips (Figure 1). By 2001, active commuting

to school had declined by 27.8% to 12.9% (95%CI!11.8–13.9) of school trips. Nearly the entire de-cline in ATS occurred between 1969 and 1983 with thesharpest change between 1969 and 1977. The decreasein walking and biking is mirrored by a rise in driving toschool. For example, 55.0% (95% CI!53.6–56.5) ofstudents reached school by private vehicle in 2001compared with 17.1% (95% CI!14.9–19.3) in 1969.Use of school buses and public transit declined duringthe study period but not as sharply as active modes.

Variation in Active Transportation Rates

Elementary students, who had the highest rates of ATSin most years, experienced the steepest decrease,34.2%, in walking and biking between 1969 and 2001(Table 1). Approximately three quarters of the overalldecline among young students occurred between 1969and 1977. Although walk and bike rates have continuedto slip for elementary students since 1977, high schoolstudents experienced the largest decline (14.9%) be-tween 1977 and 2001 of any age group.

Boys have higher rates of ATS in each year, but thedecline in walking to school has affected both gen-ders equally (z!0.74, p!0.459). Minority studentsare twice as likely to walk to school as whites across allsurvey years, likely reflecting their lower level ofautomobile ownership across all survey years (datanot shown). Although the decline in walking between1977 and 2001 is higher for minority students, thereis no statistically significant difference in the declinebetween white and minority students (z!1.54,p!0.123).

Walking accounts for more than 90% of ATS.Although walk rates have fallen sharply at all agelevels, biking had a statistically significant declinebetween 1977 and 2001 only at the high school level.For those who walk, travel times have remainedrelatively constant during the study period rangingfrom a low of 10.0 minutes (95% CI!9.2–10.8) in1990 to a high of 12.7 minutes (95% CI!11.2–14.3)in 2001. Bike trip times range from a low of 8.6minutes (95% CI!6.8 –10.5) in 1990 to a high of 13.2(95% CI!10.8 –15.6) in 2001. Walk and bike traveltimes increase slightly with the age of the students,but the differences between grade levels are notstatistically significant.

Longitudinal Effects of Distance to School

Across all years, ATS varies sharply with distance toschool. In 1969, 85.9% (95% CI!82.3–89.4) of stu-dents living less than 1 mile from school walked orbiked compared with 1.6% (95% CI!0.4–2.8) forstudents living 3 or more miles from school (Figure 2).By 2001, the pattern remained the same. However, theproportion that chose to walk for trips of less than 1

Figure 1. Standardizeda mode shares for trips to school.aStandardized to 2001 age and race distribution. Error barsrepresent the 95% confidence interval.

June 2007 Am J Prev Med 2007;32(6) 511

McDonald Am J Prev Med 2007

Perc

ent (

%)

Year

Trends in U.S. Children’s Travel to School

equivalent to asking what the mode split would be ifthere had been no change in the spatial distribution ofschools and students across survey years. The differencebetween the distance-standardized and the age–race-standardized decline in ATS represents the effect ofchanging distance between home and school on youthtravel behavior.

Logit Model

To understand the relative influence of individual,household, and trip characteristics across the studyperiod, binary logit models predicting whether a childwalked or biked to school were constructed usingindividual trip records from the 1977 through 2001surveys. The Swait–Louviere test22–24 showed that theparameter vectors, !, were different across years("2!226, p"0.01) even allowing scale factors and thealternative-specific constant to differ by year. This sug-gests that modeling each survey year separately pro-vides the best fit to the data. Wald tests are used toanalyze whether the coefficients vary across the surveyyears to assess whether the relative influence of factorshas changed. Logit analyses were conducted in Stata 9.2(Stata Corp., College Station, TX) using the logitcommand with robust standard errors using appropri-ate weighting factors. Sample sizes are reduced becauseindividuals with missing data for household incomeand vehicle ownership have not been included in theanalysis.

Results

Analysis of the NPTS data shows that walking andbiking were the most common means of getting toschool in 1969, accounting for 40.7% (95% CI!37.9–43.5) of all trips (Figure 1). By 2001, active commuting

to school had declined by 27.8% to 12.9% (95%CI!11.8–13.9) of school trips. Nearly the entire de-cline in ATS occurred between 1969 and 1983 with thesharpest change between 1969 and 1977. The decreasein walking and biking is mirrored by a rise in driving toschool. For example, 55.0% (95% CI!53.6–56.5) ofstudents reached school by private vehicle in 2001compared with 17.1% (95% CI!14.9–19.3) in 1969.Use of school buses and public transit declined duringthe study period but not as sharply as active modes.

Variation in Active Transportation Rates

Elementary students, who had the highest rates of ATSin most years, experienced the steepest decrease,34.2%, in walking and biking between 1969 and 2001(Table 1). Approximately three quarters of the overalldecline among young students occurred between 1969and 1977. Although walk and bike rates have continuedto slip for elementary students since 1977, high schoolstudents experienced the largest decline (14.9%) be-tween 1977 and 2001 of any age group.

Boys have higher rates of ATS in each year, but thedecline in walking to school has affected both gen-ders equally (z!0.74, p!0.459). Minority studentsare twice as likely to walk to school as whites across allsurvey years, likely reflecting their lower level ofautomobile ownership across all survey years (datanot shown). Although the decline in walking between1977 and 2001 is higher for minority students, thereis no statistically significant difference in the declinebetween white and minority students (z!1.54,p!0.123).

Walking accounts for more than 90% of ATS.Although walk rates have fallen sharply at all agelevels, biking had a statistically significant declinebetween 1977 and 2001 only at the high school level.For those who walk, travel times have remainedrelatively constant during the study period rangingfrom a low of 10.0 minutes (95% CI!9.2–10.8) in1990 to a high of 12.7 minutes (95% CI!11.2–14.3)in 2001. Bike trip times range from a low of 8.6minutes (95% CI!6.8 –10.5) in 1990 to a high of 13.2(95% CI!10.8 –15.6) in 2001. Walk and bike traveltimes increase slightly with the age of the students,but the differences between grade levels are notstatistically significant.

Longitudinal Effects of Distance to School

Across all years, ATS varies sharply with distance toschool. In 1969, 85.9% (95% CI!82.3–89.4) of stu-dents living less than 1 mile from school walked orbiked compared with 1.6% (95% CI!0.4–2.8) forstudents living 3 or more miles from school (Figure 2).By 2001, the pattern remained the same. However, theproportion that chose to walk for trips of less than 1

Figure 1. Standardizeda mode shares for trips to school.aStandardized to 2001 age and race distribution. Error barsrepresent the 95% confidence interval.

June 2007 Am J Prev Med 2007;32(6) 511

McDonald Am J Prev Med 2007

Trends in U.S. Children’s Travel to SchoolPe

rcen

t (%

)

Year

?

Environmental health

?

Environmental DESIGN

POLICY STATEMENT

The Built Environment: DesigningCommunities to Promote PhysicalActivity in ChildrenCommittee on Environmental Health

ABSTRACTAn estimated 32% of American children are overweight, and physical inactivitycontributes to this high prevalence of overweight. This policy statement highlightshow the built environment of a community affects children’s opportunities forphysical activity. Neighborhoods and communities can provide opportunities forrecreational physical activity with parks and open spaces, and policies must sup-port this capacity. Children can engage in physical activity as a part of their dailylives, such as on their travel to school. Factors such as school location have playeda significant role in the decreased rates of walking to school, and changes in policymay help to increase the number of children who are able to walk to school.Environment modification that addresses risks associated with automobile traffic islikely to be conducive to more walking and biking among children. Actions thatreduce parental perception and fear of crime may promote outdoor physicalactivity. Policies that promote more active lifestyles among children and adoles-cents will enable them to achieve the recommended 60 minutes of daily physicalactivity. By working with community partners, pediatricians can participate inestablishing communities designed for activity and health. Pediatrics 2009;123:1591–1598

INTRODUCTIONA child’s life is affected by the environment in which he or she lives. Relationshipsbetween health and the quality of air, water, and food are well recognized.1–3 Thephysical environments of the home and school also influence health throughexposures to lead,4 mold,5 noise,6 or ambient light.7 In addition, the overallstructure of the physical environment of a child’s community (referred to as the“built environment”) can also affect health in diverse ways.

As cities have expanded into rural areas, large tracts of land have been frequently transformed into low-densitydevelopments in a “leapfrog” manner. The resultant urban sprawl can increase automobile travel, which increases airpollution8 as well as passenger and pedestrian traffic fatalities.9 Some urban areas may have few supermarkets,produce stands, or community gardens, thereby limiting access to fresh fruits and vegetables.10 The physicalenvironment of a community can support opportunities for play, an essential component of child development,11 andfor physical activity, a health behavior that not only reduces risk of excess weight gain12,13 but also has many otherbenefits for overall well-being.

Many factors influence a child’s level of physical activity, including individual-level psychosocial factors such asself-efficacy14,15; family factors such as parental support16; and larger-scale factors such as social norms.17 Althoughthese are all important contributors, this policy statement is limited to focusing on how the physical design of thecommunity affects children’s opportunities for physical activity. Opportunities for recreational physical activity arisewith parks and green spaces. “Utilitarian” physical activity, such as walking or bicycling to school and to otheractivities, is an equally important part of a child’s daily life. Environments that promote more active lifestyles amongchildren and adolescents will be important to enable them to achieve recommended levels of physical activity.

BACKGROUNDThe term “built environment” refers to spaces such as buildings and streets that are deliberately constructed aswell as outdoor spaces that are altered in some way by human activity. This term may be unfamiliar to mostclinicians, but with the high prevalence of childhood overweight and obesity,18 the subject is increasingly relevant.

www.pediatrics.org/cgi/doi/10.1542/peds.2009-0750

doi:10.1542/peds.2009-0750

All policy statements from the AmericanAcademy of Pediatrics automatically expire5 years after publication unless reaffirmed,revised, or retired at or before that time.

This document is copyrighted and isproperty of the American Academy ofPediatrics and its Board of Directors. Allauthors have filed conflict-of-intereststatements with the American Academy ofPediatrics. Any conflicts have beenresolved through a process approved bythe Board of Directors. The AmericanAcademy of Pediatrics has neither solicitednor accepted any commercial involvementin the development of the content of thispublication.

KeyWordsphysical activity, youth, neighborhood,active transport, walk to school, parks, builtenvironment, active living, urban design,pedestrian safety

PEDIATRICS (ISSN Numbers: Print, 0031-4005;Online, 1098-4275). Copyright © 2009 by theAmerican Academy of Pediatrics

PEDIATRICS Volume 123, Number 6, June 2009 1591

Organizational Principles to Guide andDefine the Child Health Care System and/orImprove the Health of All Children

at Claude Moore Health Sciences Library on June 15, 2009 www.pediatrics.orgDownloaded from

Pediatrics (June 2009)

POLICY STATEMENT

The Built Environment: DesigningCommunities to Promote PhysicalActivity in ChildrenCommittee on Environmental Health

Organizational Principles to Guide andDefine the Child Health Care System and/orImprove the Health of All Children

“...a pediatrician’s recommendation that a patient get regular physical activity loses its

salience if this patient’s everyday world lacks opportunities to walk, play, or run.”

Traffic Injury

Fatality Rate per 100 Million Miles Traveled

Driver Age (years)

Trowbridge Am J Prev Med (2008)

Urban Sprawl and Miles Driven Daily by Teenagersin the United StatesMatthew J. Trowbridge, MD, MPH, Noreen C. McDonald, PhD

Background: Urban sprawl’s association with increased automobile reliance and daily mileage is wellestablished among adults. However, sprawl’s specific impact on teen driving exposure isunknown. Teen driver fatality rates per mile driven are significantly higher than adults,making the identification of environmental influences on travel behavior particularlyimportant in this age group.

Methods: Driving and demographic data for 4528 teens (weighted!10.5 million) aged 16–19 yearswere obtained from the 2001 National Household Transportation Survey (NHTS).County-level sprawl was measured using an index developed by Ewing et al. The associationbetween daily miles driven by teens and sprawl, controlling for demographic characteris-tics, was modeled using ordinal logistic regression. The predicted probability of driving"20 miles in counties with varying degrees of sprawl also was calculated.

Results: Of the surveyed teens, 48% did not drive, 27% drove #20 miles/day, and 25% drove "20miles/day. Of the 52% of teens who reported driving, the average distance driven was 15.6miles/day. More-pronounced sprawl was associated with increased daily mileage(p#0.001). Overall, teens in sprawling counties were more than twice as likely to drive "20miles/day than teens in compact counties. This trend was most prominent among theyoungest drivers. For example, the predicted probability of boys aged 16–17 years driving"20 miles per day varied from 9% to 24% in compact versus sprawling counties.

Conclusions: Sprawl is associated with increased daily mileage by teen drivers. Given the starkrelationship between driving exposure and fatality risk among teens, increased efforts tounderstand and modify the effects of sprawl on adolescent driving behavior are necessary.(Am J Prev Med 2008;34(3):202–206) © 2008 American Journal of Preventive Medicine

Background

Despite dramatic improvements in automotivesafety engineering over the past few decades,motor vehicle crashes remain the most com-

mon cause of death among adolescents in the UnitedStates.1 More than 3500 drivers aged 15–20 years diedand more than 300,000 were injured in motor vehiclecrashes in 2004.2

Driving is a particularly dangerous activity for teens.Per mile driven, teens are involved in four to eighttimes the number of fatal crashes than mature drivers,3

due in large part to a confluence of developmentalfactors including normative risk taking and individualpersonality traits.4 As a result, dangerous driving behav-iors such as speeding, close following, or seat belt

non-use, which are prevalent among adolescents, arenot readily amenable to change.4

Given the recalcitrant nature of adolescent risk be-haviors, many teen driver safety interventions, such asgraduated drivers licensing,5 instead attempt to limitdriving exposure during this high-risk period of earlyskill building. Minimizing driving exposure among theyoungest and most novice teens (those aged 16–17years) appears to be particularly important given theirgreatly increased risk of crash involvement.6,7

In support of these efforts to reduce miles driven byteens, it is necessary to identify environmental factors,such as urban sprawl, that potentially influence adoles-cent travel behavior. Sprawl is a development patterntypified by low-density construction, poor street con-nectivity, and minimal land-use mix8 that has beenpreviously associated with increased automobile depen-dency and driving exposure among adult drivers.8,9

Sprawl’s relationship to driving exposure has not beenspecifically evaluated among adolescents. However,similar to adults, the total daily miles driven by teens arelikely increased in sprawling counties compared withteens living in more-compact counties. Confirmation ofthis hypothesis could have considerable policy and

From the Department of Emergency Medicine, School of Medicineand Center for Applied Biomechanics, School of Engineering, Uni-versity of Virginia (Trowbridge), Charlottesville, Virginia; Depart-ment of City and Regional Planning, University of North Carolina atChapel Hill (McDonald), Chapel Hill, North Carolina

Address correspondence and reprint requests to: Matthew J. Trow-bridge, MD, MPH, Department of Emergency Medicine, University ofVirginia School of Medicine, P.O. Box 800699, Charlottesville VA22908. E-mail: mtrowbridge@virginia.edu.

202 Am J Prev Med 2008;34(3) 0749-3797/08/$–see front matter© 2008 American Journal of Preventive Medicine • Published by Elsevier Inc. doi:10.1016/j.amepre.2007.11.013

per day varied from 7% in compact counties to 18% insprawling counties.

Discussion

The results of this study support the hypothesis thatsprawl is significantly associated with increased dailydriving mileage by teen drivers. Adolescents living incounties with sprawling development are more thantwice as likely to drive !20 miles each day than those inmore-compact counties. These findings are importantgiven the enormity of teen driver safety as a publichealth issue and the particularly stark relationshipbetween driving exposure and risk of severe injury ordeath among novice teen drivers.2,3 These results alsosuggest that the proliferation of sprawling developmentin the U.S. may undermine the shared goal of decreas-ing driving exposure among adolescents during theirhigh-risk period of early skill acquisition.

Why might this be the case? Studies of adults haveshown that people drive more in areas typified by low-density housing organized in “loops and lollipops” alongcentral feeder roads with poor street connectivity.8,11

Vehicle miles traveled are higher in areas with more-pronounced sprawl because trip distances are longerand alternative modes, such as walk, bike, or transit,may be impractical.15,16 These same forces likely causeteens in sprawling areas to substitute driving for walk-ing or taking public transit. In addition, living in areaswith good public transit service and proximate destina-tions actually may decrease demand for becominglicensed drivers. The relationship between teen licen-sure and the built environment remains unexaminedand is a promising area for further research.

Finally, the location of schools may affect how muchteens drive. Several authors have noted that the schoolsiting guidelines existing prior to 2004 encourageddistricts to build schools on large parcels.17–19 In manycommunities such parcels are available only on theedge of town, often removed from the residentialdevelopments they serve. This geographic arrangementincreases distance to school and it is well-establishedthat distance is the most critical factor in whether youth

walk to school.20 Recent efforts to encourage the coor-dination of school and land-use planning may lead tomore walkable schools. However, the size of highschools—often more then 1000 students—reduces op-portunities to locate the schools close to most students,except in very dense areas.

Limitations

There are several limitations to this study. The use ofcross-sectional data only allows for the determination ofsprawl’s association with increased daily driving expo-sure, rather than causation due to issues such aspotential self-selection bias. Alternatively, the use ofyouth-related data may minimize this issue given thefact that children do not make household locationdecisions.21 In addition, the county-level sprawl mea-sure used in this analysis is fairly coarse and thereforecannot distinguish the significant variation in develop-ment patterns that exist within most counties. Theeffect of using such an aggregate spatial scale will likelybe to bias the coefficients lower, that is, to underesti-mate the effect of sprawl on driving. Future analyseswould benefit from measures of sprawl at the neighbor-hood level.

Measuring distance alone also does not adequatelydescribe differences in driving contexts (e.g., trafficdensity, typical road designs, average speed) betweencompact and sprawling communities. The densest areasmay have highly congested roads that reduce averagevehicle speeds; residents of newer communities locatedat the urban edge often face less-congested conditionsand therefore drive at higher average speeds. Increasedspeed is associated with higher crash incidence andinjury or fatality risk.8

Finally, it was not possible to control for variation inlicensing restrictions that existed among states at thetime of the 2001 NHTS. Data specifying the implemen-tation date and enforcement guidelines of licensingrestrictions by state are not available. However, despitelack of adjustment for these regulatory factors, it isunlikely that the overall findings would be altered

Table 3. Predicted probability of adolescents driving !20 miles per day by county-level sprawl, age, and gendera

County-level sprawl (95% CI)

Age (years) Gender Compact (I!132)b Average (I!94)b Sprawling (I!56)b

All teens Both 21.7 (14.7–28.6) 33.0 (25.5–40.6) 46.8 (38.4–55.2)16–17 Male 9.0 (5.8–12.1) 14.9 (10.8–19.1) 23.8 (17.8–29.9)

Female 6.5 (4.2–8.9) 11.0 (7.7–14.4) 18.1 (12.7–23.6)18–19 Male 22.9 (12.3–33.4) 38.5 (24.4–52.6) 56.9 (42.3–71.6)

Female 16.4 (11.0–21.8) 25.9 (19.3–32.5) 38.4 (29.8–47.0)aAll probabilities calculated using reference household income level ($25,000–$50,000).bSprawl categories represent the mean index value (94.2) " two standard deviations (SD#19). Higher values of the county sprawl indexcorrespond to more-compact development, lower values to more sprawling development.I # county-level sprawl index value.

March 2008 Am J Prev Med 2008;34(3) 205

Trowbridge Am J Prev Med (2008)

Urban Sprawl and Miles Driven Daily by Teenagersin the United StatesMatthew J. Trowbridge, MD, MPH, Noreen C. McDonald, PhD

Background: Urban sprawl’s association with increased automobile reliance and daily mileage is wellestablished among adults. However, sprawl’s specific impact on teen driving exposure isunknown. Teen driver fatality rates per mile driven are significantly higher than adults,making the identification of environmental influences on travel behavior particularlyimportant in this age group.

Methods: Driving and demographic data for 4528 teens (weighted!10.5 million) aged 16–19 yearswere obtained from the 2001 National Household Transportation Survey (NHTS).County-level sprawl was measured using an index developed by Ewing et al. The associationbetween daily miles driven by teens and sprawl, controlling for demographic characteris-tics, was modeled using ordinal logistic regression. The predicted probability of driving"20 miles in counties with varying degrees of sprawl also was calculated.

Results: Of the surveyed teens, 48% did not drive, 27% drove #20 miles/day, and 25% drove "20miles/day. Of the 52% of teens who reported driving, the average distance driven was 15.6miles/day. More-pronounced sprawl was associated with increased daily mileage(p#0.001). Overall, teens in sprawling counties were more than twice as likely to drive "20miles/day than teens in compact counties. This trend was most prominent among theyoungest drivers. For example, the predicted probability of boys aged 16–17 years driving"20 miles per day varied from 9% to 24% in compact versus sprawling counties.

Conclusions: Sprawl is associated with increased daily mileage by teen drivers. Given the starkrelationship between driving exposure and fatality risk among teens, increased efforts tounderstand and modify the effects of sprawl on adolescent driving behavior are necessary.(Am J Prev Med 2008;34(3):202–206) © 2008 American Journal of Preventive Medicine

Background

Despite dramatic improvements in automotivesafety engineering over the past few decades,motor vehicle crashes remain the most com-

mon cause of death among adolescents in the UnitedStates.1 More than 3500 drivers aged 15–20 years diedand more than 300,000 were injured in motor vehiclecrashes in 2004.2

Driving is a particularly dangerous activity for teens.Per mile driven, teens are involved in four to eighttimes the number of fatal crashes than mature drivers,3

due in large part to a confluence of developmentalfactors including normative risk taking and individualpersonality traits.4 As a result, dangerous driving behav-iors such as speeding, close following, or seat belt

non-use, which are prevalent among adolescents, arenot readily amenable to change.4

Given the recalcitrant nature of adolescent risk be-haviors, many teen driver safety interventions, such asgraduated drivers licensing,5 instead attempt to limitdriving exposure during this high-risk period of earlyskill building. Minimizing driving exposure among theyoungest and most novice teens (those aged 16–17years) appears to be particularly important given theirgreatly increased risk of crash involvement.6,7

In support of these efforts to reduce miles driven byteens, it is necessary to identify environmental factors,such as urban sprawl, that potentially influence adoles-cent travel behavior. Sprawl is a development patterntypified by low-density construction, poor street con-nectivity, and minimal land-use mix8 that has beenpreviously associated with increased automobile depen-dency and driving exposure among adult drivers.8,9

Sprawl’s relationship to driving exposure has not beenspecifically evaluated among adolescents. However,similar to adults, the total daily miles driven by teens arelikely increased in sprawling counties compared withteens living in more-compact counties. Confirmation ofthis hypothesis could have considerable policy and

From the Department of Emergency Medicine, School of Medicineand Center for Applied Biomechanics, School of Engineering, Uni-versity of Virginia (Trowbridge), Charlottesville, Virginia; Depart-ment of City and Regional Planning, University of North Carolina atChapel Hill (McDonald), Chapel Hill, North Carolina

Address correspondence and reprint requests to: Matthew J. Trow-bridge, MD, MPH, Department of Emergency Medicine, University ofVirginia School of Medicine, P.O. Box 800699, Charlottesville VA22908. E-mail: mtrowbridge@virginia.edu.

202 Am J Prev Med 2008;34(3) 0749-3797/08/$–see front matter© 2008 American Journal of Preventive Medicine • Published by Elsevier Inc. doi:10.1016/j.amepre.2007.11.013

per day varied from 7% in compact counties to 18% insprawling counties.

Discussion

The results of this study support the hypothesis thatsprawl is significantly associated with increased dailydriving mileage by teen drivers. Adolescents living incounties with sprawling development are more thantwice as likely to drive !20 miles each day than those inmore-compact counties. These findings are importantgiven the enormity of teen driver safety as a publichealth issue and the particularly stark relationshipbetween driving exposure and risk of severe injury ordeath among novice teen drivers.2,3 These results alsosuggest that the proliferation of sprawling developmentin the U.S. may undermine the shared goal of decreas-ing driving exposure among adolescents during theirhigh-risk period of early skill acquisition.

Why might this be the case? Studies of adults haveshown that people drive more in areas typified by low-density housing organized in “loops and lollipops” alongcentral feeder roads with poor street connectivity.8,11

Vehicle miles traveled are higher in areas with more-pronounced sprawl because trip distances are longerand alternative modes, such as walk, bike, or transit,may be impractical.15,16 These same forces likely causeteens in sprawling areas to substitute driving for walk-ing or taking public transit. In addition, living in areaswith good public transit service and proximate destina-tions actually may decrease demand for becominglicensed drivers. The relationship between teen licen-sure and the built environment remains unexaminedand is a promising area for further research.

Finally, the location of schools may affect how muchteens drive. Several authors have noted that the schoolsiting guidelines existing prior to 2004 encourageddistricts to build schools on large parcels.17–19 In manycommunities such parcels are available only on theedge of town, often removed from the residentialdevelopments they serve. This geographic arrangementincreases distance to school and it is well-establishedthat distance is the most critical factor in whether youth

walk to school.20 Recent efforts to encourage the coor-dination of school and land-use planning may lead tomore walkable schools. However, the size of highschools—often more then 1000 students—reduces op-portunities to locate the schools close to most students,except in very dense areas.

Limitations

There are several limitations to this study. The use ofcross-sectional data only allows for the determination ofsprawl’s association with increased daily driving expo-sure, rather than causation due to issues such aspotential self-selection bias. Alternatively, the use ofyouth-related data may minimize this issue given thefact that children do not make household locationdecisions.21 In addition, the county-level sprawl mea-sure used in this analysis is fairly coarse and thereforecannot distinguish the significant variation in develop-ment patterns that exist within most counties. Theeffect of using such an aggregate spatial scale will likelybe to bias the coefficients lower, that is, to underesti-mate the effect of sprawl on driving. Future analyseswould benefit from measures of sprawl at the neighbor-hood level.

Measuring distance alone also does not adequatelydescribe differences in driving contexts (e.g., trafficdensity, typical road designs, average speed) betweencompact and sprawling communities. The densest areasmay have highly congested roads that reduce averagevehicle speeds; residents of newer communities locatedat the urban edge often face less-congested conditionsand therefore drive at higher average speeds. Increasedspeed is associated with higher crash incidence andinjury or fatality risk.8

Finally, it was not possible to control for variation inlicensing restrictions that existed among states at thetime of the 2001 NHTS. Data specifying the implemen-tation date and enforcement guidelines of licensingrestrictions by state are not available. However, despitelack of adjustment for these regulatory factors, it isunlikely that the overall findings would be altered

Table 3. Predicted probability of adolescents driving !20 miles per day by county-level sprawl, age, and gendera

County-level sprawl (95% CI)

Age (years) Gender Compact (I!132)b Average (I!94)b Sprawling (I!56)b

All teens Both 21.7 (14.7–28.6) 33.0 (25.5–40.6) 46.8 (38.4–55.2)16–17 Male 9.0 (5.8–12.1) 14.9 (10.8–19.1) 23.8 (17.8–29.9)

Female 6.5 (4.2–8.9) 11.0 (7.7–14.4) 18.1 (12.7–23.6)18–19 Male 22.9 (12.3–33.4) 38.5 (24.4–52.6) 56.9 (42.3–71.6)

Female 16.4 (11.0–21.8) 25.9 (19.3–32.5) 38.4 (29.8–47.0)aAll probabilities calculated using reference household income level ($25,000–$50,000).bSprawl categories represent the mean index value (94.2) " two standard deviations (SD#19). Higher values of the county sprawl indexcorrespond to more-compact development, lower values to more sprawling development.I # county-level sprawl index value.

March 2008 Am J Prev Med 2008;34(3) 205

?

Sources of Inspiration

Times Square, NYC

High Line Park, NYC

High Line Park, NYC

NYC Citi Bikeshare

Citi Bikeshare

New Tools

ACTIVE DESIGN GUIDELINES FINAL DRAFT OCTOBER 9, 2009 CHAPTER 3 BUILDING DESIGN

63

3.2 Stair location and visibility Objective Increase stair use by locating a highly visible and appealing stair within the building’s orientation areas and points of decision. Strategies

Locate stairs near the building’s entrance. Research indicates that stairs located within 25 feet of an entrance and encountered prior to the elevator are more likely to be used for everyday travel.3

Locate a stair targeted for everyday use near the elevator.

Users may consider taking a stair located near to and visible from the elevator lobby a more expedient option than waiting for the elevator.3,5 Visual and physical proximity of a stair to the elevator can be supplemented by point-of-decision signage encouraging people waiting for elevators to take the stairs.

Locate an appealing, visible stair directly on the building’s principal entrance and paths of

travel. Research indicates that stairs directly accessible and visible from a building’s elevator waiting areas, atrium, entry vestibules, and most-used public corridors are more likely to be utilized for everyday travel.3 One study found that stair use decreased as the number of turns required to access the stairs from either the building’s entrance or principal path of travel increased. Highly visible grand or ornamental stairs provide a clear indication that they are provided for use.

Design stairs to be more visible, in order to encourage their everyday use. Stairs, particularly those designated for fire egress, often are not visible from a building’s main spaces. Code-mandated fire separations are traditionally achieved by encasing stairs in opaque masonry or gypsum board assemblies with solid metal doors. However, alternative assemblies, materials, and systems are available that allow egress stairs to be more visible while meeting code requirements for fire resistance ratings. Egress stairs can be made more visible by incorporating: • Fire-rated glass enclosures instead of traditional opaque enclosures. • Open stairs between two or more floors with either the same or associated tenancies. Where

open stairs connect more than two floors, spaces, additional sprinkler systems, smoke control system, and building code variance may be required, as per NFPA 13 and the New York City Building Code.4,6

DRAFT - DO NOT DISTRIBUTE OR CIRCULATE

ACTIVE DESIGN GUIDELINES FINAL DRAFT OCTOBER 9, 2009 CHAPTER 3 BUILDING DESIGN

64

Locate stairs near the building’s entrance to encourage everyday use. Apple store, Manhattan, Bohlin Cywinski Jackson with Ronette Riley Architect

Enclosing a stair with glass increases its visibility and encourages everyday use. The glass pictured here is fire protected through the use of a sprinkler water curtain. Memorial Sloan-Kettering Cancer Center interaction staircase, Manhattan, SOM in collaboration with ZGF Architects

DRAFT - DO NOT DISTRIBUTE OR CIRCULATE

ACTIVE DESIGN GUIDELINES FINAL DRAFT OCTOBER 9, 2009 CHAPTER 3 BUILDING DESIGN

64

Locate stairs near the building’s entrance to encourage everyday use. Apple store, Manhattan, Bohlin Cywinski Jackson with Ronette Riley Architect

Enclosing a stair with glass increases its visibility and encourages everyday use. The glass pictured here is fire protected through the use of a sprinkler water curtain. Memorial Sloan-Kettering Cancer Center interaction staircase, Manhattan, SOM in collaboration with ZGF Architects

DRAFT - DO NOT DISTRIBUTE OR CIRCULATE

Manassas Park ElementaryVMDO Architects

Poquoson Elementary School

BUILDING SUSTAINABLE SCHOOLS FOR HEALTHY KIDS

Green Health

A Workshop Co-sponsored by the National Collaborative for Childhood Obesity Research and the National Academy of Environmental Design

In partnership with the U.S. Green Building Council Center for Green Schools

nccor.org/projects/greenhealth

Volume 10 — February 28, 2013TOOLS AND TECHNIQUES

Healthy Eating Design Guidelines for School Architecture

Terry T-K Huang, PhD, MPH, CPH; Dina Sorensen, MArch; Steven Davis, AIA; Leah Frerichs, MS; Jeri Brittin, MM; Joseph Celentano, AIA; Kelly Callahan, AIA; Matthew J. Trowbridge, MD, MPHSuggested citation for this article: Huang TT, Sorensen D, Davis S, Frerichs L, Brittin J, Celentano J, et al. Healthy Eating Design Guidelines for School Architecture. Prev Chronic Dis 2013;10:120084. DOI: http://dx.doi.org/10.5888/pcd10.120084 .

PEER REVIEWED

AbstractWe developed a new tool, Healthy Eating Design Guidelines for School Architecture, to provide practitioners in architecture and public health with a practical set of spatially organized and theory-based strategies for making school environments more conducive to learning about and practicing healthy eating by optimizing physical resources and learning spaces. The design guidelines, developed through multidisciplinary collaboration, cover 10 domains of the school food environment (eg, cafeteria, kitchen, garden) and 5 core healthy eating design principles. A school redesign project in Dillwyn, Virginia, used the tool to improve the schools’ ability to adopt a healthy nutrition curriculum and promote healthy eating. The new tool, now in a pilot version, is expected to evolve as its components are tested and evaluated through public health and design research.

IntroductionCreating school food environments that facilitate healthy eating among children is a recommended national strategy to prevent and reduce childhood obesity (1). According to socioecological models, the macroenvironment — the density of fast food outlets around schools (2,3), for example — affects eating behavior. Limited research has focused on the microenvironment, such as building design. School design can affect student behavior, development, and academic performance (4–6). Food displays and time allotment for school meals can also affect children’s eating behavior (7–9). A recent evaluation of a system-level healthy eating initiative in 4 California schools showed that changes in dining room design and features may have contributed to positive outcomes such as increased nutrition and knowledge of the food environment, preference for fruits and vegetables, and higher in-school and out-of-school fruit and vegetable consumption (10).

Interest is growing in how the physical design of school buildings (ie, architecture, interior design, and landscaping) affects school policies and practices and the subsequent eating behaviors and norms among children. Systematic theory- and evidence-based design strategies drawn from both public health and architecture are needed to define, test, and further develop best practices. We developed a pilot version of a new tool, Healthy Eating Design Guidelines for School Architecture. These guidelines, organized for practical use by architects, schools, and health researchers, present a set of design strategies to help promote healthy eating behaviors in schools.

Development of the Healthy Eating Design Guidelines for School ArchitectureTheoretical frameworksThe Healthy Eating Design Guidelines (Table) represent a new application of existing theoretical frameworks on the role of school building design in child development and health promotion (11). The guidelines draw on research in environmental health, environmental psychology, behavioral economics, and socioecological models.

Page 1 of 12Preventing Chronic Disease | Healthy Eating Design Guidelines for School Architecture - ...

Volume 10 - February 28, 2013[Open Access Journal]

Buckingham Elementary (VMDO Architects)Dillwyn, VA

Buckingham Elementary (VMDO Architects)

Educational Signage

Opening Day

KINDERGARTENCHILDREN_WATERING ACTIVITY

Schoolyard Garden

Matthew Trowbridge MD MPHmtrowbridge@virginia.edu