OBJECTIVES: Initiatives to increase active school transportation are popular. However, increased walking to school could increase collision risk. The built environment is related to both pedestrian collision risk and walking to school. We examined the influence of the built environment on walking to school and child pedestrian collisions in Toronto, Canada.
METHODS: Police-reported pedestrian collision data from 2002 to 2011 for children ages 4 to 12, proportion of children walking to school, and built environment data were mapped onto school attendance boundaries. Collision rates were calculated by using 2006 census populations and modeled by using negative binomial regression.
RESULTS: There were 481 collisions with a mean collision rate of 7.4/10 000 children per year. The relationship between walking proportion and collision rate was not statistically significant after adjusting for population density and roadway design variables including multifamily dwelling density, traffic light, traffic calming and 1-way street density, school crossing guard presence, and school socioeconomic status.
CONCLUSIONS: Pedestrian collisions are more strongly associated with built environment features than with proportions walking. Road design features were related to higher collision rates and warrant further examination for their safety effects for children. Future policy designed to increase children’s active transportation should be developed from evidence that more clearly addresses child pedestrian safety.
- AST —
- active school transportation
- BE —
- built environment
- CI —
- confidence interval
- DA —
- Dissemination Area
- IRR —
- incident rate ratio
- LOI —
- Learning Opportunities Index
- MPAC —
- Municipal Property Assessment Corporation
- SES —
- socioeconomic status
- SRTS —
- Safe Routes to School
- TDSB —
- Toronto District School Board
What’s Known on This Subject:
Many studies have demonstrated that the built environment is related to both collision risk and walking to school. However, little research examines the influence of the built environment on the relationship between walking to school and pedestrian collision risk.
What This Study Adds:
Increased walking was not associated with increased pedestrian collision once the effects of the built environment and socioeconomic status were modeled. Safety was related primarily to the built environment and specifically features related to road crossing.
Road traffic injuries are the leading cause of child death in most developed countries.1–3 In 2010, 61 children died and over 9000 were injured on Canada’s roads.4 Pedestrian collisions account for ∼25% of children’s road traffic fatalities.5 Much of children’s exposure to traffic as pedestrians is during school travel.6–8 Although walking to school rates have declined in Canada, ∼50% of Ontario children walk to school.9–11 In Toronto, Canada, ∼50% of child pedestrian collisions occurred during school travel times with more than one-third occurring within 300 m of a school.12
Initiatives to increase active school transportation (AST) are popular. In 2005, the US Safe Routes to School (SRTS) program received over $1 billion of federal funding.13,14 In Canada, SRTS programs have developed at a grassroots level, with limited government funding.14 Increased walking to school might benefit overall health, but also increase collision risk. A recent Canadian study revealed a possible increase in walking/cycling injuries among youth who regularly walked or cycled longer distances to school (>15 minutes), compared with those who did not walk/cycle to school.15
Many conceptual frameworks describing correlates of walking to school focus on the built environment (BE). The BE is defined as “the human-made space in which people live, work, and recreate on a day-to-day basis.”16 No framework considers child pedestrian injury, and there has been little research examining pedestrian collision as an AST outcome. The purpose of this study was to examine the effect of the BE on the relationship between observed walking to school and child pedestrian collisions.
Study Design, Setting, and Population
A cross-sectional study examined child pedestrian collisions from 2002 to 2011. Walking exposure was measured in an observational study in the spring of 2011 in junior kindergarten to grade 6 schools in Toronto. Schools were excluded if they were participating in another active transportation study or had special programs (eg, French immersion), which serve large areas. Further methodological details were reported previously.17 Ethics approval was obtained from the Toronto District School Board (TDSB) and the Hospital for Sick Children Research Ethics Boards.
Collision data were extracted from Toronto Police Service motor vehicle collision reports from 2002 to 2011 for children ages 4 to 12 years. Police-reported collisions include those resulting in no injury, minimal, minor (seen in the emergency department), major (admitted to hospital), and fatal injuries. Collision rates by study school boundary were calculated over a 10-year period and reported as an annualized rate per 10 000 children. Child population numbers were obtained for Dissemination Areas (DAs, 400–700 residents) by using the 2006 Canadian Census. DAs were mapped onto school boundaries, and population estimates were calculated by using area-weighted proportionate estimation.18,19
Two trained observers measured walking exposure by counting children either arriving to school walking, by car, or other active means (bicycle or scooter) on a single day. Observations were repeated in 22 schools (19%) where at least 1 observer rated <90% confidence in the accuracy of their counts and were repeated 1 week later in another 12 schools (10%) to examine test-retest reliability. Walking proportion was calculated from the total number observed. The majority of children counted lived within walking distance of the school (ie, ≤1.6 km), as defined by TDSB transportation policy.
BE variables were identified from a literature review and conceptually organized by using Cervero and Kockelman’s 3Ds: density of population, diversity of land use, and design of the roadway environment.20 Socioeconomic status (SES) variables were included because of previously reported correlations with child pedestrian injury.21,22 Table 1 presents each variable according to its conceptual category, level of measurement, and data source. Variables were measured at the school and school attendance boundary level.
DA data were obtained from the 2006 Canadian census. Social environment variables included the proportion of households falling below the after-tax, low income cut-offs in each school’s DA, as a proxy measure for the school neighborhood SES.
Municipal Property Assessment Corporation
Land use diversity variables were derived by using 2011 parcel level data from the Municipal Property Assessment Corporation (MPAC), which classifies and assesses properties in Ontario. Mixed land use was calculated by using an entropy index, which ranges from 0 (single land use) to 1 (equal distribution of residential, commercial, industrial, institutional, and vacant land classifications).23,24
School level design variables were obtained from site audits conducted by the observers during school drop-off time. Only adults employed by Toronto Police Services surrounding the school were identified as school crossing guards. Vehicle speed and volume were measured along a road within 150 m of the school by using manual short-based methods.25,26
City of Toronto
Collision outcome data, school boundary level design variables, and recreational facility land use were obtained from The City of Toronto (Transportation Services and the Open Data Web site).27 Densities were calculated either per school boundary area or per kilometer of roadway. Toronto consists of an older central city with many pre-World War II traditional neighborhoods and an inner ring representing newer, car-oriented post-World War II neighborhoods.28 Central city status was assigned if >50% of the school boundary overlapped with the central city area.
Toronto District School Board
The TDSB provided social environment variables and school attendance boundary data. The 2011 Learning Opportunities Index (LOI) is a composite measure of children’s SES attending a school reflecting parental education, income, housing and immigration, and ranges from 0 (high) to 1 (low SES).
School attendance boundaries were the unit of analysis. All features were mapped onto boundaries by using ArcMap version 10.29 Statistical analysis was conducted by using SAS, version 9.3.30 Mutlicollinearity was assessed by using variance inflation factors. If 2 correlated variables had variance inflation factors >10, the variable with the higher standardized unadjusted β was retained.31
Child pedestrian collision rates were modeled by using negative binomial regression. Variables with P ≤ .2 in the unadjusted analysis were entered by using forward manual stepwise regression, according to the magnitude of standardized βs.32 Variables were retained if significantly associated with the outcome at P ≤ .05, with confounding identified if the variable also changed walking estimates by ≥10%.32,33 Stratified analyses were conducted by tertiles for design features to assess for differential impact on outcome.
Model fit was assessed by using the Akaike Information Criteria.34,35 Incident rate ratios (IRRs) were calculated through exponentiation of βs from the regression models and presented with 95% confidence intervals (CIs). Sensitivity analysis was conducted by using 10, 7, and 5 years of collision data. A subanalysis was done of collisions occurring only during school travel times (7:30–9:00 am, 11:30 am–1:00 pm, 3:00–5:00 pm, weekdays, September through June). Both before and after school collisions were included, as numbers of elementary school children walking to and from school are comparable.36,37
Among 245 junior kindergarten to grade 6 schools, 126 met inclusion criteria, 8 refused, and 118 schools participated in the study. A total of 481 collisions occurred within 105 school boundaries. There were 24 collisions resulting in no injury, 191 minimal, 236 minor, 30 major injuries, and 1 fatality. The average collision rate was 7.4/10 000 per year (range, 0–27/10 000, SD = 6.7). Two outlier schools with extreme collision rates (>3 interquartile ranges above the 75th percentile) were excluded. These were inner city schools with very small attendance boundaries and few resident children contributing to low rate denominators. The mean proportion of children walking to school was 67.0% (range, 27.9%–98.2%, SD = 14.4%) with high test-retest reliability of walking counts (Pearson’s r = 0.96).
In the unadjusted analysis, increased walking proportions were associated with higher pedestrian collision rates (IRR = 3.47; 95% CI = 1.15–10.47; Table 2). This was equivalent to a 13% increase in collision rate with every 10% increase in walking.
Older housing, residential, road, and local road density were dropped from further analyses due to multicollinearity.
In the adjusted collision model, walking proportions were no longer associated with collision rates (Table 3). Collisions were less frequent in areas with higher multifamily dwelling density (IRR = 0.84; 95% CI = 0.73–0.96). Design variables, including higher densities of traffic lights (IRR = 3.20; 95% CI = 1.89–5.41), traffic calming (IRR = 1.31; 95% CI = 1.06–1.63), 1-way streets (IRR = 1.19; 95% CI = 1.03–1.36), school crossing guards (IRR = 1.45; 95% CI = 1.09–1.91), and low school SES (IRR = 2.36; 95% CI = 1.39–3.99) were positively associated with collisions and changed the walking exposure by ≥10%. Significant design variables were generally related to road crossing. Traffic light density and SES exhibited the strongest associations.
The association between walking and collisions differed by levels of traffic light density; walking was positively associated with collisions in low traffic light density areas, with no association in medium/high density areas (Table 4). Tertiles were dichotomized into low density and medium/high density areas for ease of interpretation.
Analysis using 5- and 7-year collision data revealed very similar models. School travel time collisions represented 44% (n = 214) of total collisions within 83 school boundary areas. Although there were reductions in the magnitude of effects, effect direction was similar to the full model with less precise estimates because of the smaller sample size.
A significant positive univariate association between walking and pedestrian collisions was no longer observed in adjusted models that controlled for population density, design features (primarily related to road crossings), and SES. Land use diversity was unrelated to collisions. These findings are encouraging in that it suggests that modification of the BE may both promote walking and make it safer. Although causality could not be definitively determined because of the cross-sectional and ecological study design, the results suggest a potentially important influence of the BE on walking and collisions.
Comparisons of Findings to Previous Studies
Previous studies of the association between pedestrian collisions and walking have revealed inconsistent results. In population studies, increased pedestrian volume was associated with decreased collisions leading to a “safety in numbers” effect.38,39 Other studies revealed positive relationships between walking exposure and child pedestrian collisions.15,40,41 Different environmental conditions and a lack of adequate control for confounders may contribute to these contradictory findings.
BE features have been previously correlated with both walking to school and pedestrian collisions.42–47 Major roads, urban location, street/intersection density, sidewalks, school crossing guards, population density, distance, and traffic controls have been associated with walking to school.45,47 A reduction in pedestrian injures has been associated with BE road design interventions by 50% to 75% in specific locations.43 Traffic calming has been associated with a 37% reduction in fatal pedestrian outcomes48 and a 53% reduction in child pedestrian collision risk.49 The BE influences both the walking exposure and risk, so it must be considered as a confounder in child pedestrian safety studies.
Multifamily dwelling density, SES, 1-way streets, traffic calming, and school crossing guards were confounders of the relationship between walking to school and child pedestrian collisions as they changed the walking exposure estimates by ≥10%. Higher multifamily dwelling density was associated with lower collision rates, which may indicate a safer walking environment and supports the safety in numbers concept (Fig 1). Multifamily dwelling has been associated with higher pedestrian collision risk at the individual level, unadjusted for walking exposure.50,51 At the population level, higher population density may reflect shorter distances to school with fewer road crossings, resulting in less traffic exposure and fewer collisions. Lower SES has been consistently associated with higher child pedestrian collisions,52,53 with higher rates attributed to lack of safe play areas and higher walking exposures because of lower car ownership.41,54–57 More collisions on 1-way streets have also been found in Hamilton Ontario, with a 2.5 times higher injury rate on 1-way compared with 2-way streets.58 Traffic calming devices were positively associated with collisions, whereas other studies revealed a negative association; however, no adjustment was made previously for traffic exposure.49,59,60 Finally, the presence of school crossing guards was positively associated with collisions, which has similarly been found in Montreal, Quebec.61
Features considered to be confounders in this analysis may fall on the casual pathway (particularly school crossing guards) between walking to school and pedestrian collisions. However, the relationships are complicated in that they may be bidirectional with the possibility of reverse causality. The cross-sectional nature of the data restricted the ability to determine temporality and directionality of the relationships. Casual pathways need to be established to determine if any of these features act as mediators between walking to school and child pedestrian collisions.
Overall, traffic light density was positively related to both walking and collisions. Stratified analysis revealed that traffic light density was an effect modifier, as walking was only positively associated with collision rates where traffic light densities were low. Traffic light density has been previously associated with less pedestrian collisions.62,63
There were several unexpected results of these analyses. Traffic speed and volume were not associated with collision rates, which have been previously identified as important confounders for pedestrian-related collisions.64 These null findings are likely a result of poor sensitivity of the measurements taken only on 1 busy roadway near the schools. Accurate traffic data were not available close to schools from other sources. Volume and speed were, however, likely measured by proxy through other variables, such as traffic calming and road type.
The positive associations between collision rates and both school crossing guards and traffic calming were unexpected, as these features are intended to be protective. These features may be indicators for more hazardous locations with the feature’s safety effect insufficient to reduce injury risk.50,61 These features could also potentially create hazardous traffic situations for children. Although evidence exists of reduced effectiveness of specific road design features for older adults, this has not been well studied in children.65 More rigorous study of these features with further spatial analysis is required to better ascertain the safety effects for children.
The ecological nature of the data may also have contributed to these unexpected findings. Collisions occurred at a specific location, whereas the built and social environment data were measured area-wide and may not represent the environment at the collision location. Area-level data would also not identify if a BE feature either displaced collisions or was a marker for a more dangerous traffic environment elsewhere within the school boundary. Ecological analysis was, however, felt to be most appropriate, as an understanding of both the physical and social environment measured on a geographic level is essential when examining pedestrian injury.
The possibility of reverse causality existed, with collisions occurring before the installation of these features, because of the cross-sectional design of the study. Several assumptions were made regarding the data. Walking exposure measured in 2011 was assumed to be representative of exposure throughout the 10-year collision data period (2002–2011). This was reasonable because there was evidence of stabilization of active school travel prevalence from 2001 onwards, after an earlier period of sharp decline.10,66,67 The BE was also assumed to remain unchanged over the collision data period. Substantial changes would not be expected in the BE generally because the study neighborhoods were well-established. However, this may not have been the case for features more easily implemented, such as the installation of traffic calming or school crossing guards.
Results may also have been affected by the walking exposure, which was measured only en route to school and did not include nonschool travel time collisions. The study’s intention was to examine factors related to all child pedestrian collisions in Toronto. Because school is children’s most common walking destination, walking to school is considered the best proxy for their general walking activity.6–8 Consistent evidence also exists that children who actively commute to school walk more in general.68 The inclusion of nonschool travel time collisions when school crossing guards were off-duty could potentially have produced an artificial positive association between guards and collisions. However, when restricting the data to school times, the directions of effects were maintained. The use of walking to school as a proxy for children’s general walking is currently the most feasible measurement of children’s walking exposure. Walking exposure has been poorly dealt with in the past, and creative methods of measurement are needed to most accurately evaluate pedestrian risk.
Study strengths included that it was large and population-based and directly measured walking to school as a proxy for general walking exposure. The only previous study that used direct observation of walking to school to investigate associations with BE was limited by a small sample and little geographic diversity.36 Multivariable modeling was used to test the association between walking and pedestrian collisions while controlling for the BE. The model identified BE features acting as confounders and effect modifiers of the relationship between walking exposure and collision outcome. One other recent study that modeled walking school trips and active transportation injuries examined limited numbers of potential BE confounders.15 However, none of the roadway design features significant in our study were included.
The study findings are relevant for parents and pediatricians and on the policy and public health levels. Walking to school over a short distance was found to be relatively safe in Toronto. Pediatricians can counsel parents to encourage children to walk to school as a healthy lifestyle choice. Parents can be informed that the number and type of road crossings are the major determinant of injury risk. Parents, pediatricians, and school principals can advocate for walkable environments, by approaching school boards, the police, and city officials to ensure the safety of children walking to school.
Active school travel policy is currently undergoing changes in Canada and the United States. In Ontario, Canada, the provincial government recently initiated the “Stepping It Up” school travel planning program under its regional transportation plan, intending to spend $200 million on active transportation infrastructure and research to improve safety and achieve AST >60% for all schools.9,69 In the United States, recent changes to the federal transportation bill have eliminated SRTS as a separate funding program. Alternative funding through the federal Highway Safety and Infrastructure Program now requires SRTS projects to show both evidence of increasing active transportation and a reduction in collisions.70 These changes provide an opportunity to incorporate safety evaluation into new policies.
Several important conclusions and implications have emerged from the study. Firstly, the positive relationship between walking and collision rates was no longer significant after controlling for the BE. These results are encouraging for walking promotion programs, suggesting that safety issues are concerned primarily with the BE and not the numbers walking. Secondly, design features related to road crossing exhibited the most influential effects. To increase walking safety in children, focus should be on minimizing or mitigating road crossings, as opposed to changing other factors such as land use, which may be more applicable to adults. Finally, the mechanisms of how to mitigate road crossings for children are not well understand, and well-controlled research designs must be integrated into SRTS program evaluation. Future policy designed to increase children’s active transportation should be developed from strong evidence that addresses child pedestrian safety.
The authors thank the TDSB for their participation in this project and various departments at the City of Toronto for providing data.
- Accepted February 7, 2014.
- Address correspondence to Linda Rothman, BScOT, MHSC, Child Health Evaluative Sciences, Hospital for Sick Children, 555 University Ave, Toronto, Ontario M5G 1X8, Canada. E-mail:
Ms Rothman was responsible for the conceptual framework and study design, data acquisition, data analysis and interpretation, and writing and editing of the manuscript; Dr Macarthur contributed to the conceptual framework and study design, interpretation of data, and editing of the manuscript; Dr To contributed to the conceptual framework and study design, interpretation of data, and editing of the manuscript; Dr Buliung contributed to the conceptual framework and study design, interpretation of data, and editing of the manuscript; Dr Howard contributed to the conceptual framework and study design, interpretation of data, and editing of the manuscript; and all authors approved the final manuscript as submitted.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: This work was supported by a Canadian Institutes for Health Research (CIHR) doctoral research award, the CIHR Strategic Teams in Applied Injury Research, and the Ontario Neurotrauma Association Summer Internship Program.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
COMPANION PAPER: A companion to this article can be found on page 915, and online at www.pediatrics.org/cgi/doi/10.1542/peds.2014-0600.
- ↵Leitch K. Reaching for the top: a report by the advisor on healthy children and youth, 2007. Available at: www.hc-sc.gc.ca/hl-vs/pubs/child-enfant/advisor-conseillere/index-eng.php. Accessed February 15, 2014
- Public Health Agency of Canada
- ↵World Health Organization. World report on child injury prevention. Available at: www.who.int/violence_injury_prevention/child/injury/world_report/en/index.html. Accessed February 15, 2014
- ↵Transport Canada. Canadian motor vehicle traffic collision statistics, 2010. Available at: www.tc.gc.ca/eng/roadsafety/tp-1317.htm#3. Accessed February 15, 2014
- ↵Transport Canada. Pedestrian fatalities and injuries, 1992–2001, 2004. Available at: www.tc.gc.ca/eng/roadsafety/tp-tp2436-rs200401-menu-165.htm. Accessed February 15, 2014
- ↵Cragg S. 2004 National Transportation Survey. Available at: www.cflri.ca/media/node/593/files/National%20Transportation%20survey%20final%20report%202006%20formatted%20june%202.pdf. Accessed February 15, 2014
- ↵Active Healthy Kids Canada. Are we driving our kids to unhealthy habits, 2013. Available at: www.activehealthykids.ca/ReportCard/2013ReportCard.aspx. Accessed February 15, 2014
- Warsh J,
- Rothman L,
- Slater M,
- Steverango C,
- Howard A
- ↵National Center for Safe Routes to School. History of SRTS, 2012. Available at: www.saferoutesinfo.org/about-us/mission-and-history. Accessed February 15, 2014
- Gropp K,
- Janssen I,
- Pickett W
- ↵Rothman L, To T, Buliung R, Macarthur C, Howard A. Influence of social and built environment features on children’s walking to school: an observational study. Prev Med. 2013. Available at: http://dx.doi.org/10.1016/j.ypmed.2013.12.005. Accessed February 15, 2014
- Lin J-J,
- Chang H-T
- ↵Marler N, Montgomery R. Urban road traffic surveys: overseas road note 11. Berkshire, United Kingdom, 1993. Available at: www.transport-links.org/transport_links/filearea/publications/1_708_ORN%2011-%20.pdf. Accessed February 15, 2014
- ↵City of Toronto. Open data, 2013. Available at: http://www1.toronto.ca/wps/portal/contentonly?vgnextoid=9e56e03bb8d1e310VgnVCM10000071d60f89RCRD. Accessed February 15, 2014
- ↵City of Toronto. Building the new city of Toronto: final three-year status report on amalgamation, January 1998–December 2000. Toronto, Canada, 2001. Available at: www.toronto.ca/city_manager/pdf/amalgationreport.pdf. Accessed February 15, 2014
- ↵Base SAS 9.3, Procedures Guide, Statistical Procedures [computer program]. Cary, NC: SAS Institute Inc; 2012
- Kleinbaum D,
- Kupper L
- Hosmer DW,
- Lemeshow S
- Chatterjee S,
- Hadi A
- Agresti A
- Roberts I,
- Carlin J,
- Bennett C,
- et al
- Jacobsen PL
- Sirard JR,
- Slater ME
- Rothman L,
- Buliung R,
- Macarthur C,
- To T,
- Howard A
- Mueller BA,
- Rivara FP,
- Lii SM,
- Weiss NS
- Agran PF,
- Winn DG,
- Anderson CL,
- Tran C,
- Del Valle CP
- Laflamme L,
- Diderichsen F
- Braddock M,
- Lapidus G,
- Gregorio D,
- Kapp M,
- Banco L
- Rivara FP,
- Barber M
- Roberts I,
- Norton R,
- Taua B
- Bunn F,
- Collier T,
- Frost C,
- Ker K,
- Roberts I,
- Wentz R
- Jones SJ,
- Lyons RA,
- John A,
- Palmer SR
- Cloutier MS,
- Apparicio P
- von Kries R,
- Kohne C,
- Böhm O,
- von Voss H
- ↵World Health Organization. World report on road traffic injury prevention. Geneva, Switzerland, 2004. Available at: whqlibdoc.who.int/publications/2004/9241562609.pdf. Accessed February 15, 2014
- DiGuiseppi C,
- Roberts I,
- Li L
- ↵US Department of Transportation FHA. Highway Safety Improvement Program (HSIP) MAP-21 Interim Eligibility Guidance, 2012. Available at: www.fhwa.dot.gov/map21/guidance/guidehsip.cfm. Accessed February 15, 2014
- Copyright © 2014 by the American Academy of Pediatrics