Article Text

Download PDFPDF

Neighbourhood street connectivity and injury in youth: a national study of built environments in Canada
  1. Graham Mecredy1,
  2. Ian Janssen1,2,
  3. William Pickett1,3
  1. 1Department of Community Health and Epidemiology, Carruthers Hall, Queen's University, Kingston, Ontario, Canada
  2. 2School of Kinesiology and Health Studies, Queen's University, Kingston, Ontario, Canada
  3. 3Department of Emergency Medicine, Angada 3, Kingston General Hospital, Kingston, Ontario, Canada
  1. Correspondence to Dr William Pickett, Emergency Medicine Research, Queen's University, Clinical Research Centre, Angada 3, Kingston General Hospital, 76 Stuart St., Kingston, ON K7L 2V7, Canada; will.pickett{at}


Background The influence of the built environment on health is of contemporary societal interest. The design of streets in neighbourhood settings may contribute positively to the health of populations through increased physical activity, but it may also have injury consequences.

Methods We conducted a national cross-sectional study to describe the injury experiences of 9021 students from 180 Canadian schools that participated in the 2006 Health Behaviour in School-Aged Children survey. Street designs surrounding each school (5 km circular buffer) were estimated via geographic information systems for three established measures of connectivity (intersection density, average block length and connected node ratio). A composite scale of connectivity was derived using factor analysis. Multilevel logistic regression analyses were used to examine the associations between the composite connectivity measure and students' reports of physical activity injuries occurring in the street (street injuries).

Results Students living in neighbourhoods with low versus high street connectivity reported possible increases in the occurrence of street injuries (OR, 1.38; 95% CI, 0.84 to 2.26). This relationship was mainly attributable to the occurrence of bicycle injuries (52% of all street injuries; OR, 2.33; 95% CI, 1.28 to 4.25). The population attributable risk was 20% for street injuries potentially caused by living in an area with low connectivity.

Conclusion The design of streets, as a measure of the built environment, is related to the occurrence of youth injury. Positive effects of poorly connected street designs that are likely in terms of physical activity were offset by negative injury outcomes, although the injuries observed were mostly minor in nature.

  • Street connectivity
  • injury
  • youth, population health
  • child, head injury
  • socioeconomic status
  • psychological
  • surveillance

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Built environments, defined as physical aspects of the surroundings in which we live our daily lives, have been recognized as a possible determinant of health.1 Despite the widespread acceptance of environmental modification as an effective approach to injury prevention,2 little research in the emerging field of the built environment and health has focused specifically on its potential contribution to youth injury.1

One aspect of the built environment that may be particularly relevant for etiological studies of youth injury is a construct called street connectivity, the directness of links and the density of connections in paths or road networks.3 A highly connected neighbourhood street design has many short links, numerous intersections and few dead ends.3 Such designs are correlated with positive health outcomes related to increases in walking, bicycling and overall physical activity levels in adult populations.4–6 This had led some to promote highly connected street designs as optimal for health.4–6 However, as one might expect, highly connected street designs have also been related to higher risks of fatal pedestrian injury in general and adult populations.7–9 Experiences in youth populations appear to be somewhat different. Youth living in neighbourhoods with highly connected streets engage in less physical activity outside schools than youth living in neighbourhoods with poorly connected streets.10 While the effects of street connectivity on the occurrence of youth injury have not been studied directly, previous studies have demonstrated a link between travel patterns, transportation infrastructure and traumatic injury reported for children and youth, particularly for child pedestrian injuries11 12 and bicycling injuries.12–14 This suggests that the design of street networks may contribute to varying risks for injury among youth populations.

We had the opportunity to study the relationship between street connectivity and youth injury in a national study of young Canadians. Our hope was to provide foundational information to support urban planning efforts that aim to maximize opportunities for safe and healthy living through attention to the built environment and specifically in the design of street networks in residential neighbourhoods. When considered with other evidence (eg, that surrounding street connectivity and adult injuries7–9 as well as evidence describing street connectivity and physical activity levels across the lifespan4–6), such information may point to specific community designs that are optimal for health from a broad perspective.


Data sources

This was a national study. We examined the health experiences of individual children nested within areas defined by their school neighbourhoods. Individual (level 1) data on the injury and other health experiences of Canadian schoolchildren were obtained from the 2006 Canadian Health Behaviour in School-Aged Children Survey (HBSC).15 This survey compiled information on health behaviours, health outcomes and contextual determinants of health among students in grades 6–10 (approximate ages of 11–15 years) from publicly funded schools sampled from all Canadian provinces and territories. A systematic single-stage cluster sample was used to identify the sampling unit of classes within schools.15 Exclusions were students enroled in private schools and special needs schools, those who were homeschooled and students who were absent on the day of the survey (∼9% of this population).15 These exclusions were made to maintain consistency of the 2006 sampling frame with past HBSC survey cycles in Canada. Students agreed to participate by completing and submitting the HBSC student questionnaire in classroom settings. This was done following the submission of signed consent forms from a parent or another responsible adult. The response rate at the individual student level was 74% (9672/12 800; the 9672 participating students were nested within 186 elementary and high schools). Of these, six schools were excluded due to record linkage errors associated with school identification numbers. An additional 93 students were missing data on the main injury outcome of interest. This left 9021 students (93.3%) from 180 schools (96.8%) available for analysis.

Area (level 2) measures describing street connectivity were obtained from the CanMap Streetfiles (DMTI Spatial Inc., version 2008.3, DMTI Spatial Inc., Markham, Ontario, Canada) geographic information system (GIS). Street connectivity was measured in a 5 km circular buffer around each of the 180 participating HBSC schools using the ArcGIS software (ESRI, version 9.3, ESRI, Redlands, California, USA) and the latter street files. A network layer of intersection nodes was created and added to the buffer map for each school. A 5 km buffer has been applied successfully in previous HBSC studies and is considered reliable for neighbourhood constructs.16–19

Street connectivity measures

Three standard street connectivity measures were obtained: intersection density, average block length and connected node ratio.3 Intersection density is defined as the ‘number of intersections per unit of area’3 and is calculated by dividing the number of true intersections by the total land area. A higher number indicates more intersections and higher street connectivity. Average block length is the mean length of blocks in the area and is calculated by dividing the sum of the street length per unit area by the number of intersections per unit of area. Shorter blocks mean more intersections and a greater number of routes available, showing higher street connectivity. Connected node ratio is the number of street intersections divided by the number of intersections plus cul-de-sacs.3 The maximum value for this variable is 1, with higher numbers indicating that there are few cul-de-sacs and dead ends and higher connectivity.

Street injuries

Participants were asked about their one most serious injury (if any) that was medically treated and occurred in the 12 months preceding the survey. Those who reported at least one medically treated injury answered a series of supplemental questions about their most serious injury event, starting with “where were you when this one most serious injury happened?” Those who responded as being injured “in the street/road/parking lot” and “during a physical activity (biking/cycling, playing sports/recreational activity, roller blading, skateboarding and walking/running)” were categorized as having a street injury. Injury severity was inferred from responses to questions regarding the medical treatment of injuries and the number of days missed from school or other usual activities. Injuries were considered to be more serious if they required medical treatment such as a cast, stitches, surgery, or an overnight hospital stay or if they caused the individual to miss at least 1 day of school or other usual activities.

Potential covariates

Level 1 variables considered a priori as potential confounders were gender, grade, family socioeconomic status measured using an existing affluence scale derived from individual student responses,20 perceived neighbourhood safety and perceived neighbourhood aesthetics. The latter two questions were also reported by individual students who responded to a series of Likert-like items that summarized their neighbourhood perceptions with respect to the presence (lots, some, or none) of various hazards. These items were treated as potential confounders as they could be direct risk factors for injury that are plausibly associated with the primary exposure of interest (street connectivity) but were unlikely to be in the causal chain linking exposure and outcome.

Potential level 2 confounders were neighbourhood-level socioeconomic status, geographic location and numbers of parks and recreational facilities in neighbourhoods. Level 2 variables were measured via the 2001 Canadian Census of Population,17 Statistics Canada's Census Metropolitan Area data21 and the CanMap Streetfiles.

Statistical analyses

Statistical analyses were performed in SAS version 9.2 (SAS Inc.). Exploratory factor analyses were used to develop an overall street connectivity score for the area surrounding each participating school. Bivariate multilevel models were fit to describe the relationship between measures of street connectivity and street injuries. A hierarchical series of models was then developed, following a systematic approach: (1) model 1 was controlled for all level 1 covariates, (2) model 2 was controlled for all level 1 and level 2 covariates and (3) model 3 was then fit using all level 1 and level 2 covariates and eliminating variables via backwards elimination methods. Variables whose removal from the model caused a >10% change in the street connectivity effect estimate were retained. Model 3, therefore, considered the street connectivity scale as well as one covariate that significantly contributed to the final model (parks/recreational facilities). Geographic location was not retained in the final model, although it was clearly associated with the street connectivity measure in exploratory analyses. Since the outcome of street injuries is relatively uncommon (<5%), ORs obtained from the regression models were interpreted as relative risks. As the amount of missing data was generally modest in these models (<10%), analyses were simply based upon all available data as opposed to an imputation method.

The SAS GLIMMIX procedure was used to fit multilevel (mixed) logistic regression models with a binomial distribution and the logit link specification. After testing for model fit, we employed a random intercepts model but assumed fixed effects for β coefficients. This model simultaneously accounted for the clustered nature of our sampling design (individual students nested within schools). All multilevel logistic regression models used a Newton–Raphson with ridging technique to aid convergence.22 Results for the main associations under study are presented as adjusted ORs and associated CIs. Emphasis was put on the interpretation of the effect estimates and associated CIs, as opposed to statistical significance alone.

Population attributable risk (PAR) was calculated to estimate the proportion of street injuries potentially attributable to living near a school with lower street connectivity. PAR was calculated based on the results of model 3 using the following equation: PAR=Pe(OR−1)/1+Pe(OR−1), where Pe is the proportion of youth living in an area with low connectivity, and OR is the OR of experiencing a street injury in a low- versus high-connectivity neighbourhood.23

Multivariate models were also created using the specific physical activity injury outcomes—biking/cycling injuries, sports/recreational injuries, roller blading/skateboarding injuries and walking/running injuries. For consistency, covariates from the main multivariate model (model 3) were retained in these models.


Composite measure for neighbourhood street connectivity

Each of the 180 schools was ranked on each of the three connectivity measures. Principal component factor analysis suggested strong agreement between the three ranked variables; factor loadings were 0.96, 0.95 and 0.73 for the intersection density, average block length and connected node ratio variables, respectively (Cronbach's α=0.86). The ranked variables were combined with equal weight to create the composite measure, which was then divided into quartiles. Due to the homogeneity of risk estimates in quartiles 1–2 then 3–4, these quartiles were combined into two groups (high connectivity vs low connectivity) to maximize statistical power. Figure 1 illustrates differences in street layout and design between high- and low-connectivity neighbourhoods.

Figure 1

Comparison between street layouts and designs in typical neighbourhoods with high connectivity (1) versus low connectivity (2).

Sample description

Distributions of the participants according to the characteristics of students and school neighbourhoods are shown by level of street connectivity in table 1. Differences in street connectivity were observed by grade, family affluence (socioeconomic status), neighbourhood safety, perceived amounts of litter in neighbourhoods and perceived housing conditions (all p<0.01). The number of parks and recreational facilities was significantly associated with street connectivity (p<0.0001); neighbourhoods with the highest number of parks and facilities all had high street connectivity scores (p<0.0001). Conversely, all neighbourhoods in rural geographic locations had low street connectivity scores (p<0.0001).

Table 1

Relationships between street connectivity and characteristics of individuals (level 1) and schools (level 2), 2006 Canadian HBSC survey (n=9021 students)

Street injuries

Distributions of street injuries are further described by gender and grade in table 2. Overall, only 2.5% of the students reported having at least one medically treated street injury, and 52.0% of these occurred while bicycling. Most injured students (80.7%) did not report missing any days of school or other regular activities due to their injury.

Table 2

Description of street injuries by gender and grade, 2006 Canadian HBSC Survey (n=9021 students)

Multilevel logistic regression analysis

Table 3 presents a hierarchical series of logistic regression models. Based upon the direction of the ORs and their asymmetric 95% CIs, models 2 and 3 suggest that students living in neighbourhoods with low versus high street connectivity may have experienced higher relative odds for street injury, although this analysis was somewhat limited in terms of statistical power.

Relationships between street connectivity and the occurrence of specific types of street injury, defined by activity, are shown in table 4. Compared to students in neighbourhoods with high street connectivity, those in low-connectivity neighbourhoods were more likely to report a bicycle injury that occurred in the street (OR, 2.33; 95% CI, 1.28 to 4.25). None of the other effect estimates were suggestive of an increased risk for injury associated with low street connectivity.

Table 3

Multivariate analysis: associations between street connectivity and street injuries, 2006 Canadian HBSC Survey (n=9021 students)

Table 4

Relationships between street connectivity and activity at the time of injury, 2006 Canadian HBSC Survey (n=9021 students)


Study findings suggest that young Canadians who live in neighbourhoods with poor street connectivity may have experienced higher relative odds of street injury compared with their peers from more connected neighbourhoods. Assuming that the relationship is causal, up to one in five of all street injuries was potentially attributable to students living in such low-connectivity areas. Relationships between street connectivity and street injuries mainly resulted from injuries sustained while bicycling. When applied to the Canadian youth population as a whole, only 2.5% of the population under study were injured in the street, and 1.3% were injured in a cycling event. In addition, the majority of the observed street injuries (∼81%) did not require medical treatment or result in missed activities. The overall impact of low street connectivity on injury morbidity in youth appears to be quite modest.

Strengths of this analysis warrant comment. This study was national in scope, involving the experiences of approximately 9000 children nested with 180 schools sampled from across our country. Second, we employed contemporary multilevel modelling and GIS techniques to examine the injury experiences of young people as a function of their personal and neighbourhood circumstances; such analyses are rare in the paediatric injury control literature. Third, our study represents one of the few existing analyses of built environment features as a possible determinant of health, with a specific focus on the injury experiences of young people. This attempts to fill a void in the literature encompassed by the field of built environments and health and in the more general field of population health research.

Limitations of this study include the cross-sectional nature of HBSC survey data, which limits our ability to make causal inferences. Second, due to the low observed prevalence of street injuries, this analysis was limited in terms of statistical power, and not all effects interpreted as positive achieved statistical significance. Third, there is potential for area-level associations to be residually confounded by variables not captured in this research. An example of this is ethnicity, which has been shown to affect risks for youth injury.24 25 The analysis also does not account for amounts of exposure to cycling and other activities that occur in street environments. Fourth, students who had a severe injury prior to survey would be more likely to be absent on the day of the survey. Such missing data would be likely to attenuate any of the observed effect estimates likely towards no effect. Fifth, the 5 km radius used around schools was a proxy for residential neighbourhoods; this would have resulted in area-level characteristics being improperly ascribed to those students who did not live within this radius. The resulting exposure misclassification would also lead to bias in the effect estimates, likely towards no effect. Studies of built environments are in their infancy, and it was unclear what type of buffer should be used (radial vs street network buffers)26 and what radial distance around schools would be most appropriate in this context. In addition, our analysis assumes that the 5 km buffer adequately represented the neighbourhood where the observed street injuries occurred. If this assumption is not valid, it, too, would lead to non-differential misclassification. Finally, as this was a Canadian study, care should be taken in generalizing findings beyond our national context due to likely variations in child physical activity, built environment features and other factors that might influence the relationship under study and would be expected to vary across countries.

Our findings suggested that street connectivity was highly correlated with urban–rural geographic location. There is a possibility that both a geographic effect and a street connectivity effect exist and that these two effects are inter-related. Street connectivity could be one of several mechanisms by which geographic locations could be related to injury outcomes.

Because most of the observed street injuries did not result in medical procedures or multiple days missed from school or usual activities, they were considered to be of minor severity. Such injuries are unlikely to involve collisions with motor vehicles, which has been the focus of most existing analyses of this type.27–29 It is therefore difficult to compare our findings with such existing studies, as these existing studies mainly focus on experiences in general populations, pedestrian fatalities and major trauma.7–9 Major pedestrian injuries were reported only rarely by members of our study population, and mechanisms that underlie the association between street connectivity and injury are likely different among young people than within adult and more general populations. To our knowledge, no existing studies have focused upon street designs as a specific determinant of injury in youth populations, making our study unique to the biomedical literature.

Our study of the built environment and street injuries among the youth has implications for prevention. The non-severe nature of the reported injuries was actually a positive finding in terms of population health. As such, it may be more prudent to focus on encouraging safe transport, play and physical activity rather than advocating a change in neighbourhood design. For example, if the increase in bicycling injuries in low-connectivity neighbourhoods is largely attributable simply to the frequency of bicycle use, then a logical intervention would be to increase the safety of both the rider and the bicycle itself.30 This focus may be the most practical and cost-effective method to reduce youth physical activity injuries occurring in the street, potentially targeted towards poorly connected neighbourhoods. In the 2001/2002 cycle of the Canadian HBSC, approximately 70% of bicycle injuries occurred during the warm weather months of April through August, suggesting a need to consider seasonal effects in such interventions.

With regard to built environment interventions, increased safety of neighbourhoods, through alterations such as improved street maintenance and segregated bicycle lanes/paths,14 has obvious potential to decrease youth street injuries as well. While these interventions would be more costly to implement, they may ultimately be more effective in altering youth behaviour and minimizing injuries.31


This national, multilevel analysis suggested that youth living in neighbourhoods with lower street connectivity were more likely to experience an injury while playing in the street. This relationship was largely attributable to bicycling injuries, which accounted for 52% of the physical activity injuries that occurred in the street. Because most observed injuries were minor, continued development and improvement of safe play and bicycle safety initiatives is suggested as the most immediate way to reduce the occurrence of street injuries. This should be combined with known efficacious infrastructure changes that promote separation of bicycles from traffic, even in poorly connected neighbourhoods.

What is already known on this subject

  • The built environment and its relationship with injury are an emerging health topic. Few studies of the built environment as a determinant of youth injury exist in the biomedical literature

  • Adult studies have demonstrated that features of the built environment can lead to fatal pedestrian injuries. One built environment feature worthy of study is street connectivity

  • Few studies have examined street connectivity as a potential determinant of injury, especially in youth populations

What this study adds

  • In this national study, youth living in neighbourhoods with poor connectivity were more likely to experience a street injury. This relationship was largely attributable to bicycling injuries

  • Public health interventions should focus on the optimization of street designs but more immediately on safe play and bicycle safety interventions in neighbourhoods where street injuries are common


We thank Laura Seliske and Marianne Nichol for their intellectual contributions with respect to measurement of the built environment.



  • Funding This study was funded by an operating grant from the Canadian Institutes of Health Research (MOP 97962) and a second operating grant cofunded by the Canadian Institutes of Health Research and the Heart and Stroke Foundation of Canada (PCR 101415). In addition, Graham Mecredy was supported by a Master's Studentship Award from the Heart and Stroke Foundation of Ontario. Ian Janssen was supported by investigator awards from the Canadian Institutes of Health Research and the Ontario Ministry of Research and Innovation. The Health Behaviour in School-Aged Children Survey (HBSC), a World Health Organization/European Region collaborative study, was funded in Canada by the Public Health Agency of Canada (Contract: HT089-05205/001/SS). The principal investigator of the 2006 Canadian HBSC was William Boyce, and HBSC is coordinated internationally by Candace Currie (University of Edinburgh).

  • Competing interests None.

  • Ethics approval This study was approved by the Queen's University Research Ethics Board.

  • Provenance and peer review Not commissioned; externally peer reviewed.