Are school-based measures of walkability and greenness associated with modes of commuting to school? Findings from a student survey in Ontario, Canada

Abstract

Objectives

In Canada, students are increasingly reliant on motorized vehicles to commute to school, and few meet the recommended overall physical activity guidelines. Infrastructure and built environments around schools may promote active commuting to and from school, thereby increasing physical activity. To date, few Canadian studies have examined this research question.

Methods

This study is a cross-sectional analysis of 11,006 students, aged 11–20, who participated in the 2016/2017 Ontario Student Drug Use and Health Survey. The remote sensing-derived Normalized Difference Vegetation Index (NDVI), at a buffer of 500 m from the schools’ locations, was used to characterize greenness, while the 2016 Canadian Active Living Environments (Can-ALE) measure was used for walkability. Students were asked about their mode of regular commuting to school, and to provide information on several socio-demographic variables. Multivariable logistic regression models were used to quantify associations between active commuting and greenness and the Can-ALE. The resulting odds ratios, and their 95% confidence intervals, were adjusted for a series of risk factors that were collected from the survey.

Results

Overall, 21% of students reported active commuting (biking or walking) to school, and this prevalence decreased with increasing age. Students whose schools had higher Can-ALE scores were more likely to be active commuters. Specifically, the adjusted odds ratio (OR) of being an active commuter for schools in the highest quartile of the Can-ALE was 2.11 (95% CI = 1.64, 2.72) when compared with those in the lowest. For children, aged 11–14 years, who attended schools in high dwelling density areas, a higher odds of active commuting was observed among those in the upper quartile of greenness relative to the lowest (OR = 1.41; 95% CI = 0.92, 2.15). In contrast, for lower dwelling density areas, greenness was inversely associated with active commuting across all ages.

Conclusion

Our findings suggest that students attending schools with higher Can-ALE scores are more likely to actively commute to school, and that positive impacts of greenness on active commuting are evident only in younger children in more densely populated areas. Future studies should collect more detailed data on residential measures of the built environment, safety, distance between home and school, and mixed modes of commuting behaviours.

Electronic supplementary material

The online version contains supplementary material available at 10.17269/s41997-020-00440-0.

Introduction

In Canada, students’ commuting behaviours to school have evolved over the last 25 years and an increasing percentage of students now rely on motorized private vehicles. For example, in the Greater Toronto Area, the percentage of students who walk to school decreased from 53% in 1986 to 42.5% in 2006 among those aged 11–13, and from 38.6% to 30.7% among those aged 14–17 (Buliung et al. 2009). More recent trends are not widely reported; however, the number of students in the Greater Toronto and Hamilton areas who walk to school has decreased from 56% in 1986 to 39% in 2011 among those aged 11–13 and from 36% to 28% among those aged 14–17 (Metrolinx 2015).

Active commuting behaviours, such as walking and biking to school, provide important opportunities for children and youth to incorporate physical activity into their daily routines. Several studies have shown that increased physical activity improves physical and mental health (Warburton et al. 2006), and reduces the risk of developing several chronic diseases (Booth et al. 2012). Despite these benefits, only an estimated 9.3% of Canadian children and youth between the ages of 5 and 17 meet the recommended physical activity guidelines of 60 mins of moderate-to-vigorous physical activity per day (Public Health Agency of Canada 2016).

There are several strategies to promote increased physical activity. From a public health perspective, there has been a recent shift in focus where researchers and policy makers have begun to explore physical activity through a holistic lens by looking at “where we live” (i.e., our environment) rather than just “how we live” (i.e., health choices individuals and communities make) (James et al. 2017; Villeneuve et al. 2018). Creating urban environments with features that include green spaces, wider streets, well-lit sidewalks, parks, and gardens, and proximity to schools, offices, and shops has been shown to promote active commuting among adults (Saelens et al. 2003; Cerin et al. 2017).

A number of studies have assessed the impact of features of the built environment on utilitarian commuting in adults (Wang and Wen 2017; Smith et al. 2017; McCormack et al. 2019). Far fewer studies have explored how the built environment influences active commuting to school in children and youth. Those that have report a higher prevalence of active commuting to school among students who have better access to safe routes (Ewing et al. 2004), shorter routes (Oliver et al. 2014; Macdonald et al. 2019), and better neighboured cohesion (McDonald 2007). Together, these findings suggest that more walkable neighbourhoods may confer important health benefits by increasing physical activity levels in youth.

Ideally, both school and residential measures of the built environment are needed to understand the intricacies of active school commute (Larsen et al. 2009; Carver et al. 2019). Studies of walkability and active commuting to school have mostly relied on residentially based measures (Veitch et al. 2017; Barnett et al. 2019); very few have used both due to the complex nature of study design. One such recent Australian study that used both school and residential measures found that active school commute was associated with higher walkability around both home and school, direct travel, and close proximity to school (Carver et al. 2019). Similarly, a Canadian study reported a higher mix of land use around the school, residential population density, short distance, and being male were positively associated with active school commuting (Larsen et al. 2009). In Canada, publicly funded schools have catchment areas where the proximity of the student’s residence to the school usually determines where they receive their education. Therefore, school and residentially based measures of the built environment share many similarities, particularly in urban areas.

The beneficial effects of greenspaces for physical activity have been reported in children (Bringolf-Isler et al. 2018; Grigsby-Toussaint et al. 2011; Benjamin-Neelon et al. 2019). However, few of these studies investigated associations between green space and active commuting. A Canadian study found that the presence of street trees was positively associated with active school commute (Larsen et al. 2009). In contrast, in the Netherlands studies have found that both subjective (Aarts et al. 2013) and objective measures of greenness (Helbich et al. 2016) were inversely associated with active commuting.

Taken together, previous research suggests that some features of the built environment may encourage active commuting in youth. As youth attend school 5 days a week for 10 months a year, active commuting to school provides an important opportunity for increasing physical activity. Herein, we describe these associations using school-based, rather than residentially based, measures. Evidence-based findings from studies such as ours can support interventions at community levels that may contribute to improved population health.

Materials and methods

Study population and design

This is a cross-sectional study design that used data collected from participants of the Ontario Student Drug Use and Health Survey (OSDUHS). This survey has been conducted biennially since 1977 (Boak et al. 2018). The 2016–2017 OSDUHS survey collected data from 11,435 students (grades 7–12) in 214 publicly funded schools in Ontario between November 2016 and June 2017, using a stratified (region by school) and two-stage (school, class) clustering design. The sampling frame of the survey was designed to be representative of the approximately 1 million students in grades 7–12 in Ontario. The primary aim of the OSDUHS was to assess alcohol and other drug use among students. However, the self-administered questionnaire also collected data for a wide range of topics, including socio-demographics, mental health, physical health, and modes of commute to school (Boak et al. 2018). Sixty-one percent of students who were approached participated in the survey. More detailed information on survey design and data collection has previously been published (CAMH 2020; Boak et al. 2018).

Carleton University’s Research Ethics Board (CUREB) provided ethical approval for this project. The Research Ethics Boards of the Centre for Addiction and Mental Health (CAMH), York University, as well as 31 school board research review committees approved the OSDUHS 2017 protocol (Boak et al. 2018).

Ascertainment of commuting status

Participants were asked “How do you usually travel to school?” and the available response options were (a) by car, van, truck, SUV (as a passenger); (b) by car, van, truck, SUV (as a driver); (c) by school bus; (d) by public bus; (e) by walking; (f) by bicycling; (g) by subway or streetcar; or (h) other. We classified students as active commuters if they indicated that they either walked or cycled to school. Those who indicated that they used motorized transportation methods, such as a car, bus, or public transit, were classified as non-active commuters. Of the 11,435 participants, we were unable to classify the active commuting status of 422 students. Of these, 278 (2.4%) indicated that they used “other” or a “combination” of commuting modes, while the remaining 144 (1.2%) students did not provide a response.

Measures of built environment

The Canadian Active Living Environments (Can-ALE) dataset (Ross et al. 2018) was used to describe the “walkability” for the area around each school. The Can-ALE dataset has been developed for 2006 and 2016 and these metrics have been applied to cross-sectional data (Colley et al. 2019; Lukmanji et al. 2020). We applied the 2016 Can-ALE dataset as it was closest in time to our survey period. The Can-ALE dataset contains active living environment measures for all 56,589 Canadian census dissemination areas (Ross et al. 2018; Herrmann et al. 2019; DMTI Spatial Inc. 2016). Dissemination areas are the smallest reportable census area units and usually contain between 400 and 700 individuals (Statistics Canada 2017). These census-based geographies were used to create Can-ALE scores at a six-character postal code level across Canada (Ross et al. 2018).

The Can-ALE summary score is an aggregate measure of the three component scores for intersection density, dwelling density, and points of interest (Ross et al. 2018). Intersection density was defined as the connectedness of streets and walking paths through a community, captured by the number of three- or more-way intersections within a 1-km buffer from the centroid of a dissemination area. Data for intersection density, including national road networks and spatial data for transport infrastructure, administrative boundaries, and natural and topological features, were derived from Statistics Canada and OpenStreetMap (Ross et al. 2018). Dwelling density was defined as the average dwelling density of the dissemination area based on 2016 census data (Statistics Canada 2016). Points of interest (POIs) (e.g., shops, parks, schools, business, landmarks) were defined as the number of such places within 1-km buffers from the centroid of a dissemination area. Amenity features for this measure were derived from OpenStreetMap (Ross et al. 2018).

For our analyses, we assigned to the participants a school-based measure of greenness by using the Normalized Difference Vegetation Index (NDVI). During photosynthesis, plants prevent overheating by absorbing photosynthetically active radiation and reflecting near-infrared radiation. Characterization of the NDVI is done by measuring the infrared radiation and visible radiation reflected by the plant. The NDVI ranges in value from − 1 (absence of greenness) to 1 (full greenness). These metrics were derived using summertime observations of the NDVI when the cloud cover was less than 10% (Robinson et al. 2017). We used postal code-specific measures of the NDVI for 2016 that were assembled by Canadian Urban Environmental Health Research (CANUE) project (CANUE 2020). NDVI data were collated by CANUE from United States Geological Survey, Landsat 8 satellite via Google Earth Engine (Gorelick et al. 2017; United States Geological Survey 2017; United States Geological Survey 2015). Mean and maximum of annual mean NDVI were calculated at buffer distances of 100 m, 250 m, 500 m, and 1000 m from the school. Our analysis of greenness and active commute was done using a commonly used distance buffer of 500 m (Larsen et al. 2009) which would represent an approximately 15-min walk. Both Can-ALE scores and greenness measures were assigned to the centroid of the six-character postal codes for each school.

Other variables related to active commuting

Our analyses sought to control for a series of covariates that could confound the association between features of the built environment and active commuting. Covariates that were identified from survey responses and included were age, biological sex, ethnicity, region, and the month that the questionnaire was completed. Students were classified into four age groups: 11–12, 13–14, 15–16, and ≥ 17. Ethno-racial background was categorized into a binary variable: Caucasian and other (Asian [Chinese, South Asian, Filipino, Southeast Asian, West Asian/Arab, Korean/Japanese], Black, Indigenous, Latin American, and multiple). We also considered the length of time lived in Canada and living situation (i.e., lives in multiple homes) as potential confounders.

The survey was administered to students at various schools across the province, and a classification variable was created to describe these four areas (Greater Toronto Area, and Northern, Western, and Eastern Ontario). In addition, as active commuting behaviours vary by season, we created a variable that captured whether students completed the survey during the winter/fall (November–February) or spring/summer (March–June).

We also evaluated the role of contextual measures of socio-demographic characteristics on active commuting. Specifically, we modelled the Material and Social Deprivation Index (MSDI), an area-level measure of socio-economic index which is calculated from socio-economic indicators such as neighbourhood income, education, and proportion of single-parent families (Gamache et al. 2017). We adopted a similar approach used in previous analyses by classifying students into five MSDI categories (most deprived, deprived, middle, privileged to most privileged) (Gamache et al. 2017; Pampalon et al. 2012).

Statistical analysis

We first performed descriptive statistics of the survey responses, including tabulations of the number of active commuters across socio-demographic characteristics. Chi-square tests were applied to assess whether there were statistical significance differences in the percentage of active commuters across these characteristics.

As the frequency distributions of Can-ALE scores and greenness were not normally distributed, we used the Kruskal-Wallis method (Leon 1998) to test for differences in the medians across different modes of commuting.

Covariates for the final adjustment models were selected based on their association with both the outcome and the exposure using chi-square tests based on two-sided level of statistical significance (p < 0.05). A generalized estimating equation (GEE) population-averaged model with a binomial family, logit link function and an exchangeable correlation was used to account for possible correlations in the data among students in the same class. These equations were fit with quartile-based measures to describe the association between greenness, Can-ALE scores, and active commuting. Results were further stratified by age and sex.

All analyses were conducted using Stata version 15 (StataCorp LLC, College Station, TX, USA). Graphs were made in Microsoft Excel 2019 (Microsoft Corp., Redmond, WA, USA).

Results

Approximately one fifth (21.1%) of students reported that they actively commuted (walked or biked) to school. Younger students (aged 11–14) reported higher active commute (24.9%) than older students (aged 15–20) (17.4%). School bus (37.6%) and car (35.3%; 30.5% as passenger and 4.8% as driver) were the most commonly reported modes of commuting (Table 1).

Table 1.

Sample characteristics of participants of the 2016–2017 OSDUHS cycle (n = 11,006) by active commuting status

		Commuting status
Characteristic	Value	Active (n = 2323)		Non-active (n = 8683)		pa
		N	%	N	%
Age	11–12	435	18.7	1074	12.4	0.001
	13–14	914	39.4	2983	34.3
	15–16	643	27.7	2943	33.9
	17+	331	14.2	1683	19.4
Sex	Male	1128	48.6	3651	42.1	0.001
	Female	1195	51.4	5032	57.9
Race	Caucasian	1279	55.1	5010	57.7	0.048
	Other	924	57.7	3287	37.9
	Missing	120	5.2	386	4.5
Years lived in Canada	Have lived all their life in Canada	1885	81.2	7299	84.1	0.001
	Have not lived all their life in Canada	432	18.6	1378	15.9
	Missing	6	0.3	6	0.1
Living situation	Live only in one home	7391	86.1	1905	83.1	0.001
	Split between two homes	1196	13.9	388	16.9
Regionb	Greater Toronto Area	1103	47.5	3465	39.9	0.001
	Northern Ontario	227	9.8	1209	13.9
	Western Ontario	476	20.5	1521	17.5
	Eastern Ontario	517	22.3	2488	28.6
Material and social deprivationc	Most deprived	432	18.6	1699	19.6	0.054
	Deprived	580	25.0	2222	25.6
	Middle	381	16.4	1241	14.3
	Privileged	345	14.9	1199	13.8
	Most privileged	512	22.0	1993	23.0
	Missing	73	3.1	329	3.8
Month of participation	Nov–Feb	1388	59.7	4852	55.9	0.001
	Mar–June	935	40.3	3831	44.1

Active commuters are those who responded that they either walk (n = 2213) or bike (n = 110) to school. Non-active commuters are those who took public transit (n = 662), school bus (n = 4133), or travelled in a car (as passenger (n = 3358), as driver (n = 530))

^ap value derived from chi-square tests to assess differences in the distribution of each categorical variable between active and non-active commuters

^bGreater Toronto Area (GTA) (including Toronto, Durham, York, Peel, and Halton); Northern Ontario (including Parry Sound, Nipissing, and farther north); Western Ontario (including Dufferin County and farther west); and Eastern Ontario (including Simcoe County and farther east)

^cThe Material and Social Deprivation Index (MSDI), an area-level measure, was linked to the school address

Regionally, the overall proportion of students who were active commuters was highest in the Greater Toronto Area (24.2%) though the proportion was similar in Western Ontario (23.8%). In Eastern and Northern Ontario, there was higher reliance on automobiles. Graphical distribution of active commute showed students in the least walkable schools (quartile 1) were less likely to actively commute to school (Fig. 1). The proportion of students who were active commuters rose in the first 3 quartiles of Can-ALE index then dropped in the upper quartile (12.2%, 21.1%, 27.7%, and 23.0%). A similar trend was apparent for greenness with the percentage of active commuters across quartiles being 21.4%, 23.9%, 24.4%, and 14.8% (Fig. 1).

Fig. 1.

Percentage of participants who actively commute to school by region, and school-based quartiles of Can-ALE index, NDVI, and dwelling density

Distributional characteristics of measures of greenness and the Can-ALE index, including the mean, median, and standard deviations across different modes of commute, are presented in Table 2. There were statistically significant differences in the median values (p < 0.05) between active and non-active commuters for both Can-ALE (its components) and the NDVI.

Table 2.

Distributional characteristics of school-based measures of the Can-ALE index and NDVI for active and non-active school commuters among participants of the 2016/17 OSDUHS (n = 11,006)

Mode of transportation to schoolf	n	NDVIa		Intersection densityb		Dwelling densityc		Points of interestd		Can-ALE indexe
		Median	Standard deviation	Median	Standard deviation	Median	Standard deviation	Median	Standard deviation	Median	Standard deviation
Active commuteg	2323	0.41	0.08	57.3	37.4	926.6	637.6	39.0	51.8	0.19	1.45
Non-active commuteh	8683	0.42	0.07	43.2	42.7	730.5	680.0	34.0	62.1	− 0.38	1.70

^aBased on a 500 m buffer from the school location

^bIntersection density—number of ≥ 3-way intersections per square kilometre in the buffer around a dissemination area centroid

^cDwelling density—number of dwellings per square kilometre in the buffer around a dissemination area centroid

^dPoints of interest—number of points of interest in the buffer around dissemination area centroid

^eCan-ALE index—sum of the z scores of intersection density, dwelling density, and points of interest for 2016

^fKruskal-Wallis tests between active and non-active commute were statistically significant at a 5% level for school-based measures of the Can-ALE index, its components, and NDVI

^gActive commuters are those who responded that they either walk (n = 2213) or bike (n = 110) to school

^hNon-active commuters are those who took public transit (n = 662), school bus (n = 4133), or travelled in a car (as passenger (n = 3358), as driver (n = 530))

School-based measures of Can-ALE index were positively associated with active commuting. Students whose schools were in higher active living environments were more likely to be active commuters. Overall, we found that the odds of active commute were 2.11 times (OR = 2.11, 95% CI = 1.64, 2.72) higher for students in schools of the upper quartile of overall Can-ALE scores relative to the lowest (Table 3). The strength of the association was slightly stronger among younger children (Table 3), and among females when compared with their male counterparts (Supplementary Table 1).

Table 3.

Adjusted odds ratios of active commuting across quartiles of school-based measures of Can-ALE index and NDVI among participants of the OSDUHS, 2016–2017

			11–14 years of age (n = 5406)		15–20 years of age (n = 5600)		All students (n = 11,006)
		Active commutersa (%)	Odds ratiob	95% CI	Odds ratiob	95% CI	Odds ratiob	95% CI
Can-ALEc	Lowest quartile (0.379–1.165)	12.2	1.00		1.00		1.00
	Second quartile (1.166–2.253)	21.8	2.07	1.47–2.89	1.84	1.33–2.53	1.98	1.56–2.50
	Third quartile (2.254–3.287)	27.7	2.89	2.08–4.01	1.97	1.42–2.72	2.49	1.96–3.13
	Highest quartile (3.288–9.880)	23.0	2.23	1.57–3.18	2.06	1.45–2.92	2.11	1.64–2.72
NDVId
High dwellinge density areas (above the median)	Lowest quartile (0.233–0.330)	22.9	1.00		1.00		1.00
	Second quartile (0.332–0.384)	25.9	0.99	0.64–1.54	0.64	0.43–0.95	0.82	0.61–1.11
	Third quartile (0.386–0.425)	30.0	1.24	0.81–1.91	0.75	0.50–1.13	1.05	0.77–1.42
	Highest quartile (0.427–0.524)	34.5	1.41	0.92–2.15	0.58	0.35–0.94	1.06	0.77–1.46
Low dwellinge density areas (below the median)	Lowest quartile (0.249–0.417)	14.3	1.00		1.00		1.00
	Second quartile (0.419–0.487)	20.2	0.87	0.56–1.34	1.15	0.77–1.71	1.06	0.79–1.44
	Third quartile (0.490–0.523)	19.3	0.70	0.43–1.14	0.87	0.57–1.32	0.84	0.61–1.17
	Highest quartile (0.524–0.684)	13.9	0.40	0.24–0.68	0.78	0.49–1.24	0.59	0.41–0.83

Bold indicates statistical significance

^aActive commuters included those who walked (n = 2216) or biked (n = 110) to school. Non-active commuters included those who took public transit, school bus, or travelled in a car to school. Those who indicated they used mixed or other modes of commuting were excluded from analyses

^bThe odds ratios have been adjusted for age, ethnicity, biological sex at birth, years lived in Canada, living situation, region, and season

^cWalkability quartiles are based on the frequency distribution of the Can-ALE (Canadian Active Living Environments) index (sum of the z scores of intersection density, dwelling density, and points of interest for 2016)

^dNDVI (Normalized Difference Vegetation Index)/greenness based on 500 m buffer from the school location

^eBased on the dwelling density component of the Can-ALE index

Associations between school-based measures of greenness and active commuting were modified by the dwelling density of the school area. For example, among younger students (aged 11–14) who attended schools in high dwelling density areas, we found greenness was positively associated with odds of active school commute OR = 1.41 (95% CI = 0.92, 2.15), but did not reach statistical significance (p > 0.05) (Table 3). However, for the same age group, for those who attended schools in low-density areas, we found that the odds of active commuting was reduced when comparing those in the upper quartile of greenness with those in the lowest (OR = 0.40, 95% CI = 0.24–0.68) (Table 3).

Discussion

This cross-sectional study examined associations between school-based measures of the built environment, namely Can-ALE index and greenness, and self-reported active commuting to school. Students whose schools were in higher active living environments were more likely to be active commuters. The association between school-based measures of greenness and active commuting was modified by the number of dwellings, a surrogate for population density, around the school.

Perhaps our most important finding is the low prevalence of active commuters. Only one in five students (21.1%) were found to be actively commuting to school. This estimate falls outside the lower end of the range of findings from other Canadian studies that reported prevalence estimates between 23% and 72% (Wong et al. 2011; Larsen et al. 2009; Cozma et al. 2015; Buliung et al. 2009). Wong et al. (2011), who used earlier data from the OSDUHS (2009 cycle), reported that 47% and 38% of elementary school students actively commuted to and from school, respectively, while the corresponding estimates were 23% and 32% among secondary school students. Our lower estimates may be explained by continuing secular trends of decreasing active commuting over time, and suggests that further rapid declines in active commuting in Ontario have occurred over the last decade. There are likely a wide range of factors that have contributed to these trends, including individual- and community-level characteristics (Larsen et al. 2009). While Wong et al. highlighted the need for an analytic distinction between morning and afternoon commuting behaviours, the more recent cycles of the OSDUHS do not collect separate data for before and after school commuting, and therefore, we were unable to explore these differences.

We recognize that distance to school is perhaps the most important predictor of active school transportation (Aarts et al. 2013; Trapp et al. 2012). For example, Larsen et al. (2009) reported 72% of students actively commuted from and 62% to school and Cozma et al. (2015) found an overall prevalence of 63% for students who resided within 1 mile (1.61 km) of their school. While distance to school information was not collected in our survey, dwelling density around the school likely represents a proxy-based measure of average distance to the school. In our study population, the prevalence of active commuting across increasing quartiles of high dwelling density was 22.9%, 25.9%, 30.0%, and 34.5%. We advocate that future cycles of the OSDUHS consider capturing students’ residential location so that distance and travelling time to school can be modelled.

Our study also found that younger students reported a higher overall percentage of active commuters than older students (24.9% vs 17.4%). Similar findings by age have been reported in Canada (Buliung et al. 2009; Wong et al. 2011). There are a number of possible factors that contribute to these age-related differences in active commuting. Younger children will be incapable of driving themselves to school, and for many younger children, parents may be reticent to have them take public transportation. Previous research has shown that having an older sibling, or a nearby child of the same age, is an important factor that may increase active commuting in younger children (Aarts et al. 2013). Unfortunately, these data were not available to evaluate.

The active living environment measure (Can-ALE) that we modelled was positively associated with active commuting indicating that three components that comprise this measure are correlated with active commuting. Similar associations have been reported elsewhere, for instance, active school commute has been positively associated with street connectivity (Giles-Corti et al. 2011), traffic exposure (Christiansen et al. 2014; Giles-corti et al. 2011), proximity to walking tracks (Trapp et al. 2012; Barnett et al. 2019), speed limits (Jauregui et al. 2016), diversity of routes (Aarts et al. 2013), residential density (Christiansen et al. 2014), population density, land-use mix (Jauregui et al. 2016), road connectivity, and proximity to amenities/city centre/schools (Barnett et al. 2019; Trapp et al. 2012). The different methods that have been used to describe features of the built environment on active commuting make it challenging to compare findings across studies. Nonetheless, future analyses of the OSDUHS should explore to what extent different components of walkability are associated with these different features of the built environment. This would require more detailed survey data on commuting routes, and mixed modes of commuting (i.e., walk to bus stop and bus route).

We found that associations between school-based measures of greenness and active commuting were modified by both age and dwelling density. As described above, these differences may reflect the fact that dwelling densities around a school represent the average distance to the school for the students. The other Canadian study we identified reported similar findings, namely, that the presence of street trees was positively correlated with active commute among students who resided within 1 mile (1.6 km) of their schools (Larsen et al. 2009). When evaluating the associations with greenness, it is important to recognize that the characteristics of green spaces may differ substantially based on the population density. The NDVI that we used in our analyses captures overall vegetation based on a sky view metric. Other features of greenness, including accessibility, aesthetics, safety, and design (parks, walking paths), may be important determinants in promoting active transportation to school. While we were unable to describe these features given the data we had, future studies should consider exploring the role of these characteristics of greenness, or modelling other greenness-based metrics such as the Green View Index measure (Villeneuve et al. 2018).

Stratified analysis showed that the positive associations of the Can-ALE index and active commuting were modestly stronger among female students than among their male counterparts. This finding differs to other findings in the UK (Potoglou and Arslangulova 2017), Canada (Larsen et al. 2009), and Australia (Trapp et al. 2012), but is consistent with a Japanese study (Susilo and Waygood 2012). There are many factors that could contribute to differences between the sexes, including external factors such as parental perception of safety, their own self-efficacy, availability of a walking companion, and family time constraints (Potoglou and Arslangulova 2017; Larsen et al. 2009; Trapp et al. 2012; Susilo and Waygood 2012). Further efforts to better understand how features of the built environment differentially impact active commuting between boys and girls would be helpful.

Conclusion

Our study is among the few to assess school-based walkability, greenspace, and active school commute among children and youth in Canada. Overall, our findings suggest that students who attend schools that are located in areas of high-active living index scores are more likely to actively commute to school. Furthermore, they suggest that the beneficial impacts of greenness on active commuting vary by age, and population density. Efforts to enhance walkability around existing schools or to build new schools in areas that support active living may help increase physical activity through active commuting to school in urban areas. Other strategies should be pursued for those who live in suburban or rural areas to promote increased levels of physical activity.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflict of interest.