Associations between the built environment, total, recreational, and transit-related physical activity

The environment in which we live is a major determinant on overall health and well-being [1]. Ecologic models acknowledge that there are individual and community-level factors that may influence health behaviours [2]. While the pathways to health and disease may be both direct and indirect, when taken together, emerging evidence suggests that patterns of discretionary physical activity (PA) participation may be significantly influenced by those features of the physical and social environment associated with PA accessibility [3, 4]. In order for the inactive portion of our society to effectively shift towards one that is more active, strategies for PA promotion must be targeted beyond individual behavior change to incorporate strategies that engage whole health organizations, institutions, and communities.

Current estimates suggest that between ~3% and 28% of the variance in PA participation can be accounted for by neighbourhood-level differences in the built environment [5, 6], consistent with ecologic models which view PA as having multiple influences; including intrapersonal, interpersonal, policy, and environmental components [7]. Moreover, research suggests that the role of the built environment in influencing opportunities to engage in PA varies [8], as neighborhoods may provide constraints and opportunities for different forms of PA [9]. While many vigorous forms of PA are engaged in for recreational or health-related purposes, walking and cycling remain two of the most common forms of PA [10] and can be done for leisure, recreation, exercise, occupational, or transport-related reasons. In turn, this may make walking and bicycling more susceptible to environmental influences [11]. In addition to studying leisure-time PA, transport-related activities, such as walking or cycling, are alternate means to increase total daily energy expenditure [12].

While a number of studies have assessed the impact that both individual-level and neighborhood-level characteristics have on the frequency, duration, and engagement in PA [5, 6, 13, 14], this work has largely been limited to assessing walking behaviour [5, 6], or used multiple data sources to additional questionnaires to obtain demographic and PA data [13, 14]. Moreover, little work has been done within the Canadian context with demographic and PA information obtained during the same question period. Therefore, the aim of the current analysis is to quantify the association between different measures of the built environment with leisure-time and transport-related PA.

The demographic characteristics of the respondents and characteristics of the built environment are presented in Table 1. A total of 1 158 respondents were included in the analyses (μ_age = 47.9 y; μ_BMI = 25.6 kg · m^-2, and; 49.6% female). In general, the majority of the sample was white (69.6%) and well-educated (80% with at least post-secondary education). Nearly half of the sample had never smoked (44.8%) and were either moderately or sufficiently active (47.7%).

Table 2 gives an overview of the outcome variables. Respondents spent an average of 17.7 minutes/day either walking or cycling for leisure and their median daily energy expenditure was 1.3 KKD for leisure-time pursuits. Overall, 69.8% of respondents participated in any form of walking or cycling for leisure and 14.4% engaged in any form of walking or cycling for transport-related pursuits. The PA variable was not transformed since the continuous KKD was calculated based on categorized approximations for the amount of time spent in certain activities and was not originally assessed on a continuous scale. Furthermore, it would be more difficult to apply the results from a transformed-PA variable to real-world applications.

It has been recommended that people who are physically inactive should start with short sessions (5–10 minutes) of PA before building-up to longer durations of activity [18], and even walking at a brisk pace for 5 (~500 m) [13, 14, 21] or 10 minutes (~1 km) may be associated with significant health benefits [18]. In a buffer zone of 500 m fixed at the centroid of the respondent’s postal code address, higher residential density and intersection frequency was associated with greater odds of engaging in walking or cycling for transportation purposes, whereas intersection frequency was associated with walking or cycling for leisure-time. Consistent with previous literature [5, 6, 11] hectares of building area were found to be negatively associated with energy expenditure. Although some previous studies have also demonstrated significant associations between walking and bicycling for both leisure-time and commuting purposes with area of green and recreational space [6, 14], no significant associations were found with either parks/green space measure and any form of walking or cycling for leisure-time or transport-related purposes.

These previous results notwithstanding, a greater amount of land dedicated to building area was associated with lower overall LTDEE, a finding that may be accounted for in part by an increase in sensitivity of the retail measure to identify pedestrian activity. Previous research calculated a ratio of the retail building floor area to the retail land floor area footprint with the rationale that a low ratio would indicate more parking, and a high ratio would indicate less surface parking and fewer setbacks which in turn promotes pedestrian activity [22]. However, the current analyses could not distinguish for what purpose the building area was designated, nor could a ratio be calculated, as only ground-level area footprint data was available. Therefore, the finding that higher levels of building area was related to lower levels of LTDEE may suggest that building area impacts the frequency or duration that respondents can engage in leisure-time PA. This may be possible if we assume higher levels of building area also require more space dedicated to parking or streets dedicated to vehicle traffic without accounting for pedestrian activity. While the significant association between building area and LTDEE was small, the results suggest that increasing the area within a local neighborhood dedicated to building space may hinder PA levels. However, without taking into account destinations to travel or a building footprint ration, further research is necessary to clarify the association between building area and LTDEE.

National health surveillance surveys have relied on sampling strategies that typically draw data from administrative boundaries or sociodemographic variables, thereby disregarding variations in distribution of built environment variables. Using spatial sampling techniques over neighborhood-based methods may require smaller sample sizes, be more cost-effective, and may be more generalizable to the broader population since spatial sampling does not impose a pre-specified neighborhood from which to draw a study sample [3]. Therefore, a major strength of the current analyses was the use of a population-based sample of respondents from the CCHS; however, in order to more fully characterize these relationships, future studies would benefit from exploring interactions between age, sex, socioeconomic, and ethnic subgroups to better understand factors that may account for additional neighbourhood-level variation in PA. Previous research has shown that rates of walking and cycling of children to school are inversely related to socioeconomic status [23, 24]. Understanding features of the built environment that either promote or constrain PA may assist in closing the gap in patterns of PA participation in traditionally marginalized groups. Finally, the CCHS sampled respondents in all 9 municipalities of York Region during each month of the 2-year collection cycle. Although this mode of sampling is unlikely to impact on the spatial relationships observed here, future research would benefit by taking into account the month of data collection to account for seasonal effects on physical activity.

Several limitations of this study also warrant mention. First, while these analyses included a diverse age range and the sampling strategy incorporated a representative socioeconomic sample of the York Region population, it remains unclear whether the observed associations are similar across all subgroups of the population. Therefore, further research is necessary to examine whether different populations experience the same built environment differently (i.e. do subpopulations living in the same neighborhood and sharing the same space interact with the built environment distinctly from one another?). Due to sample size restrictions it was not possible to explore these interactions in the present analysis. Second, since these analyses were conducted using cross-sectional survey data, causality cannot be inferred. However, when taken together, they highlight the need for additional multilevel analyses to confirm the associations between physical activity and the built environment within rapidly growing, diverse regions of Canada. By combining multiple CCHS survey cycles, a clearer understanding of subgroup variation may be possible. While this approach (with different respondents for each survey cycle) could not be interpreted as longitudinal in design, when repeated in other regions in Canada, it could provide period estimates that could provide important insight for exploring changes in physical activity with changes in neighborhood landscape. Although the mode of sampling is unlikely to impact on the spatial relationships observed here, future research would benefit by taking into account the month of data collection to account for seasonal effects on physical activity. In addition, as with all survey data, social desirability bias should be acknowledged. Finally, it is important to note that respondents are not asked the reason for selecting their current address as the question is not addressed during the CCHS interview process. It is unclear if the resident has selected their household for reasons based on neighborhood safety, proximity to schools, traffic flow, socio-cultural factors, aesthetics, or a variety/combination of many other factors that relate to homeownership and neighborhood dwelling. While perceived neighbourhood safety has been found to be associated with physical activity [4, 25‐27], it could not be assessed within these analyses.

Future work should explore expanding the buffer zone to cover a larger walking distance provided sufficient data can be obtained to incorporate more measures of the built environment, thus providing a more robust estimation of the local landscape. It is possible that the magnitude of effects found within these analyses would differ depending on the size of the buffer zone. It would be helpful to assess points of interest in regard to walking distance within the local neighborhood. Researchers investigating associations between leisure- and transport-related physical activity with places to commute would benefit from calculating buffer zones based on network distance as this provides a more accurate indication of the routes realistically travelled. Road network data containing information on sidewalks and park trails/paths would help define pathways that would be exclusively used by cars and those that could be used for pedestrian and cyclists.

The current analyses found that living in the highest regions of intersections and residential density was associated with higher levels of LTPA and TRPA, respectively, and a modest association between number of intersections within one’s neighbourhood and the likelihood that they engaged in walking/cycling for transport-related activities. Taken together, this suggests that respondents in York Region living in areas similar to major urban centres that have a greater number of intersections (increased street connectivity) and a higher level of residential density were more likely to engage in PA. While neighborhood level covariates could not assessed, these analyses still demonstrate the need to consider variability between local neighborhoods when exploring the associations of built environment measures and PA. In particular, these analyses found that certain features of the built environment (i.e. residential density, intersections, and building area) were independently associated with PA behaviours; however, while no associations were found with the amount of green space within a 500 m buffer zone. Further research is necessary to clarify the extent to which modifications to such features of the built environment may improve physical activity participation in similar suburban communities.