The key findings of our study are as follows. First, by using a focused measure of where students eat their lunch, we were able to demonstrate that the food retail environment surrounding schools is strongly related to student’s eating behaviours during the school day. Second, our findings suggest that the geographic boundaries used to assess the food retail environment are better captured using road network buffers rather than circular buffers.
At 26.3%, the amount of variation in students’ lunchtime eating behaviours attributable to school-level factors was noteworthy. Although there are currently no studies with a similar lunchtime eating behaviour outcome, a similar Canadian study of the school food environment and obesity by Leatherdale et al. [34
] had an ICC of 5.4%. Our comparatively higher ICC value indicates that the school food environment accounts for a notable proportion of the variation in students’ lunchtime eating behaviours. Furthermore, the relationship between food retailers surrounding schools and students’ eating behaviors observed in our study were much stronger than those previously reported [17
]. This difference may partly be explained by our use of a precise measurement of food consumption at food retailers during the school day, rather than a more general measure of food consumption patterns reported for the entire day or week used in past studies [17
]. It is important to consider the specific context of food consumption, including where and when the food was eaten, in order to evaluate the importance of a specific food environment. By accounting for the particular context in which food was consumed, the potential for the misclassification in the measurement of the food environment is greatly reduced.
Our comparison of the model fit provided by circular and road network-based buffers provided findings consistent with similar research, despite the fact that previous research has used different behavioural outcomes in varying populations. Within a sample of adults residing in Vancouver, B.C. Oliver et al. [20
] examined the relationship between land use mix in the 1
km surrounding each participant’s home, measured using 1
km circular and road network buffers, and their walking behaviour. The road network-based measures were consistently related to walking behaviours while the circular-based measures were not. Taken together, the findings from these two studies suggest that researchers should consider investing the time in obtaining road network buffers when they want to measure the association between built environment constructs and health-related behaviours. However, it is important to note the scarcity of studies directly comparing buffer types. Future studies comparing these measures are needed to confirm this observation, particularly when assessing the food environment.
Despite the implementation of a new policy restricting fast food restaurants in a socioeconomically disadvantaged area in California [35
], there is currently no evidence evaluating its effectiveness. In fact, to our knowledge, no existing studies have examined whether policies aimed at restricting the number of food retailers that sell primarily unhealthy foods (e.g., fast food restaurants) in a given region impacts people’s dietary behaviours. However, there is analogous evidence from interventions and policies put in place to address the lack of supermarkets – the main source of reasonably priced fresh fruits and vegetables – in socioeconomically disadvantaged neighbourhoods. For instance, the introduction of a new supermarket in a deprived neighborhood in Leeds, UK positively impacted the fruit and vegetable consumption of the adults residing in that neighborhood, particularly those with the worst diets, whose fresh fruit and vegetable consumption doubled [36
]. This demonstrates the potential to influence peoples’ eating behaviours by modifying their food environment through policies and interventions. Given the preponderance of unhealthful food retailers near schools [15
] and the strong associations we found with students’ eating behaviours, there is a need for future research to evaluate whether restrictions on food retailers near schools affect lunchtime eating behaviours of young people.
While the food retail environment within the school has an important impact on student’s eating behaviors and food choices [12
], approximately one third of the grade 9 and 10 students in our national study did not usually eat their lunch at school and almost one in ten usually ate their lunch at a food retailer. Interestingly, we observed that the presence of school cafeterias and certain vending machines were positively associated, albeit not statistically significant, with eating lunch at food retailers. The positive relationships suggest that despite having the option to purchase food directly within their school, some students prefer to purchase their lunch at nearby food retailers. Furthermore, the population attributable risk calculations suggest that 58% of the study outcome (eating at food retailers during the school week) was attributable to students being exposed to 3 or more food retailers within a 1
km travel distance of their school. Therefore, policies and programs directed at eliminating unhealthy food choices within school cafeterias and vending machines may be undermined by the availability of less nutritious food at nearby food retailers.
If additional studies provide evidence of a strong and consistent relationship between the food environment surrounding schools and students’ eating behaviours, municipal planners should consider implementing policies that would limit the number of food retailers in close proximity to schools. A second strategy to limit students’ consumption of food from nearby food retailers would be the implementation of policies preventing students from leaving school grounds during the school day. However, this would not be effective in preventing students from purchasing food at nearby food retailers before or after school.
Some important strengths and limitations merit consideration. Key strengths of this study included our specific measurement of eating behaviours during the school day and the large and geographically diverse study sample. A limitation of this study was that the food environment was measured with an online GIS database, which may not provide a completely accurate and up-to-date measure of the food environment. Furthermore, only the top 75% of chain fast food and donut/coffee retailers were included in the exposure measure. There was no information from the HBSC survey on which food retailers the students actually went to. Also, all data were obtained by self-report, and may be subject to bias introduced by the social desirability of eating healthy meals. Furthermore, the HBSC survey did not collect information on food preparation practices at home which may influence whether students brought their lunch from home or purchased their lunch from a nearby food retailer. Finally, the study was cross-sectional and therefore temporality between the presence of food retailers and eating behaviours cannot be directly established. However, it is unlikely that students chose to attend schools based on the presence of nearby food retailers.