Physical inactivity has an annual financial impact totalling ~$2.1 billion in Canada . As such, developing a better understanding of the factors that inhibit or promote PA among youth populations should be a public health priority. We identified that even when controlling for individual student characteristics, the characteristics of the school a student attends were associated with his/her likelihood of being either moderately active or highly active. This finding is consistent with previous empirical research which suggests that characteristics of the school environment can have an important impact on the PA level of a student [13–15]. We also identified that the school characteristics associated with PA were not the same for differentiating youth who were moderately active from youth who were considered highly active. This suggests that a targeted approach may be required for PA promotion depending on what youth sub-populations school-based programs may be trying to impact. For instance, our results suggest that while programs using PA as a reward may increase the likelihood of students being moderately active, schools may need to implement community partnership programs if they are interested in increasing the likelihood of students being highly active. Such a targeted approach to program implementation would require evaluation. Considering that there is more evidence from secondary school settings compared to elementary school settings , and there is even less evidence simultaneously examining how multiple school characteristics (programs, policies and resources) are associated with PA , our findings provide valuable new insight to both researchers and practitioners.
Consistent with research demonstrating that community coalitions can affect youth behaviour [7, 12, 17, 28, 29], we identified that students were more likely to be highly active if they attended a school that had established community partnerships. This includes partnering with public health units, partnering with community-based recreation clubs and organizations, and providing staff with ongoing training and support . Given the importance of community-based support and reinforcement on establishing effective school initiatives , it was promising to see that over three quarters of the schools in our Ontario sample were in the action or maintenance phase for developing community partnerships. This was substantially higher than the published results from the American Trial of Activity for Adolescent Girls study, where just over a third of schools had collaborated with groups in the community to provide students with PA programs . The Action Schools! BC program provides a good model for understanding the mechanisms by which schools can collaborate with community stakeholders to promote PA among youth . A survey that employed the same SHAPES school-level tool found that schools with a "healthy school committee", often including community members, was more likely to achieve a maintenance classification for offering students a healthy school environment (unpublished data).
In this study, we also identified that some students were more likely to be highly active as a function of both their individual behaviour and whether or not the school they attended had established community partnerships. As illustrated in Figure 1, although students who participate in league sports outside of school were more likely to be highly active than students who do not participate in league sports outside of school, the strength of the association appears to be moderated by community partnerships. Attending a school that was in the action or maintenance phase for community partnerships was associated with a substantially larger likelihood of being highly active for students participating in league sports and a modest increase in the likelihood of being highly active for students who do not participate in league sports outside of school relative to students who attend a school in the initiation phase for community partnerships. Considering the physical activity levels of our respondents were similar to those of a large sample of secondary school students in Ontario  and data from adolescents in the United States , this is an important finding for practitioners interested in tailoring and/or targeting PA promotion programs to consider. For instance, there may be a larger impact by targeting programs designed to enhance community partnerships to schools in need rather than tailoring programs to all schools.
To the best of our knowledge, this is the first study to identify that students were more likely to be moderately active if they attended a school that used PA as a reward and not as discipline. It makes sense that if encouraging physical activity is the goal then physical activity experiences need to be made as positive and reinforcing as possible as opposed to allowing an association between physical activity and pain or punishment to be established . Additional research is required to evaluate the potential mechanisms for using PA as a reward and the impact of such novel interventions on student PA. For instance, rewarding student behaviour by providing additional supervised areas for kids to play during the school day  or providing additional after school programs  may promote active choices in students' discretionary time.
Sedentary behaviours, such as screen time, are distinct from PA and do not necessarily replace time spent being active [8, 33]. This distinction is important as research suggests that the largest public health benefit with respect to PA promotion will come from having sedentary individuals become more active rather than having active individuals become more active . The American Academy of Pediatrics recommends that children's total screen time be limited to no more than 1 to 2 hours of quality programming per day . In alignment with these recommendations, we identified that students with three or more hours of screen time per day were less likely to be either moderately active or highly active. However, research reviews have previously concluded that there was a zero to small association between television-based screen time and PA among youth . Discrepancies between our findings and their conclusions may be due to our inclusion of other sedentary screen time behaviours, such as computer use. A recent study of sedentary behaviours in Canadian adolescents reported that computer usage was associated with physical activity among males, and reading was associated with physical activity among females . We recommend that future research consider the relationship between multiple screen time behaviours rather than focusing exclusively on television/video use.
Unlike research from a provincial survey of key school informants at the elementary  and secondary school levels  in Ontario, and research from the US NHANES III  which identified that the majority of students report that they do not play on school based sports teams, we identified that the majority of the students in our elementary school sample reported participating in sports teams at school. This is important considering that in the present study those students who participated in intramural sports at school were almost twice as likely to be considered moderately active and three times more likely to be highly active. This finding is consistent with previous research [5, 6] and research highlighting that one of the most preferable methods for engaging in PA among youth is via playing sports . Alleviating barriers to the provision of intramural activities may represent an ideal opportunity for schools to intervene . For instance, some schools provide students with activity buses that allow them to participate in intramural sports after schools hours . Research is required to evaluate if developing programs or policies to promote student participation in intramural sports at school has an impact on increasing PA levels among students.
Behavioural theories consistently highlight the important role that influential social models surrounding youth (e.g., friends) can have on their behaviour [40, 41]. In general, social models can influence behaviour through modelling, through social norms, or through providing support for the behaviour [40, 41]. Empirical research has also demonstrated that the behaviour of peers are associated with higher levels of physical activity among youth [5, 6] and friends' influence on physical activity levels may be higher than parental influence at least for adolescents . Research has rarely considered peers as a target for PA intervention studies and this issue deserves attention in promoting a school climate that values PA.
This study is subject to some limitations. Almost 50% of the data for BMI were missing, so we could not robustly understand the association between weight status and PA in this sample. Since no data on ethnicity or socioeconomic status are available within our measurement tools, it was not possible to examine how PA varied across ethnic groups or social economic strata. Our ecological data were from the school environment, and it is possible that characteristics from other ecological contexts (e.g., home) may also be important to consider. Causal relationships can not be inferred from these cross-sectional data. Considering that these data were drawn from a convenience sample of schools, we can not infer that these results would be representative of the general student population in Ontario. Although data were based on self-reports, the measures in the PAM have been previously demonstrated to be reliable and valid , and honest reporting was encouraged by ensuring confidentiality during data collection. However, by using a measure of physical activity based on energy expenditure, we have not provided information regarding the frequency, duration or intensity of physical activity which may also be important details relevant to practitioners .