The negative effect of sedentary lifestyles on children's health is a source of concern (1); the increasing prevalence of obesity among North American youths coincides with an increasing prevalence of high screen time (typically defined as any combination of activities such as watching television or playing video games) in this age group (2). National organizations have developed recommendations to limit sedentary behavior among youths (3). For instance, the American Academy of Pediatrics recommends that children's screen time be limited to no more than 1 to 2 hours of quality programming per day (3). Few children or adolescents meet these guidelines (4), and activities designed to reduce sedentary behavior in this age group should be a public health priority.

Defining sedentary behavior in terms of the absence of physical activity (PA) (5) fails to acknowledge the range and complexity of sedentary pursuits that can be modified. For example, research has focused on the relationship of obesity and PA to television watching (6) and more recently to other types of screen-time sedentary behavior, such as playing video games and using a computer (7). Research suggests that most children spend 1 to 3 hours in these types of screen-time sedentary behaviors per day (1).

Social cognitive theory (8) posits that sedentary behavior is influenced by personal beliefs (believing parents do not encourage PA), physical characteristics (being overweight), and other related behaviors (frequency of regular participation in PA). Empirical research has demonstrated support for these relationships with respect to screen time (9,10).

An ecologic approach to youth inactivity acknowledges that sedentary behaviors are a function of not only individual characteristics but also context (11). The school environment may be associated with the time youths spend in screen-time sedentary behavior (11). We still know very little about the associations between the school environment and sedentary behavior, except that 1) youths spend approximately 25 hours per week in school throughout the school year, 2) school-based interventions to reduce sedentary behavior are likely more effective than individual-based interventions (12), 3) the most prevalent behavior among youths after school is screen-time sedentary behavior (13), and 4) other behaviors, such as smoking (14) and PA (15) vary among elementary schools. We sought to better understand through multilevel analyses the school- and student-level characteristics associated with screen-time sedentary behavior among children in grades 5 through 8. Such insight would be valuable to appropriately tailor school-based interventions to reduce sedentary behavior.

The study sample was approximately evenly divided between boys and girls (Table 1), and average age was approximately 12 years. Most students reported 1 to 3 hours of screen time per day; boys were more likely than girls to report 3 or more hours. More boys than girls were highly active. BMI could not be calculated for 47% (n = 1,149) of the sample because of missing height or weight measurements. Among respondents who provided sufficient data to calculate BMI, the mean BMI for boys was 19.5 kg/m² (standard deviation [SD], 3.8 kg/m²) and for girls was 19.1 kg/m² (SD, 4.1 kg/m²). Boys were more likely than girls to be overweight or obese, although girls were more likely to have missing BMI data. Most students were white, and significantly more boys than girls were white.

Most schools were in the action phase for healthy physical environment and supportive social environment and the maintenance phase for community partnerships (Table 2). Conversely, most schools were in the initiation phase for instruction and programs. None of the schools was in the maintenance phase for the overall scores for healthy physical environment, instruction and programs, or supportive social environment.

We identified no significant between-school random variation for moderate screen time compared with low screen time. Because the school a student attended was not related to his or her likelihood of reporting moderate screen time, school characteristics were not examined in this model. The student-level characteristics associated with moderate screen time included sex, number of friends who are active, and parental encouragement of PA (Table 3).

We identified significant between-school random variation for high screen time compared with low screen time. Our analysis suggests that school-level differences accounted for 12% of the variability in the odds of a student reporting high screen time. Univariate analyses revealed that 2 school characteristics were associated with the odds of reporting high screen time; however, in the multivariate analyses, only 1 school characteristic remained significant. If a student attended a school that was in the action phase for the indicator "Emphasis placed on maximizing participation in PA through school programs," he or she was less likely to report high screen time than a similar student attending a school that was in the initiation phase for this indicator (Table 3). The student-level characteristics associated with high screen time included sex, parental encouragement of PA, parental support of PA, and race/ethnicity. No significant contextual interactions between school- and student-level characteristics were found.

The students in our sample were highly involved in screen-time sedentary behavior, and our findings provide further support for the recommendation that intervention efforts to reduce sedentary behavior must begin before adolescence (23). Two student-level characteristics were consistently related to screen time. First, boys were more likely than girls to report high screen time. Similar patterns have been found elsewhere (1,11). Second, consistent with research among secondary school students (9,20), we found that parental support and encouragement for PA are associated with screen time among elementary school students. Behavioral theories consistently highlight the important role that influential social models surrounding youths (eg, parents) can have on their behavior (8). In general, social models can influence behavior through modeling, social norms, or providing support and encouragement for the behavior (8). Both direct support (overt provision of assistance) or indirect support (encouragement and emotional support) may be associated with more PA among youths (24); a lack of television rules set by parents may also facilitate more screen time (25). Considering that youth become socialized to be active or inactive by the encouragement and support provided by their parents during the elementary school years (26,27), interventions designed to reduce screen time sedentary behavior should target elementary school children and be tailored to include the participation and involvement of parents whenever possible.

Notably, we found that even when controlling for individual student characteristics, the characteristics of the school a student attends were associated with high levels of screen time. Because youths spend up to 12 years of their life in school, the school environment is an ideal place to promote more active lifestyles in this population. To the best of our knowledge, this is the first study to find that the likelihood of a student reporting levels of screen time that exceed existing public health recommendations (3) was associated with the characteristics of the school he or she attended. Specifically, students were less likely to have high screen time if they attended a school that emphasized participating in PA through school programs. These findings are consistent with existing research (11) and guidelines that recommend schools provide opportunities for students to be active during and after school hours to reduce time spent in sedentary behaviors (12); these opportunities can consist of providing students with access to resources before, during, or after school or working with community partners to ensure students have access to resources and facilities.

School-based strategies designed to alter the school environment have a more pronounced effect on increasing PA levels than do individual approaches targeting knowledge and beliefs (28). For instance, schools can compensate for the increase in sedentary behavior as children get older by offering students the opportunity to participate in other activities such as varsity and intramural teams at school (20). Such activities may be particularly important after school because this is the time of day when sedentary behaviors commonly occur (13). Rewarding student behavior by providing additional supervised areas for children to play during the school day (29) or providing additional after-school programs (30) may promote active rather than sedentary choices during students' discretionary time. Additional research is required to evaluate the effect of such initiatives.

Our findings, together with previously published empirical research (14,31), suggest that program planners should target additional prevention resources to the schools that are putting students at the highest risk for being sedentary (ie, schools that do not encourage activity through school programs). By targeting these "high-risk" schools, intensive prevention programs could be implemented where they are most likely to influence students' behavior (31). It may also be beneficial to tailor programs to the needs of high-risk students. For instance, our finding that nonwhite race/ethnicity is associated with more sedentary behavior suggests that prevention programs may need to be tailored to the needs of nonwhite students.

Limitations to our study should be acknowledged. First, the SHES is designed to assess PA environments and policies. Accordingly, there is a lack of correspondence between the SHES and our outcome of interest, screen-time sedentary behavior. This may explain why school-level differences accounted for modest variability in the odds of a student reporting high screen time. Some school-level factors may have been related to sedentary behavior but not assessed by the SHES.

Nearly half of participants in the host study did not report height or weight, preventing us from calculating their BMI. Previous research on older adolescent populations has also reported large amounts of missing self-reported BMI data and that BMI data are more likely to be missing among younger respondents (32,33). Consistent with those studies, additional analyses performed with the PLAY-On data identified an age-related trend in which the prevalence of missing BMI data was higher for younger respondents (34). Therefore, it is possible that some of the results identified in our study may be biased (eg, the lack of associations between BMI and screen time); the results presented in relation to BMI should be interpreted with caution. The missing BMI values may be a result of motivated nonresponding, which would have important implications for the feasibility of using self-reported height and weight in future surveillance efforts.

Causal relationships can not be inferred from these cross-sectional data. Although data were based on self-reports, the measures in the student questionnaire have been previously demonstrated to be reliable and valid (16), and honest reporting was encouraged by ensuring confidentiality during data collection.

Because of the increasing prevalence of obesity among youth populations, a better understanding of sedentary behavior is necessary. We found that even when controlling for individual student characteristics, the characteristics of the school a student attends were associated with his or her likelihood of having high screen time. Future research should evaluate whether the optimal population impact for school-based activity promotion might be achieved most economically if interventions selectively target the schools that are putting students at the greatest risk for being sedentary.