School and class-level variations and patterns of physical activity: a multilevel analysis of Danish high school students

Physical inactivity is a major independent risk factor for the development of several lifestyle-related diseases and for global mortality [1‐4]. The burden of non-communicable diseases on populations’ health necessitates the identification and understanding of physical activity (PA) levels and trends, enabling the development and implementation of effective PA-promoting programmes. Global estimations of PA indicate that nearly a third of individuals, aged 15 years or older, do not comply with recommendations [5]. There is limited knowledge of PA patterns among young individuals (aged 15–25 years) compared to those among adults and children [5].

The high school period is a time of transition characterised by psychological, biological and social changes [6, 7]. Focus on and knowledge of the consequences of these changes is important, as the effect on health behaviours and patterns of declining PA trajectories, which persist into adulthood, might be established during adolescence and young adulthood [8‐10]. Gender differences in PA patterns are observed already during childhood, with girls being less active than boys, and inactivity patterns increasing with age [5, 8]. These findings are also observed in the Health Behaviour in School-aged Children (HBSC) study [8, 11] and in objectively measured PA in school children [12, 13]. However, the causes of these differences and to what extent they persist in high school students remains to be elucidated.

High school is an important context for the students, as they spend a large part of their day interacting with the school environment and with other students. It is uncertain to what extent the school influences PA behaviour in this age group, but the school might be an important determinant of the development of PA patterns. In children and adolescents, studies that have assessed the impact of school related factors on PA, have reported large variations in the PA that could be accounted for by schools [14‐21]. Except for two [18, 20], these studies were conducted among students attending schools primarily located in the same geographical area or municipality; thus, the impact of the school environment might not be fully reflected in these data. School-level variances might be higher when several municipalities are included, due to greater variations in environmental factors and health policies between schools.

Assessments of different measures of PA participation and the variation between schools and classes have, to our knowledge, not been conducted in a larger sample of high school students. Thus, the purpose of this study is to describe patterns of PA and assess variations between schools and classes in measures of PA, in a national cohort of Danish high school students.

This study described patterns of PA and quantified the variations of measures of PA between schools and classes among 70,674 students attending high school in Denmark.

Our findings in relation to PA patterns were consistent with findings in school children and adolescents showing that ethnic minority [20, 29, 30], female sex [5, 8, 12‐14, 19‐21, 30, 31], higher age [8, 19] and indicators of low parental socio-economic status [8, 13, 14, 20] were negatively associated with PA levels or the likelihood of participating in organised PA. Although a low proportion of students participated in PA at school during leisure time (13%), this finding is encouraging as girls of perceived ethnic minority simultaneously had higher odds of participating in PA at school. This finding could indicate that facilitation of opportunities of PA at the schools during leisure time could potentially activate this otherwise inactive group of girls. This is further emphasised by the finding that 9% of the variance in this outcome was accounted for by school characteristics. After the separate controlling for factors relating to school leadership, we found an attenuation of the ICC for PA during leisure time, suggesting that at least some of the identified school-level variance is due to facilities or opportunities offered by the schools. In support, secondary schools that offer daily PA and have high quality and accessible PA facilities have been found to be predictive for Canadian students’ time spent on moderate to vigorous PA [20, 21]. In this study, perceived ethnic minority, having parents with lower education, higher age and lower school urbanisation degree were independently associated with lower likelihoods of walking or cycling to and from school. This is consistent with active commuting studies in children and adolescents that have shown that ethnic minority [32], female sex [33‐36], indicators of low parental socio-economic status [32, 34], higher age [37], longer distance between school and home [33, 37, 38] and lower residential densities [36] negatively influence the use of active transportation to and from school. It is reasonable to assume that students attending schools located in intermediate or thinly populated areas generally have to travel longer distances to school, which explains why fewer students actively commute to school. The emerging pattern of students of a higher age being less likely to actively commute is discouraging, but is presumed to be indicative of students acquiring their driving licence (possible when turning 18 years old) and thus, using passive forms of transport. We found that 8% of the variance in active transport could be explained by school factors, implying that schools have the potential to influence students’ behavioural patterns in relation to active travel. School promotion of active transportation in Canada has been positively associated with higher PA levels in children [19]. Thus, by encouraging and facilitating initiatives of walking and cycling to school in this group of young individuals, the school has the potential to positively influence students’ modes of active travel.

High school-level ICCs indicate that student PA behaviour is influenced by mutually shared social and environmental circumstances at the schools. To our knowledge only Kristensen et al. (2013) has reported ICCs for the separate and combined effect of both the school and the class in children. They showed that classes had a consistently larger impact on PA (accounting for 6–14%), emphasising the importance of psychosocial factors rather than physical conditions of the schools. In line with this, consistently higher ICCs for school classes were observed in most of our PA measures, except for PA at school during leisure time and active transportation where school-level ICCs were higher. This is not that surprising, since students attending the same school are exposed to the same conditions and offered the same possibilities of participating in PA during leisure time at the school regardless of class relations. Like suggested by Kristensen et al. (2013), the higher class ICCs could indicate that circumstances within a class to a large extent influence PA patterns. However, it remains to be elucidated which are the specific class-related psychosocial factors that cause the clustering of these PA behaviours in this age group.

Compared to other health-related behaviours [14, 15, 18] school-level ICCs for PA behaviours are inconsistent and tend to be less than 0.07 [14‐21], suggesting little influence on PA behaviour by school characteristics. Although some of our school-level ICCs were low, they were in fact significant for all PA measures, and comparable to that observed in a nationally representative study in secondary school students in England [18] and Canada [21]. The between school variance of 9% and 8% for PA at school during leisure time and active transportation are noteworthy and suggest that schools can influence these behaviours. However, the small between school variance for inactive, very active, organised PA and self-organised PA, suggests that these outcomes are mainly individually determined and that only a small part of these PA characteristics is shared among students at the same school. This is perhaps not surprising as the outcomes address leisure time PA, which might not noticeably be influenced by school factors. Also, the low ICCs could potentially be explained by the fact that high schools in Denmark are largely governmentally administered, ensuring a certain degree of homogeneity. The student composition at the schools and the PA opportunities that they offer are probably, to a large degree quite similar.

A major strength of this study is the sample size and number of schools and classes included, which allowed us to assess separate associations for boys and girls and clustering effects for several measures of PA behaviour. With 70,674 students attending 119 different schools and 3213 school classes, this is to our knowledge the largest survey in relation to young individual’s PA patterns in Denmark and one of the largest globally. In 2014 there were 106,268 students enrolled in Danish high schools [39], of which about 67% were included in this study. Thus, the study population is assumed to represent students in Danish high schools, as student (85%), school (87%) and class (96%) participation were high. Not only is this study nationally representative, the extensive dataset, containing a number of PA questions, permitted detailed description of several patterns of PA participation in this age group.

Some limitations should be noted. Data are based on self-reported questionnaires, which tend to overestimate PA [40, 41]. However, we tried to minimise the overestimation in the outcomes by coding PA variables conservatively and by describing patterns of PA participation instead of levels or intensities of PA. The study is also limited by its cross-sectional design, as conclusions on cause-effect relations cannot be made. Longitudinal investigations are required in assessing the development of PA patterns over time and whether changes in school environmental factors precede change in PA in high school students.

Residual confounding cannot be dismissed. Few studies are available on school-level clustering effects in PA addressing specific factors that could account for the data clustering. Student demographics such as ethnicity, socio-economics and gender have shown to reduce school-level ICCs [14, 16, 18]. PA clustering effects have also proven not to be influenced by adjustments for gender, socio-economic status, parental exercise habits, grade, year, and season of measurement [17]. Likewise, we did not find that ICCs decreased in size for our outcomes when adjusting for socio-demographic variables, which suggest that the observed school-level variance is related to unidentified factors that are beyond the scope of this paper. It is possible that school-level effects in high school may reflect residual school effects from primary school, since students often transition together from primary to high school.

Our results contribute new and important knowledge about PA patterns and the impact of schools in young individuals attending high school in Denmark. The magnitude of the school-level variance of PA measures and how it is influenced by the inclusion of school characteristics is important for the development of school-based interventions aimed at increasing participation in PA and decreasing inactive behavioural patterns in young individuals. Girls of ethnic minorities, students with lower parental education and older students represent vulnerable groups with lower participation in several PA forms. Identification and circumventions of barriers complicating the participation in PA should be further identified in these groups.

This study identified girls of perceived ethnic minority, students whose parents exhibit lower levels of education and older students as generally less likely to participate in several forms of PA during leisure time. However, the described groups did also have higher odds of participating in PA at school during leisure time. Underlined by the noteworthy school-level ICC of 9% for PA at school during leisure time, this implies that schools have the potential to activate these otherwise inactive groups of students by facilitating opportunities of PA at the schools. Also, school factors explained 8% of the variance in active transport to and from school, implying that schools have the potential to influence students’ behaviour in relation to active travel. Thus, high schools should facilitate and encourage activities such as leisure time PA within the framework of school and active transportation in all ages and vulnerable groups, as regular active commuting is an important contributor to overall PA levels [42, 43].