Participation in organized leisure-time activities and risk behaviors in Czech adolescents

We used data from the Health Behaviour in School-Aged Children (HBSC) study conducted between April and June 2014 in the Czech Republic. Schools were the primary sampling unit. They were selected randomly from the database of the Czech Ministry of Education, Youth and Sports, after being stratified by region and type of school (primary vs. secondary schools). Out of 244 contacted schools, 243 granted consent (response rate 99.6%) to carry out the survey. One class per 5th, 7th, and 9th grade, which, in general, correspond to the age categories of 11, 13, and 15 years in the Czech Republic, was then selected at random at each of the participating schools. Questionnaires were administered during regular class time by trained research assistants in the absence of a teacher. Participation in the study was voluntary and anonymous, with no incentives offered to the participants. Parents or legal guardians of the adolescents were notified of the study and its purpose by the school management in advance and could withdraw their child if they disagreed. Prior to administration of the questionnaires, respondents were also given the opportunity to opt out of the study or skip questions that made them uncomfortable. The study design was approved by the Ethics Committee of the Faculty of Physical Culture, Palacky University, Olomouc (No. 17/2013).

We obtained data from 14,539 respondents (response rate 89.2%). Approximately, 10% of the children were not present at school during the survey. Thirty children refused to fill in the questionnaire (0.2%). First, in line with the HBSC protocol, we selected only adolescents aged 11, 13, and 15 years (n = 10,795). Next, we excluded 516 cases due to missing data on all OLTA items (n = 287), more than half (i.e. 5 or more) of risk behaviors (n = 4), contradictory responses on reported prevalence of risk behaviors in lifetime versus the last 30 days (n = 170), or other unlikely responses throughout the questionnaire (n = 55). For instance, a respondent who indicated drinking alcohol (get drunk or smoke cigarettes) more than three times in the past month but never during his or her lifetime was excluded from the analyses. The final sample comprised 10,279 adolescents.

Participation in OLTA was assessed using the question: In your free time, do you do any of these organized activities? which was proven to have a good reliability (Bosakova et al. 2016). We investigated participation in the following six activities: team sports, individual sports, art school, youth organizations, recreation/leisure centers or after-school clubs, and church meeting/singing. Missing answers were considered to represent a ‘no’ unless all six OLTA items were missing. Then, the respondent was excluded (n = 287). First, the OLTA participation was investigated using a dichotomous variable (participation in at least one OLTA vs. in no OLTA). Next, the respondents were split into five groups of OLTA participation patterns (Table 1) based on a cluster analysis (Badura et al. 2015).

As indicators of risk behaviors, we selected eight distinct phenomena with the following cut-off points: current smoking (at least once a week), alcohol consumption (at least 3 days in the last 30 days), drunkenness (at least twice in the last 30 days), injuries (at least twice in the past 12 months), physical fighting (at least twice in the past 12 months), truancy (at least once in the past 12 months), and bullying others—perpetration and being bullied—victimization (at least twice a month). Using the last two items, as Solberg et al. (2007) recommended to distinguish between bullies and bully-victims (i.e., those being both bully perpetrators and victims), we derived a categorical variable of youth involved in bullying others (bully and bully-victim vs. those not involved in bullying others). Cut-off points regarding substance use and fighting were set to capture only more severe (recurrent, i.e., more than once) forms and to avoid undesired inclusion of their occasional occurrence. Items were part of the HBSC mandatory questionnaire, except for truancy. This was assessed as ‘During the past 12 months, did you skip school without a proper excuse for at least a whole day?’ The detailed description and scientific rationale of the other items can be found in the HBSC International Protocol (Currie et al. 2014).

Socioeconomic status of adolescents’ families, as a confounder, was measured by the Family Affluence Scale (FAS) developed for the purposes of the HBSC study (Currie et al. 2014). The responses on six items exploring various indicators of family socioeconomic status (car ownership, holidays abroad, having one’s own bedroom, number of computers in the household, number of bathrooms, and dishwasher ownership) were summed up and transformed to a fractional rank score ranging from 0 to 1. The score was then trichotomized to classify respondents into groups of low (0–0.333), medium, (0.334–0.666), and high (0.667–1) socioeconomic status.

First, we described the composition of the sample and its involvement in various risk behaviors. Second, using binary logistic regressions, we assessed the associations of the dichotomized overall OLTA variable (at least one activity vs. none) with cigarette smoking, alcohol use, drunkenness, involvement in physical fights, truancy, and injuries. The association of OLTA participation with bullying perpetration was analyzed using multinomial logistic regression. In the first step, we assessed crude associations per risk behavior and bullying categories, respectively (Model 1). Next, we adjusted for age and gender (Model 2), and then also for FAS (Model 3). At last, we tested interaction effects of gender (Model 4) and age (Model 5) on these associations. Then, using a two-step cluster analysis, we derived five distinct clusters of adolescents based on their pattern of OLTA participation. This number of clusters was the smallest possible that led to a reasonably high homogeneity within the clusters and reasonably high differences between particular clusters, shown by an average silhouette width over the value of 0.5 (Badura et al. 2015). All the regression analyses were then run again for each separate cluster of OLTA as an independent variable. The data were analyzed using ordinary single-level regression, because multilevel analyses yielded no indication of clustering by school. Prior to conducting the regression analyses, we assessed the random variance of specific risk behaviors at the level of schools, and for none of them, it was statistically significant. The statistical analyses were carried out using IBM SPSS 22 for Windows (IBM Corp. Released 2013) and MLwiN Version 2.02 (Centre for Multilevel Modelling, University of Bristol).

The large and representative sample is the most important strength of this study. Furthermore, it was based on the well-established and recognized HBSC study, with a strong methodological background regarding data collection procedures and construction of the questionnaire, which is being developed by an international expert team on a continuous basis.

This study also has some limitations. First, due to its cross-sectional design, it is impossible to determine causality of the associations. Second, the use of self-reported data is, in general, more susceptible to being biased. However, the HBSC mandatory items on substance use, injuries, and violent behaviors have been shown to have validity and reliability (Currie et al. 2014). Third, we lacked more detailed information on specific types of OLTA, as it has been proven that, e.g., there is a difference between American football and baseball in aggressive behaviors (Kreager 2007), while our measures did not allow capturing such differences. Finally, we also did not have data on other dimensions of OLTA, such as intensity, duration, and engagement, which might play unique roles and could have affected the associations found.

In general, the findings of this study demonstrate the associations between OLTA participation and lower occurrence of regular substance use and truancy. This indicates that conclusions drawn previously in the US context might be applicable to the European context as well. The structured content of OLTA could, therefore, serve as one of the options to address these issues, especially in girls, who seem to profit more from their participation in terms of lower occurrence of risk behaviors. In particular, art activities, either alone or in combination with sports, appear to be the most beneficial OLTA context. Future research should concentrate on revealing the causal pathways between OLTA participation and risk behaviors.

OLTA participants are less likely to smoke regularly, get drunk repeatedly, and skip school. On the other hand, OLTA involvement is associated with higher risk of physical fighting and injuries. Girls, in general, seem to be at lower risk of risk behaviors when participating in OLTA than boys. However, the specific OLTA pattern is important, because we found no significant association with reduced occurrence of risk behaviors among adolescents involved only in team sports.