Study design
Data were obtained from the WHO collaborative HBSC study conducted in 2002–2014. HBSC is an international survey on adolescent (11-, 13-, and 15 year olds) health and the context of health, conducted every 4 years since 1983 across European and North American countries (Inchley et al. 2016). Sampling procedure and questionnaire are based on a standardized research protocol (Currie et al. 2014). Each participating country obtained approval to conduct the survey from their ethics review board or equivalent regulatory institution. Data were collected through a school-based survey, using an anonymous self-completion questionnaire. Under supervision of the national research teams, a translation/back-translation procedure was applied to guarantee language equivalence of the questionnaires. Representative samples on each national or regional level were selected using a clustered sampling design where the initial sampling unit was either the class or the school.
Participants
Thirty HBSC countries collected data on EMC, FTF and substance use in 2002, 2006, and 2010. Separate studies were carried out in different parts of Belgium (Flanders and Wallonia), and the UK (England, Scotland and Wales). The data for these countries were analysed at the national level. Danish data were based on mainland Denmark (data from Greenland were excluded from the analyses). In total, the analyses for tobacco and alcohol use were based on N = 445,827 adolescents (51% girls, Mage = 13.5 years old) across 26 countries. As cannabis use was assessed in 25 countries and only among 15 year olds, analyses on cannabis use were based on N = 137,398.
In 2014, 42 HBSC countries participated in the HBSC study. Slovakia and Lithuania did not have (reliable) data on EMC; Spain and Ireland lacked data on FTF. These countries were removed from the analysis. This resulted in a sample of N=191,727 adolescents in 34 countries. The mean age of adolescents was 13.5 years old (ranging from 13.1 in Armenia to 13.8 in Canada). 51% were girls (ranging from 44% in Russia to 53% in Austria and Denmark). As cannabis use was only measured among 15 year olds and two countries (Norway and Greece) did not assess cannabis use, this resulted in N = 56,159 in 32 countries for the analyses on cannabis use.
Measures
Substance use Across all waves, substance use was assessed as follows. Weekly alcohol use. Students were asked how often they drank beer, wine, and liquor/spirits. For each type, response options were “1 = never”, ”2 = rarely”, “3 = every month”, “4 = every week”, “5 = every day”. This variable was dichotomized by combining options 1 to 3 (indicating less than weekly alcohol use, coded as 0) and 4 to 5 (to reflect at least weekly drinking, coded as 1). Weekly smoking. Smoking status was defined based on the question “How often do you smoke tobacco at present?” Original answer categories (never, less than weekly, weekly but not daily, daily) were recoded into weekly smoking (1) and less than weekly smoking (0). Lifetime cannabis use. Students reported the frequency of cannabis use in their lifetime on a scale from 1 to 7, with 1 = never and 7 = 40 times or more. Answers were recoded into 0 = never and 1 = at least once.
Electronic media communication (EMC) In 2002, 2006 and 2010, electronic media communication was measured by the item “How often do you talk to your friend(s) on the phone or send them text messages or have contact through the internet?” Response options were “never or rarely”, “1 or 2 days a week”, “3 or 4 days a week”, “5 or 6 days a week”, and “every day”. Answers were recoded into 0 = less than daily and 1 = daily. In 2014, electronic media communication was measured by means of three items: How often do you a) contact your friends using texting/SMS?; b) actively contact your friends using instant messaging (e.g. BBM, Facebook chat)?; c) contact your friends using other social media, such as Facebook (posting on wall, not chat), My Space, Twitter, apps (e.g. Instagram), games (e.g. Xbox), YouTube? Answer categories were “hardly ever or never”, “less than weekly”, “weekly”, “daily”. Answers were recoded into 0 = less than daily and 1 = daily.
Face-to-face (FTF) contact with peers in the evening was measured by the item “How many evenings per week do you usually spend out with your friends after 8 PM?” in 2002, 2006 and 2010. The response options ranged from zero to seven evenings a week. Answers were recoded into 0 = less than daily and 1 = daily. In 2014, the question was changed into “How often do you meet your friends outside school time after 8 PM?” Answer categories were hardly ever or never, less than weekly, weekly, and daily. As prevalence rates in the first categories were relatively low, we combined them into the category “less than once a week”.
Confounders In our analyses, we controlled for age, gender and family affluence. Family affluence was measured by the ridit-based relative HBSC Family Affluence Scale (Torsheim et al. 2016).
Statistical analyses
Analyses were performed on two distinct datasets, as EMC was assessed differently in 2014, compared to the other years (2002, 2006, 2010). Using the trend data (2002–2006–2010), we assessed descriptive statistics of trends over time in adolescent EMC, FTF, and substance use (Table 1). Correlations were calculated between two (absolute) difference scores (2002–2006 and 2002–2010) of EMC, FTF and adolescent substance use across countries. Next, hierarchical multilevel analyses (with the second-level variable being country) were performed to test the association of daily EMC with three indicators of substance use (tobacco, alcohol and cannabis; Table 2). Analyses were controlled for age (tobacco and alcohol), gender, and family affluence (tobacco). In Model 1, survey year was added to the model, followed by daily EMC (Model 2), daily FTF (Model 3), and the interaction of survey year × EMC (Model 4).
Table 1 Adolescent face-to-face contact with peers in the evening, electronic media communication and substance use in 26 countries, 2002–2010 (%) Table 2 Results of multilevel logistic regression analyses predicting adolescent substance use in 26 countries, 2002–2010, odds ratios We also provide descriptive statistics of adolescent EMC, FTF, and substance use, by country, in 2014 (Table 3). Also here, hierarchical multilevel logistic regression models (with the second-level variable being country) were applied (Table 4). Analyses were controlled for age (tobacco and alcohol), gender, and family affluence (tobacco). In Model 1, daily EMC was added to the model, followed by FTF (Model 2), and the country-level mean prevalence of daily EMC (Model 3). Lastly, we added a cross-level interaction between daily EMC and the aggregated (average) EMC use per country (Model 4). Here, we divided countries into tertiles based on the percentage of youth that engaged in EMC at a daily basis. All analyses were performed using Stata SE 12.1 (College Station, Texas, USA).
Table 3 Adolescent face-to-face contact with peers in the evening, electronic media communication and substance use in 34 countries in 2014, % Table 4 Results of multilevel logistic regression analyses predicting adolescent substance use in 34 countries in 2014, odds ratios