Effects on alcohol use of a Swedish school-based prevention program for early adolescents: a longitudinal study

Internationally, mean alcohol consumption among adolescents varies substantially among countries. The ESPAD study [29] found that among 16-year-old adolescents, girls’ alcohol consumption ranged from 21 centiliters of 100% alcohol per year in Romania to 137 centiliters on the Isle of Man. For boys it ranged from 71 centiliters per year in Ukraine to 173 centiliters on the Isle of Man. Boys generally report higher annual alcohol consumption than girls in all countries except Iceland and Norway. When “heavy episodic drinkers” (five or more drinks in a row during the last 30 days) are considered, the frequency for girls varied from 1.5 times in the last 30 days in Romania to 3.7 times on the Isle of Man. The corresponding figures for boys were 2.2 (Iceland, Romania and Switzerland) and 3.7 (Estonia and Malta) times in the last 30 days. Overall, boys reported a significantly higher frequency of heavy of drinking than did girls in all countries except Sweden and Iceland.

Swedish law, as in many countries, requires a person to be 18 years old to purchase alcoholic beverages containing more than 2.25% alcohol by volume. At the state-monopoly alcohol stores, the age limit for purchasing alcoholic beverages containing over 3.5% alcohol by volume is 20 years [30]. The national trend in Sweden is toward decreasing alcohol consumption among adolescents, however this could be explained by a polarization of drinking [31, 32]. Currently, data show that 47% of 15-year-olds (43% of boys and 50% of girls) in Sweden consume alcohol [33].

Substance use prevention programs in schools were initially characterized by short term interventions and fear arousal, lecturing students about the dangers and long-term health consequences of substance consumption [34, 35]. Such interventions were rarely successful and generally lacked a theoretical foundation that considered developmental factors, social influences, and other potential contributors to adolescent substance consumption. Today, we know more about which components are successful in interventions, but few of the many extant programs include these features or evaluate their effects with well-designed research [36]. From a learning perspective it has been argued that interactive strategies, combining practice in social skills and educational knowledge transmission, increase the likelihood of success in a prevention program [24, 25, 37, 38]. Interventions that target more than one risk factor also seem to be more effective [34, 39]. Methods for strengthening social, emotional, behavioral, cognitive, and moral competencies; building self-efficacy; improving social relations with adults, peers, and younger children; providing structure and consistency in program delivery; and intervening with youth for a longer period of time are further effective components [40‐42]. There is no evidence that resistance skills training has an effect on reducing substance use [36]. There is also no evidence that more intensive programs are more effective than less intensive ones [36]. Reviews have also tried to elucidate the effectiveness of school, family, and multicomponent programs, with some ambiguous results. While a Cochrane review [21] reported little evidence of the effectiveness of multicomponent interventions, compared with single-component interventions, in reducing alcohol misuse among youths, e.g., Catalano et al. [40]. Another program, originally from Finland but used in many European countries, targets pupils directly in class and uses competition as motivation. This is known as the “Smoke-free class competition” (SFC). The SFC targets only tobacco use and has been evaluated positively in a meta-analysis [43]. The core of the SFC is that entire school classes compete by signing a contract in which they promise to be smoke-free for the next 6 months. If they succeed they can win prizes in a lottery. The methodological approach of using prizes as rewards relies on frameworks such as learning theory [44] and social learning theory [45], where the behavior of remaining smoke-free becomes attractive and worthwhile; also, classmates are important to adolescents and function as influential role models for smoking initiation. Incentives, such as prizes or rewards, may be effective in improving single health behaviors, as shown by J Kavanagh, A Oakley, A Harden, A Trouton and C Powell [46]. However, this has not been the case for complex health behaviors.

Although Swedish schools have adopted a range of different prevention programs, published evaluations on programs targeting the pupils directly, in class, are scarce. Programs shown to be effective internationally lack counterparts in Sweden [47, 48]. One exception is the “Unplugged” program, which aims to postpone substance use and reduce future use. The program was implemented in 14 Swedish schools and in another set of schools in seven European countries. The program employs a high degree of interactive pedagogy, e.g., by involving pupils as leaders, providing education in life skills, holding parent meetings, and training social skills and norms related to alcohol and narcotics [47, 49]. On a European level, the results showed that the program was more effective than traditional substance use education. However, in Sweden there were no significant differences between control and treatment schools. It was argued that the program’s lack of measurable effect may be due to the fact that many of the ordinary prevention strategies implemented in control schools offer a high degree of interactivity similar to that in the Unplugged program [49].

The Triad consists partly of components that have been shown to be successful but also includes some weaknesses (described under Structure and Components of the Program). Our investigation will add to the body of evidence related to school-based intervention programs. To our knowledge, this is the first study to evaluate a Swedish program targeting this age group during school hours. This study focuses on one of the three components of the Triad program – preventing substance use (“Fighting Drugs”) – by following pupils from school year 6 (baseline) to school year 9.

The intervention the Triad is administered by the non-governmental organization (NGO) Team 49. The NGO has no religious or temperance-related affiliation. The purpose of the Triad is to increase pupils’ resistance skills through understanding and awareness of the harmful consequences of drugs and delinquent behavior, which in turn is believed to influence them to refrain from drugs and crime. The entire class participates and constitutes a team in arranged competitions with prizes. The program is built on different themes for each grade: Grade 4 (10- and 11-year-olds) includes traffic education; grade 5 (11- and 12-year-olds) aims to strengthen pupils’ morality and ethics and to prevent non-normative behavior such as delinquency, bullying, and shoplifting; and grade 6 (12- and 13-year-olds) encourages pupils to refrain from using alcohol, tobacco, and other drugs. The project started in 2011 with these three grade cohorts. This particular study is based on the grade 6 cohort, which only received the substance use theme of the prevention program in grade 6, and a follow-up at the beginning of grades 7, 8 and 9.

The information and educational materials are available to the teachers on the Triad website (via a personal login). A brochure can be downloaded with information about such things as how to use the materials, news, teacher tutoring (including pedagogical tips for planning the lesson), and fact sheets. Support is available if needed, and the teacher decides how to schedule the tasks and how much time each task will take.

The teachers are provided with suggested tasks for the class that aim to increase awareness in a certain area (e.g., drugs). The tasks align with the content and timeframe of the school curriculum (4–5 tasks, estimated time 80 min/task). An example of a task connected to the subject of math is: “Calculate, based on the knowledge that alcohol cost the society 5 billons Euro, how much money each individual would get out of this sum. It lives 9 billion inhabitants in Sweden”. A further example, connected to art, is “Draw a transparent person and show where in the body different diseases can strike us if we begin to drink. Present the results in front of the class.” In addition, about four 10-min informational films/grade about the current topic shows the consequences and hazards of, e.g., using drugs. Powerpoint slideshows with discussion material are also available. The Triad program also seeks to involve parents by allowing the students do tasks at home (decided by the teacher), e.g., discussing or investigating a certain topic together with the family: “Ask a grownup in your home if they know why people drink. Ask them to tell you about a situation where some grownups they met have behaved badly while drinking. Write down their story and tell the story to the class.” Or, “Ask a grownup in your home if they know why other grownups buy alcohol for underage adolescents despite knowing about the hazards of drinking alcohol.” When a task is completed, the teacher reports this via the website, and also evaluates each task.

The incentive for participating and completing the tasks is a lottery where the whole class can win prizes at the end of the school year; all tasks must be completed in order to participate. The program includes three competitions, one for each grade, and ends with a grand finale together with the parents, where the prizes are diplomas and grants dedicated for a particular experience. The finale is arranged in cooperation with the municipality and/or local businesses.

This program includes some components that have been shown to be effective in other preventive substance use programs, and some that have not. The main components, shown in Table 1, include (1) increasing resistance skills through understanding and awareness of the harmful consequences of drugs and delinquency, (2) involving parents, and (3) a lottery contest where the whole class can win prizes, which seems to be an incentive for participants to complete the tasks.

Eleven intervention schools (285 intervention pupils, participation rate 93%) and three control schools (159 control pupils, participation rate 92%) participated at baseline in 2011 (school year 6). When pupils in the intervention schools left grade 6 and went to grade 7 (junior high school) they attended the same junior high school. At baseline, participation was somewhat higher in the control schools than in the intervention schools (91 and 85% respectively; Table 1). Attrition was also higher in the intervention schools. Table 2 shows demographic data for the participants. The intervention and control schools were similar with regard to gender and ethnic background, but pupils in the intervention schools had better housing conditions and greater cultural capital, and more often lived in a nuclear family.

Swedish municipalities, which are responsible for both funding and arranging primary and secondary education, were informed about the study and invited to participate. The municipalities decided which schools would participate, and then informed the schools about the program. Hence, schools did not self-select into the intervention; therefore, the principal and teachers could be positive, neutral or negative toward the program. This could arguably improve external validity of the design relative to that of a typical randomized controlled trial in terms of predicting the effectiveness of the program if it were to be rolled out on a larger scale. All participating schools signed a contract whereby intervention schools undertook to conduct the Triad program in its entirety for three years, and control schools promised to continue with their usual preventive work but not to implement any specific program.

The pupils’ parents received information about the study’s aim, the voluntary nature of their compliance, and their right to withdraw from the study at any time. Parents who did not want their children to participate were asked to contact the research team. Participants were asked during school hours to fill in a questionnaire about family background, tobacco, alcohol, and self-rated health. The data collection was implemented by the research team.

This study is part of a larger study focusing on “School as a setting for alcohol and drug prevention” in a special venture financed by the Swedish government [50, 51]. The program was designed and implemented by the NGO, and the research team was responsible for the research and evaluation activities.

The intervention is conducted at the school level, and the adolescents could not influence whether they participated in the intervention or control group. Thus, we estimated a standard Intention-To-Treat (ITT) Difference-in-Difference (DD) logit model (given that the outcome variables are binary) as described in equation (1), where Y _it represent the different outcome variables Alcohol Consumption and Intoxication for adolescent i in grade t:

Intervention is a dummy variable indicating whether an adolescent attends an intervention school (equal to 1 if so, and 0 otherwise), and Post is a dummy variable for post-treatment data (i.e., equal to 1 for grade 7, 8 or 9, and equal to 0 for grade 6). The interaction term Intervention × Post is the causal effect of the program on the outcome variables given our assumed identification strategy, and thus β ₂ is the coefficient of main interest. Finally, X _it is a vector of individual-specific control variables and ε _it is the error term. Considering that we had repeated measures at the individual level (panel/longitudinal data), we clustered the standard errors at the individual level. Failure to do so would bias our standard errors and thus the statistical significance.

The main standard identifying assumption for drawing causal conclusions in our model, as in the typical DD setup, is that intervention and control schools would follow the same trend in alcohol consumption and intoxication albeit from different starting points (the common-trend assumption). One source of concern is that we know that the adolescents in the intervention and control schools differ slightly on a number of observable characteristics (Table 3). For example, there is a higher proportion of adolescents in nuclear families in the intervention schools. If such adolescents have a significantly different trend in their alcohol consumption behavior over time, this may bias our results. Note however that such differences affecting levels of alcohol intake are not a problem, given that our identification relies on differences in the trend (not levels). Still, in order to examine the possible influence of such confounders on differential trends in alcohol consumption, we introduced a number of control variables (X _it ) in the regressions. As a robustness check we also used propensity score matching techniques to create a balance between treatment and control schools in observable characteristics and ran the analysis on the balanced sample. We also conducted multilevel regressions, directly taking the hierarchical nature of the data into account, which yielded the same qualitative interpretations and results. ¹