The study is based on two cross-sectional datasets. Both surveys were conducted in 2021 by the Swedish Council for Information on Alcohol and Other Drugs (CAN). Both web-based, anonymous surveys were completed in the classroom. The first dataset (1) consists of data from the 2021 iteration of the Swedish National School Survey among second-year students enrolled in national programmes (VP and HEP) in upper secondary school (17- to 18-year-olds). A stratified sampling procedure ensured that all regions in Sweden were represented, and the school class (one per school) was used as the unit when selecting the sample.
The second dataset (2) consists of data from the IP in upper secondary school. To ensure anonymity, only schools with at least 20 students enrolled in the IP were included in the sampling frame. The survey, which was offered to all students in the randomly selected IP, was designed to generate results comparable to those from the annual Swedish national school survey. Students enrolled in language introduction who recently arrived in Sweden were not included since they were unable to answer the survey in Swedish or English.
The participation rate in dataset (1) was 74% on the school/class level and the response rate on the individual level (students who were present and chose to participate) was 81% [34]. The participation rate in dataset (2) was 56% at the class level and 67% at the individual level [34]. The analytical sample consisted of alcohol consumers (all students who stated that they had consumed at least one alcoholic beverage in the last 12 months) who had completed all of the survey items: 2124 students in HEP, 676 students in VP and 244 students in IP (see Table 1).
Measurements
Exposure
Information on academic orientation was provided by each school prior to the survey. The school classes included in the analysis were coded as either VP or HEP [in dataset (1)] or IP [in dataset (2)].
Table 1
Sample description - distribution of key variables across academic orientation. Among consumers.
Participants | (n) | 2124 | 676 | 244 |
Alcohol-related harm | mean | 2.41 | 3.47 | 4.26 |
Gender | | | | |
Men | % | 42 | 55 | 62 |
Women | % | 58 | 45 | 38 |
Heavy episodic drinking | | | | |
6 times/year or less | % | 43 | 41 | 42 |
Once a month or more | % | 25 | 33 | 32 |
Truancy | | | | |
Once/month or more often | % | 10 | 14 | 30 |
Family content | | | | |
Discontent | % | 12 | 13 | 19 |
CAST-6 | | | | |
Yes (≥ 3 problems) | % | 14 | 21 | 16 |
Early onset substance use | | | | |
Onset < 14 years old | % | 15 | 25 | 40 |
Outcome
Self-reported alcohol-related problems (hereafter referred to as alcohol-related harm, ARH) were measured using the following question:
Has any of the following things happened to you in relation to your alcohol consumption during the past 12 months?
1. Quarrel | 2. Deliberately harmed yourself |
3. Physical fight | 4. Victim of robbery or theft |
5. Accident or injury | 6. Trouble with the police |
7. Engaged in sex you regretted the next day | 8. Needed go to hospital or emergency room |
9. Deliberately harmed someone else | 10. Driving a moped or other motor vehicle |
11. Victim of violence | 12. Lost money or valuables |
13. Riding a moped or other motor vehicle with a drunk driver | 14. Been photographed/filmed in an embarrassing or humiliating situation |
15. Ruined clothes or other belongings | 16. Gone swimming in deep waters |
17. Problems with relations to parents | 18. Problems with relations to friends |
The response options ‘never’, ‘once’ and ‘twice’ were coded as 0, 1 and 2, respectively. Respondents who selected the highest value on more than 16 of the problem-related questions were deemed unreliable and excluded from the study. The total variable score therefore ranged from 0 to 33, with higher scores indicating more ARH. The internal reliability was good: Cronbach’s alpha = 0.86 (0.82 for HEP, 0.89 for VP, 0.92 for IP).
Explanatory variables
We included a number of risk factors that were associated with both academic orientation and ARH (Table 2).
Heavy episodic drinking (HED) was measured using a question regarding how often the student (during the past 12 months), on one continuous occasion, had consumed an amount of alcohol equivalent to at least a whole bottle of wine or four cans of beer/mixed drinks or six cans of medium-strength beer (3.5% per volume) or 18 cl of spirits. The measure intends to describe a large amount of alcohol consumed on one occasion, making it less important for respondents to know the exact amount they consumed. The responses were coded into three categories: no HED during the last 12 months, HED 6 times a year or less, and HED once a month or more often.
Table 2
Bivariate analysis, among consumers, of the explanatory variables and…
... alcohol-related harm |
| IRR | 95% CI |
Gender | 1.07 | 0.95 | 1.21 |
Heavy episodic drinking | 2.64 | 2.38 | 2.92 |
Truancy | 2.15 | 1.85 | 2.50 |
Family - discontent | 1.51 | 1.31 | 1.74 |
CAST-6 | 1.60 | 1.39 | 1.84 |
Early onset substance use | 2.68 | 2.40 | 3.00 |
... academic orientation |
| OR | 95% CI |
Gender | 1.85 | 1.40 | 2.43 |
Heavy episodic drinking | 1.25 | 1.11 | 1.42 |
Truancy | 2.23 | 1.69 | 2.93 |
Family - discontent | 1.29 | 1.03 | 1.62 |
CAST-6 | 1.47 | 1.20 | 1.78 |
Early onset substance use | 2.44 | 1.98 | 3.00 |
IRR = incidence rate ratio, OR = Odds ratio. | | |
Truancy Responses ranging from once a month to several times a week were coded as recurrent truancy.
CAST-6. To investigate the extent to which students felt that their parents’ alcohol use was problematic, the Children of Alcoholics Screening Test (CAST-6) was used. This instrument consists of six questions and is a shortened version of the original 30-question test. CAST-6 identifies children who perceive their parents’ alcohol consumption as problematic equally well as does the original 30-question test [37]. CAST-6 poses six questions regarding the negative consequences of parents’ drinking. If three of these six negative consequences were met, the student was coded as having at least one parent with problematic alcohol use, which is an established cutoff [37].
Early-onset substance use was defined as the use of any of the substances stated below before the age of 14 years and was measured using five questions: How old were you when (if ever) you did the following things for the first time? (1) Drank at least one glass of alcohol, (2) Got drunk on alcohol, (3) Smoked a cigarette, (4) Used moist snuff, (5) Used marijuana or hashish. Individuals who responded yes to any of these questions were defined as having an early onset of substance use.
Family discontent was measured using the question ‘How content are you with your relationship with your family in general?’ Those who reported being discontented, very discontented or neither contented nor discontented were coded as being discontented with their relationship with their family.
In addition, all analyses were adjusted for gender, measured using a question where the students could choose between male, female and other gender identities. The respondents who chose one of the two first alternatives were included in the study.
Statistical analyses
Since the outcome was a count variable with an overdispersed distribution (mean = 2.80, variance = 19.44), negative binomial regressions were used to assess the association between academic orientation and ARH. The results section presents incidence rate ratios (IRRs) and 95% confidence intervals.
First, we estimated the bivariate association between each explanatory variable and ARH. Next, to explore the social gradient in ARH, we estimated the first model, which included academic orientation and gender (included in all the models). To assess the contribution of the explanatory factors, we estimated models with separate adjustments for HED (Model 2), truancy (Model 3), early-onset substance use (Model 4), family discontent and CAST-6 score (Model 5). The final model included all the explanatory variables.
To compute the attenuation (in percentages) of the crude association between academic orientation and ARH in each of the consecutive models, the following formula was used: IRRcrude-IRRadjusted/IRRcrude-1 × 100.
Since the students in dataset (1) were clustered in school classes and the students in dataset (2) were clustered in schools, we used the vce (cluster) option in Stata version 17 to account for the clustering. This procedure relaxes the usual assumption of independence between observations and affects the standard errors but not the estimates [38].
To test for effect modification, we tested for interactions between academic orientation and HED, both on the multiplicative and the additive scales.
Multiplicative interactions were tested by including an interaction term between academic orientation and HED (academic orientation × HED). To test for additive interactions, we included a variable that combined information on academic orientation and HED. We then calculated the relative excess risk due to interaction (RERI) using the formula RERI = IRR++ – IRR+− – IRR−+ + 1. This measure assesses whether the IRRs of the combined exposures (i.e., IP students and HED, IRR++) are greater or smaller than the sum of the IRRs for each exposure individually (i.e., IP students with no HED or IRR+− and not IP student and HED, IRR−+).
We also calculated the proportion of ARH among those with both exposures that was attributable to their interaction; AP (attributable proportion) = RERI/IRR++.