The contribution of drinking culture at comprehensive school to heavy episodic drinking from adolescence to midlife

Abstract

Background

The school context is associated with adolescent alcohol use, but it is not clear whether this association continues into adulthood. This study examined whether exposure to drunkenness oriented drinking culture in 9th grade school class is associated with individuals’ heavy episodic drinking (HED) from adolescence to midlife.

Methods

Participants in the ‘Northern Swedish Cohort’ study aged 16 years in 1981 were followed-up when aged 18, 21, 30 and 43 (N = 1080). Individual-level factors were HED, positive attitudes towards drunkenness, early initiation of HED and peer-oriented spare-time. School class-level drinking culture was measured as classmate reported HED, positive attitudes, early initiation of HED and peer-oriented spare time. Multilevel log-binomial regression analyses were adjusted for gender, parental socioeconomic background, family structure and HED at age 16.

Results

After adjustment for sociodemographic factors several cross-sectional and longitudinal associations were found between class-level indicators of drinking culture and individual HED. After additional adjustment for age 16 HED, most associations attenuated. The risk ratio (95% confidence interval) for engaging in HED at age 43 was 1.58 (1.03–2.42) times higher for those who at age 16 had many classmates reporting positive attitude towards drunkenness.

Conclusions

These findings suggest that drinking culture in school may have a long-lasting impact on drinking habits in adulthood. The associations with HED at follow-ups are likely mediated by HED in adolescence. Studies on alcohol use would benefit from taking into account both individual and contextual factors in a life course perspective.

Introduction

Problematic alcohol use such as heavy episodic drinking (HED, usually defined as drinking 4–6+ alcohol units at one time) has been identified as a public health concern that is affected by risk factors at multiple levels.¹ Several individual and contextual factors have been found to be associated with alcohol use.² However, in public health literature on alcohol, the focus has mainly been on the individual over context,³ even though a thorough understanding of the complex developmental processes regarding alcohol use requires more in-depth information on both the individual and contextual factors and their interplay in a life course perspective. Alcohol-related contextual factors can be characterized, e.g by a ‘drinking culture’. The concept of ‘drinking culture’ refers to norms concerning patterns, practices, settings and occasions in relation to alcohol use that operate on a societal macro-level (e.g. national culture) or on a micro-level in a subgroup within society (e.g. culture among a group of friends).³ Most previous studies on drinking culture have focused on the macro-level and national drinking cultures,³ although especially in adolescence, when drinking habits are usually adopted, the micro-level immediate environment is especially relevant⁴ and the focus in this study.

Bronfenbrenner’s bioecological theory is a framework that can be used for analysing the interplay between an individual and the environment (including other persons, objects and symbols) through time.⁵ Individual characteristics (e.g. age, sex, drinking habits) are associated with context, which can be viewed from several different levels such as micro (immediate environment) and macro (the broadest layer). In this framework, the temporal aspects of development (either as chronological time such as age of a person or historical time) are also important.

Drinking habits are often adopted and maintained in a social context⁶ and especially the features of the surrounding drinking culture may influence the person’s own alcohol use. For adolescents, relevant micro-level contextual settings are family, peer group and school, and in this study, we focus on the latter as several measures have been taken in schools to tackle detrimental alcohol use among adolescents.⁷

Previous studies have used multilevel modelling to examine the association between contextual factors and adolescent alcohol use, but they have mostly been cross-sectional. Some studies have found school-level alcohol related and socioeconomic factors to be associated with adolescent alcohol use^8–11 (suggesting mixed associations), while others have not found any association.^12,13

Some multilevel studies have examined longitudinal associations between school-related characteristics and alcohol use,^14–16 but very few have a follow-up beyond adolescence. However, schools and school classes form important social groups, which have been found to be associated with adolescent drinking.⁴ It is not known whether these early-life associations with drinking habits extend into adulthood. Particularly, there are few studies extending follow-up to midlife, when HED usually decreases.¹⁷ A longitudinal study from the UK did not find any association between type of primary school attended in the 1960s and adult alcohol use.¹⁸

The term ‘drinking culture’ encompasses both aspects, ‘drinking customs’ and ‘expectations about drinking’,³ and in the school environment this could mean, e.g. school- or class-level patterns of alcohol use and HED, as well as pupils’ attitudes towards alcohol use, e.g. acceptance of drunkenness. A follow-up study of adolescents found that the school-level prevalence of alcohol use, but not the perceived ‘coolness’ of substance use, was associated with both current alcohol use and HED 2 years later.¹⁹ This association attenuated after both measures were included in the model simultaneously, taking into account individual scores of alcohol use and perceived coolness of alcohol use at baseline. This may suggest a possible mediating effect, where school’s drinking culture influences adolescent alcohol use, which in turn has an impact on alcohol use in adulthood. This may explain the lack of independent association found in these few previous longitudinal studies. In addition to these clearly alcohol-related aspects of drinking culture in adolescents, the aspects of spare time are also relevant. Previous studies have concluded that those adolescents who are more peer-oriented tend to use more alcohol,^20–22 whereas adolescents engaged more in other activities use less, perhaps because of fewer opportunities to use alcohol and/or less peer pressure.

Preventive measures to tackle risky and harmful alcohol use require knowledge of the effects of early environments on individual alcohol use, in addition to traditional targets regarding behaviour change and unhealthy lifestyles at individual level.³ This study continues the context-focused research conducted by multilevel school studies and takes a step further by examining the association between school context and alcohol use extending into adulthood.

Using Bronfenbrenner’s bioecological theory as a framework, this study examines the associations between alcohol-related characteristics in the school environment and individual-level HED during the life course. The more specific study question is: ‘Is exposure to drunkenness oriented drinking culture in class during adolescence associated with individuals’ HED from adolescence to midlife (ages 16–43)?

Methods

Population

The present study includes information at two levels: individual and 9th grade school class. The target group for individual-level data included all 9th grade pupils (N = 1083) attending (or those who should have attended) comprehensive school in 1981 in Luleå, Sweden. At age 16, 1080 (99.7%) pupils completed questionnaires during school hours. The participants were followed-up by collecting survey questionnaires at class reunions or using postal questionnaires in the years 1983, 1986, 1995, 2008, at the respective ages of 18, 21, 30 and 43 years. The participation rate in the latest follow-up was 93.3% (N = 1010). The number of schools in Luleå region was 9 and there were 46 classes in these schools. The mean number of pupil respondents per school was 91 (range 79–170) and per class 22 (range 5–27). The study was approved by the Regional Ethical Review Board in Umeå, Sweden. Class-level data included aggregate variables comprising the prevalence of several different characteristics of pupils.

Measures

Individual-level factors

HED was measured at ages 16, 18, 21, 30 and 43. The respondents reported the frequency of drinking occasions (every day or every second day, one to two times weekly, a couple of times monthly, more seldom, never) and average intake of beer (number of bottles), wine (number of glasses) and strong beverages (number of drinks) on each occasion.

HED was defined so that respondents who reported drinking on average ≥5 bottles of beer, ≥5 glasses of wine or half a bottle (37 cl) or more of strong beverages at one time and drinking at least a couple of times monthly were classified as heavy episodic drinkers. Participants who reported drinking the aforementioned amounts more seldom (or not drinking) were classified as ‘not heavy episodic drinkers’.

Attitude to drunkenness was based on a question ‘what do you think about adolescents of your age that get drunk?’ The scale was 1 = ‘Should never drink’, 2 = ‘Never should drink to be drunk’, 3 = ‘OK if it happens occasionally’, 4 = ‘OK once a month’, 5 = ‘OK once a week’ and 6 = ‘OK more often than once a week’ and was dichotomized to negative (Options 1–3) and positive (Options 4–6) attitude. Participants were also asked about their age when they were drunk for the first time. Participants who had been drunk at age 13 or earlier were considered as initiating HED early.

Peer-oriented spare time was based on a question ‘are you most often by yourself or with friends in your spare time?’(‘almost always with friends’ to ‘almost always alone’). Almost always spending time with friends was considered peer-oriented spare time.

Contextual, class-level characteristics

Class-level variables indicating drinking culture were derived from measures of individual responses about HED, attitude towards drunkenness, early initiation of HED and peer-oriented spare time. Aggregated variables were calculated as the prevalence of individual classmates reporting HED/positive attitude to HED/early initiation of HED/peer-oriented spare time in class (self-estimation excluded) and then these percentages were linked to each pupil using class codes. Aggregate-level scores were divided into tertiles for the analysis, the highest tertile indicating the culture most favourable to drinking.

Control variables

Parental social class was based on survey responses of father’s and mother’s occupation. Parental occupations were classified as ‘manual’ or ‘non-manual’ according to the Swedish classification system.²³ Participants were categorized into ‘low parental social class’ if both their mother and father had a manual occupation. Living in a single-parent family was based on survey responses (yes/no).

Class-level prevalence of pupils living in families with only manual workers and the prevalence of pupils in single-parent families were used as control variables in class-level analyses.

Statistical analyses

The SAS 9.3 GLIMMIX procedure in multilevel log-binomial regression analyses was used to assess class- and individual-level effects on individual-level HED outcomes (i) cross-sectionally at age 16 and (ii) longitudinally between ages 18–43 years in a hierarchical context. Analyses were performed separately for each drinking-related variable at age 16 (HED, attitude, early initiation of HED and peer-oriented spare time). Risk ratios (RRs) with 95% confidence intervals (CIs) were calculated. Analyses were conducted in three stages. First, only gender was adjusted for (Model 1). Then, additionally baseline socioeconomic background and single-parent status were included (Model 2). Finally, all covariates (including baseline HED) were included in the model (Model 3). In individual-level analyses, parental social class and single-parent family status were used as individual-level control variables and correspondingly in class-level analyses they were used as class-level variables (self-estimation included).

Analyses stratified by gender were used as sensitivity analyses and results that differ from analyses conducted in the total population are presented when applicable.

Results

The individual- and class-level characteristics for women and men are shown in table 1 and the associations between control variables and individual alcohol-related variables at baseline in Supplementary appendix table S1. In women, 10.6–13.5% were heavy episodic drinkers, except at age 43, when the prevalence was 7.9%; in men, the prevalence varied from every fifth at ages 16 and 43 to almost half at age 21. Over a third had both parents from manual occupations and about 10% lived in a single-parent family. The proportions of drinking culture indicators between each individual’s classmates were 0–32.0% for HED, 0–45.5% for positive attitude towards drunkenness, 0–62.5% for early initiation of HED and 4.8–50.0% for spending peer-oriented spare time.

Table 1

Descriptive characteristics of participants and school classes

	Women (N = 496)	Men (N = 543)	Total (N = 1039)
	% (n)	% (n)	% (n)
Individual-level
Heavy episodic drinking (16)	13.5 (65)	17.4 (91)	15.5 (156)
Heavy episodic drinking (18)	12.1 (58)	32.3 (169)	22.6 (227)
Heavy episodic drinking (21)	10.6 (51)	43.7 (229)	27.9 (280)
Heavy episodic drinking (30)	11.3 (54)	29.4 (154)	20.7 (208)
Heavy episodic drinking (43)	7.9 (38)	20.4 (104)	14.3 (142)
Positive attitude towards drunkenness (16)	15.1 (72)	23.2 (120)	19.3 (192)
Early initiation of HED (16)	30.0 (143)	25.7 (132)	27.8 (275)
Peer-oriented spare time (16)	24.4 (117)	32.4 (169)	28.6 (286)
Low parental SEP (16)	34.4 (165)	37.4 (195)	36.0 (360)
Single-parent family (16)	10.3 (49)	12.0 (63)	11.2 (112)
Class-level	Mean (SD)	Mean (SD)	Mean (SD)
Prevalence of heavy episodic drinkers	15.5 (8.4)	15.3 (8.4)	15.4 (8.4)
Prevalence of positive attitudes towards drunkenness	19.7 (11.5)	19.0 (11.6)	19.4 (11.6)
Prevalence of early initiation of HED	27.9 (13.7)	27.3 (13.1)	27.6 (13.4)
Prevalence of peer-oriented spare time	28.5 (10.0)	28.4 (9.9)	28.4 (9.9)
Prevalence of manual background families	36.1 (14.6)	35.9 (14.5)	36.0 (14.5)
Prevalence of single-parent families	10.8 (7.8)	11.6 (8.3)	11.2 (8.0)

Table 2 shows the results of multilevel log-binomial regression models for the associations between individual baseline alcohol-related variables and HED at ages 16–43. In cross-sectional analyses, all variables (positive attitude towards drunkenness, early initiation of HED and peer-oriented spare time) were associated with increased risk of age 16 HED in Model 1 and 2 (Model 3 not applicable). After adjustment for gender, parental SEP and family structure, age 16 HED, positive attitude towards drunkenness and early initiation of HED were associated with increased risk of HED at all follow-up ages (18–43). However, the association between positive attitude towards drunkenness and HED at age 30 did not research statistical significance (P < 0.05) and showed somewhat smaller RR than at other ages. Peer-oriented spare time was associated with HED at ages 18 and 21, although in women the association attenuated at age 21. Additional adjustment for age 16 HED attenuated the associations between early initiation of HED and HED at ages 21–43 for the total population, but the risk remained high in men. In addition, the association between positive attitude towards drunkenness and HED at ages 21–43 attenuated although remained statistically significant and strong for women at age 43.

Table 2

Risk ratios (RR) [95% confidence intervals (CI)] for the associations between individual-level exposures and HED at ages 16, 18, 21, 30 and 43^a

	HED 16	HED 18	HED 21	HED 30	HED 43
Model 1b	RR (95% CI)	RR (95% CI)	RR (95% CI)	RR (95% CI)	RR (95% CI)
HED 16	Non-applicable	3.71 (3.06–4.50)	2.17 (1.85–2.54)	1.80 (1.40–2.31)	2.30 (1.68–3.15)
Positive attitude towards drunkenness	5.53 (4.17–7.34)	2.36 (1.91–2.93)	1.73 (1.45–2.06)	1.34 (1.03–1.74)c	1.93 (1.40–2.65)
Early initiation of HED	5.75 (4.19–7.88)	2.26 (1.82–2.81)	1.60 (1.33–1.91)c	1.51 (1.19–1.93)c	1.52 (1.09–2.11)c,d
Peer-oriented spare time	1.71 (1.28–2.29)	1.50 (1.20–1.87)	1.27 (1.06–1.53)c	1.11 (0.86–1.43)	1.25 (0.91–1.72)
Model 2e
HED 16	Non-applicable	3.49 (2.85–4.28)	2.09 (1.77–2.47)	1.71 (1.33–2.21)	2.16 (1.56–2.97)
Positive attitude towards drunkenness	5.23 (3.92–6.99)	2.17 (1.75–2.70)	1.65 (1.38–1.98)	1.29 (0.99–1.68)	1.79 (1.30–2.47)
Early initiation of HED	5.55 (4.03–7.65)	2.15 (1.72–2.68)	1.56 (1.30–1.87)c	1.44 (1.13–1.84)c	1.43 (1.03–2.01)c,d
Peer-oriented spare time	1.69 (1.27–2.27)	1.54 (1.24–1.90)	1.29 (1.07–1.54)c	1.09 (0.85–1.40)	1.21 (0.88–1.66)
Model 3f
HED 16g	Non-applicable	Non-applicable	Non-applicable	Non-applicable	Non-applicable
Positive attitude towards drunkenness	Non-applicable	1.43 (1.05–1.95)d,h	1.18 (0.97–1.43)	1.04 (0.78–1.39)	1.37 (0.96–1.97)i
Early initiation of HED	Non-applicable	1.70 (1.13–2.55)c,d	1.21 (0.99–1.46)j	1.20 (0.92–1.58)j	1.06 (0.73–1.53)
Peer-oriented spare time	Non-applicable	1.32 (1.003–1.73)c,d,h	1.19 (1.02–1.38)c,d	1.03 (0.80–1.32)	1.13 (0.82–1.55)

a

Drinking culture variables separately in the model.

b

Model 1: adjusted for gender.

c

In gender-stratified analysis, not statistically significant in women.

d

In gender-stratified analysis, not statistically significant in men.

e

Model 2: Model 1 + adjusted for parental socioeconomic background and single-parent family.

f

Model 3: Model 2 + adjusted for heavy episodic drinking at baseline (16 years).

g

Model 3 is not applicable, because baseline HED is already included in the model.

h

The estimates were produced using the Poisson distribution due to conversion problems with binomial distribution.

i

In gender-stratified analysis, statistically significant in women.

j

In gender-stratified analysis, statistically significant in men.

Table 3 shows the results for associations between class-level drinking culture variables and individual HED at ages 16–43. In the cross-sectional analyses, the higher class-level prevalence of HED, positive attitudes towards drunkenness and early initiation of HED were associated with high risk for individual HED at age 16 after adjustment for sociodemographic variables. Regarding the follow-ups, after adjustment for sociodemographic factors the pupils whose classmates belonged to the middle or highest tertile of HED had an increased likelihood of HED at age 21 and also at age 43 for the highest tertile. Positive attitude towards drunkenness in class (middle and highest tertiles) indicated a greater likelihood of HED at ages 18 and 43. Associations with HED at ages 21 and 30 showed lower and statistically non-significant results. Early initiation of HED (middle tertile) was statistically significantly associated with HED at age 21 and at age 43 (highest tertile P < 0.1). After additional adjustment for baseline HED the statistically significant associations mostly diminished and remained only between class-level HED (highest tertile) and HED at age 43 (P < 0.1) and class-level positive attitude towards drunkenness (middle and highest tertile) and HED at age 43. Supplementary appendix table S2 summarizes this information and suggests that few associations provided strong estimates (RR ≥1.5) and remained statistically significant in the fully adjusted model. In gender-stratified analyses, these associations in full models were not statistically significant, although effect sizes remained somewhat similar. Positive attitude towards drunkenness (middle and highest tertile) remained nearly and fully statistically significantly associated with increased likelihood of HED at age 18 in men (RR 1.50, 95% CI 0.994–2.27; RR 1.50, 95% CI 1.003–2.26, respectively) in the final model. In women, class-level peer-oriented spare-time (middle and highest tertiles) showed increased risk for age 18 HED (RR 2.29, 95% CI 1.18–4.46; RR 2.13, 95% CI 1.13–4.02, respectively).

Table 3

Risk ratios (RRs) [95% confidence intervals (CIs)] for the associations between class-level exposures and individual HED at ages 16, 18, 21, 30 and 43^a

		HED 16	HED 18	HED 21	HED 30	HED 43
Model 1b		RR (95% CI)	RR (95% CI)	RR (95% CI)	RR (95% CI)	RR (95% CI)
Prevalence of HED	1 Low	1.00	1.00	1.00	1.00	1.00
	2	1.40 (0.95–2.08)	0.98 (0.71–1.34)	1.23 (0.97–1.55)	1.09 (0.81–1.48)	1.06 (0.70–1.60)
	3 High	1.75 (1.21–2.55)c	1.11 (0.81–1.52)	1.30 (1.04–1.64)e	1.27 (0.95–1.70)	1.60 (1.10–2.33)e
Prevalence of positive attitudes towards drunkenness	1 Low	1.00	1.00	1.00	1.00	1.00
	2	1.53 (1.02–2.28)c,e	1.48 (1.08–2.03)e	1.28 (1.01–1.62)c,e	1.11 (0.82–1.51)	1.74 (1.14–2.66)e
	3 High	1.80 (1.23–2.64)e	1.47 (1.07–2.01)e	1.23 (0.97–1.56)	1.29 (0.96–1.72)	1.77 (1.17–2.69)e
Prevalence of early initiation of HED	1 Low	1.00	1.00	1.00	1.00	1.00
	2	1.92 (1.30–2.82)	1.16 (0.83–1.62)	1.26 (0.99–1.59)	0.99 (0.74–1.34)	1.15 (0.76–1.74)
	3 High	1.72 (1.16–2.55)e	1.21 (0.87–1.68)f	1.20 (0.95–1.53)	1.06 (0.80–1.42)	1.47 (1.0003–2.16)c,e
Prevalence of peer-oriented spare time	1 Low	1.00	1.00	1.00	1.00	1.00
	2	0.79 (0.54–1.15)	0.94 (0.69–1.29)	0.87 (0.69–1.11)	0.88 (0.66–1.18)	0.93 (0.62–1.40)
	3 High	1.09 (0.77–1.56)	1.23 (0.90–1.68)g	1.10 (0.87–1.37)	0.99 (0.75–1.32)	1.31 (0.89–1.92)
Model 2h
Prevalence of HED	1 Low	1.00	1.00	1.00	1.00	1.00
	2	1.36 (0.91–2.05)c	0.98 (0.71–1.36)	1.28 (1.004–1.63)c,e	1.11 (0.81–1.51)	1.09 (0.71–1.68)
	3 High	1.69 (1.16–2.47)c	1.09 (0.80–1.49)	1.30 (1.03–1.63)c,e	1.27 (0.94–1.70)	1.61 (1.10–2.35)e
Prevalence of positive attitudes towards drunkenness	1 Low	1.00	1.00	1.00	1.00	1.00
	2	1.53 (1.02–2.30)c,e	1.44 (1.05–1.96)e	1.20 (0.95–1.52)	1.09 (0.80–1.49)	1.70 (1.11–2.61)c,e
	3 High	1.81 (1.23–2.67)e	1.42 (1.05–1.94)e	1.18 (0.94–1.49)	1.28 (0.95–1.71)	1.75 (1.15–2.66)e
Prevalence of early initiation of HED	1 Low	1.00	1.00	1.00	1.00	1.00
	2	2.09 (1.40–3.11)	1.27 (0.91–1.77)	1.33 (1.05–1.68)e	1.04 (0.77–1.42)	1.19 (0.78–1.81)
	3 High	1.63 (1.09–2.42)c,e	1.21 (0.87–1.67)f	1.18 (0.93–1.49)	1.05 (0.78–1.40)	1.47 (0.998–2.17)
Prevalence of peer-oriented spare time	1 Low	1.00	1.00	1.00	1.00	1.00
	2	0.77 (0.53–1.13)	0.96 (0.70–1.31)g	0.89 (0.70–1.12)	0.89 (0.66–1.20)	0.95 (0.63–1.44)
	3 High	1.03 (0.72–1.45)	1.18 (0.87–1.62)g	1.05 (0.84–1.31)	0.98 (0.73–1.31)	1.32 (0.89–1.96)
Model 3i
Prevalence of HED	1 Low	Non-applicabled	1.00	1.00	1.00	1.00
	2		0.94 (0.67–1.33)j	1.22 (0.89–1.66)j	1.08 (0.80–1.48)	1.03 (0.69–1.58)
	3 High		1.04 (0.75–1.43)j	1.23 (0.91–1.65)j	1.20 (0.90–1.61)	1.46 (0.999–2.13)
Prevalence of positive attitudes towards drunkenness	1 Low	Non-applicabled	1.00	1.00	1.00	1.00
	2		1.25 (0.89–1.77)j	1.15 (0.85–1.55)j	1.05 (0.77–1.43)	1.58 (1.03–2.43)c,e
	3 High		1.17 (0.83–1.64)f,j	1.09 (0.81–1.48)j	1.20 (0.90–1.61)	1.58 (1.03–2.42)c,e
Prevalence of early initiation of HED	1 Low	Non-applicabled	1.00	1.00	1.00	1.00
	2		0.98 (0.75–1.30)	1.22 (0.90–1.66)j	0.97 (0.71–1.31)	1.07 (0.70–1.64)
	3 High		1.15 (0.90–1.47)	1.12 (0.83–1.51)j	0.99 (0.74–1.32)	1.35 (0.91–1.99)
Prevalence of peer-oriented spare time	1 Low	Non-applicabled	1.00	1.00	1.00	1.00
	2		1.05 (0.80–1.38)g	0.94 (0.69–1.26)j	0.93 (0.70–1.25)	1.00 (0.66–1.50)
	3 High		1.20 (0.92–1.57)g	1.08 (0.80–1.44)j	0.96 (0.72–1.29)	1.33 (0.91–1.95)

a

Drinking culture variables separately in the model.

b

Model 1: adjusted for gender.

c

In gender-stratified analysis, not statistically significant in men.

d

Model 3 is not applicable, because baseline HED is already included in the model.

e

In gender-stratified analysis, not statistically significant in women.

f

In gender-stratified analysis, statistically significant in men.

g

In gender-stratified analysis, statistically significant in women.

h

Model 2: Model 1 + adjusted for class prevalence of manual worker parents, and class prevalence of single-parent families.

i

Model 3: Model 2 + adjusted for heavy episodic drinking at baseline (16 years).

j

The estimates were produced using the Poisson distribution due to conversion problems with binomial distribution.

Discussion

Applying the bioecological framework and integrating individual, contextual and life course aspects, this study examined whether exposure to drunkenness-oriented drinking culture in comprehensive school was associated with HED in later life. Regarding different aspects of drinking culture, after considering class-level confounding and baseline HED, positive attitudes towards HED in class showed an association with HED at age 43, while the association with HED at age 18 was shown only in men. In women, class-level peer-oriented spare time showed an association with HED at age 18. No statistically significant class-level associations were found regarding HED at ages 21 and 30. This highlights the importance of employing a life course perspective—time—as well as contextual factors, in addition to individual effects. HED is very common when people are in their twenties and early thirties and HED usually decreases with age. It may be that those who still often drink heavily in their forties deviate from others regarding, e.g. the ways they have adopted their drinking habits.

After adjustment for all covariates, including baseline HED, very few associations provided strong estimates (RR ≥1.5) and remained statistically significant. Several explanations may be possible. First, it may be that the association with school class diminishes over time. Secondary school forms a context for a limited time in a person’s life and is eventually replaced with other contexts such as other schools and workplaces. It may be that other contexts become more important with age. This would suggest that in order to successfully prevent risky alcohol use during the life course, the interventions should perhaps be targeted on contexts relevant in these later life stages as well, and not just on schools earlier in life. However, even though the association with context in adolescence mostly diminished over time, our results show that the school context may play a role in the development of alcohol habits over decades.

Second, our findings suggest that school class may still not be the most relevant setting in adolescents’ life to be analysed. Although peers have been found to play an important role in adolescent alcohol use,^24,25 the closeness of school classes as social units may vary. We did not have information on whether the classmates considered themselves close friends, thus in some cases, the closest friends could have been in other classes or schools. The indicators of drinking culture may influence individuals’ alcohol use in many ways. An adolescent’s alcohol use may be influenced by peer pressure, peer alcohol norms and spending time with substance-using friends,²⁵ who are not necessarily attending the same school class, as well as by social control and social learning.²⁶

A third explanation for not finding many associations in this study may be that we examined context only at the micro-level, i.e. school class. In the bioecological theory, context can be viewed from several levels and some macro-level aspects such as period effects, alcohol prices and advertising restrictions may be more relevant than micro-level culture.^25,27,28 On a national level drinking culture is similar for all participants (who did not move abroad), but there may be area-level differences between those who moved to other parts of Sweden after adolescence and those who stayed.²⁹

Fourth, the lack of strong associations between school-related contextual factors and adulthood HED may simply be that individual factors (e.g. gender) are most important in the development of drinking habits. Class-level peer-oriented spare time was statistically significantly associated with HED at age 18 only in women. In the final model, we took into account individual HED at baseline, which diminished many associations that showed increased risk for HED at ages 18, 21 and 43 in Model 2. It is not possible to draw firm conclusions about the direction of associations as the participants may have engaged in HED already before they were exposed to drunkenness-oriented drinking culture in class. However, it may also be that the association between class-level drinking culture and HED in adulthood may be almost fully mediated by HED in adolescence, which highlights the importance of targeting interventions on this life stage.

Methodological considerations

The main strengths of this study are the long follow-up time and the very high participation rate throughout the study. However, some issues should be taken into account when interpreting the findings. The measure of HED does not consider persons who combine different beverages as heavy episodic drinkers, if the amounts within one beverage type did not exceed the limits used in this study. Because of this, the prevalence of HED may be underestimated; however, the frequencies observed correspond well to those found in other population-based Swedish studies that measure HED in more detail.³⁰

The focus in this study was on class-level in schools, not examining broader, e.g. community-, neighbourhood- or country-level effects, which may have caused risk for unmeasured confounding. Future studies could benefit from analysing several different contextual levels simultaneously. In some cross-sectional studies, community-level variation in alcohol use has been found.³¹ In addition, it should be acknowledged that social contexts of adolescents today are somewhat different from those of participants of this study, who were adolescents in the 1980s. New contexts, such as online social networking, likely influence alcohol use.³

In some analyses, the lack of statistically significant associations may have been due to lack of power. For example, middle and highest tertiles of class-level attitude towards drunkenness was associated with the individual HED at age 43 in the total population, but in the analyses stratified by gender, statistical significance could not be verified to either gender.

Conclusion

In conclusion, the findings of this prospective study suggest that exposure to drinking-oriented culture in the 9th grade of comprehensive school has some associations with drinking habits lasting into adulthood, with regard to e.g. positive attitudes towards drunkenness among schoolmates. Schools are well-suited settings for preventive interventions against HED, because most of the age group can be reached there. However, in order to gain long-term effects, the actions in substance use prevention should be more holistic and also include other contexts (e.g. family, workplace) and extend to several life stages. Studies on alcohol use as well as interventions would benefit from taking into account both individual and contextual factors in a life course perspective.

Funding

This work was supported by the Swedish Research Council Formas [259-2012-37 to A.H., N.B., M.V.], the Swedish Research Council for Health, Working Life and Welfare [2011-0445 to A.H.] and the Cutting Edge Medical Research granted by the County Council of Västerbotten [VLL-355661 to A.H.].

Conflicts of interest: None declared.

Key points

• School context is associated with adolescent alcohol use, but it is not clear how strongly this association continues into adulthood.

• Exposure to drinking-oriented culture in adolescence in school has some associations with drinking habits lasting into adulthood.

• In order to gain long-term effects, actions in substance use prevention should be more holistic and also include other contexts (e.g. family, workplace) and extend to several life stages.