Alcohol habits in Sweden during 1997–2018: a repeated cross-sectional study

Abstract

Aims: This study describes changes in alcohol habits in age and gender strata among the Swedish general population during the 21-year period following the Swedish EU membership in 1995.

Methods: The Alcohol Use Disorders Identification Test (AUDIT) was distributed as a postal questionnaire to randomly selected cross-sectional samples every fourth year, starting in 1997.

Results: Six samples were included in this study. A comparison between 2014 and 2018 showed a tendency to decreasing average total AUDIT scores, particularly among men and in the youngest age group (Cohen’s d = 0.28).

Conclusion: The results should be understood in the context of declining response rates in this type of survey.

Keywords:

• AUDIT
• alcohol drinking
• response rates

Introduction

Swedish alcohol policy has traditionally been oriented towards reducing overconsumption of alcohol by restricting access, and viewing alcohol as a beverage intended for specific social occasions. High taxes and limited opening hours for the national alcohol retailing monopoly have been used as policy mechanisms to restrict the availability of alcohol [1]. Since becoming a member of the European Union (EU) in 1995, Sweden has increasingly abandoned the tradition of restrictiveness and approached the European view of alcoholic drinks as a commodity that should be accessible in the same way as other goods. As neighboring countries reduced alcohol taxation, Swedish import quotas were increased, contributing to a loss of legitimacy for Swedish alcohol policy with its retailing monopoly. In the late 1990s and early 2000s, harmonization with EU alcohol legislation and the increased availability of alcohol coincided with a shift in alcohol consumption patterns from intensive drinking on weekends to a more balanced drinking pattern over the days of the week. The incremental increase in import quotas and the liberalization of alcohol legislation led to price cuts on alcohol and a resulting increase in purchasing power. These changes were associated with a continuous increase in average alcohol consumption in Sweden over the years 1996–2004 from 8.0 to 10.6L, followed by a period of decrease to about 9 L in 2012, where it has remained somewhat stable [2]. This latter decrease may have resulted from people adapting to the increased accessibility of alcohol. The changes may also have resulted from the above-noted shift to a continental method of drinking alcohol characterized by a larger proportion of wine consumption [2] and decreased intensive consumption in the weekends [3]. Another influential factor may have been the retailing monopoly’s launch of bag-in-box wines, which accounted for more than 40% of total alcohol sales in 2010 [4].

The Skog theory about collectivity of drinking cultures states that changes in the aggregate alcohol consumption are uniformly distributed across different subgroups [5]. Due to the interconnectedness between individuals in the population, alcohol habits will spread through social contagion so that if the population average consumption changes, then all subgroups also make a similar change. Skog and Rossow [6] further elaborate this theory and propose that it is valid for different aggregated consumption levels but not for all groups. However, it has been shown that the incidence of alcohol-related problems can increase in some groups, although the entire cohort on average shows decreasing alcohol consumption. Indeed, two studies have shown that although reduced alcohol consumption between 2001 and 2014 occurred among high school students in Stockholm, alcohol-related injuries increased, negating theoretical predictions [7,8]. This occurred because a large proportion of individuals who initially drank relatively little decreased their consumption, while a small proportion who initially had high levels of consumption increased it further and also suffered more alcohol-related injuries. A study by Raninen et al. [9] showed that the distribution of alcohol consumption patterns among groups in the Swedish population does not significantly differ, thus lending support to the Skog theory. When Sweden joined the EU, Swedish alcohol researchers expressed concern that alcohol consumption would increase, but so far, studies show that alcohol consumption remained more stable than initially expected, although at somewhat higher level than before, after an initial increase.

The Alcohol Use Disorders Identification Test (AUDIT) is a 10-item instrument that is well-adapted for assessing risky and harmful alcohol habits, and also for mapping alcohol-related problems, both in clinical contexts and population samples. The AUDIT was developed in a collaborative effort in six different countries with differing cultures, for use primarily in primary care and is recommended by the World Health Organization (WHO). The AUDIT has been psychometrically tested and validated in various survey materials [10–12]. In Sweden, the test has been validated in a patient sample with high prevalence of alcohol problems [13] as well as in the general population [14,15]. Since 2007, national guidelines published in Sweden by the National Board of Health and Welfare recommend that AUDIT be used for screening of risk consumption. Also, a published manual describes the practical use of AUDIT and, in addition, its equivalent in the drug field, DUDIT [16,17].

wThis study describes the development of alcohol habits in Sweden over a 21-year period between 1997 and 2018, with repeated cross-sectional surveys using the AUDIT instrument. The primary aim is to report changes in alcohol habits between 1997 and 2018. Since changes in alcohol habits between 1997 and 2014 have been reported elsewhere [18–20], the focus of this study is on the period between 2014 and 2018.

Methods

Participants and procedure

Data collection in the years 1997, 2001, 2005 and 2009 was performed using a systematic random selection process, where a number, n, was chosen randomly, from an official national register (SPAR) with equal proportions of men and women, and aged between 17 and 71 years. Each n:th person was then selected to be included in the study for a total sample of 1250 individuals with a permanent address across Sweden. In 2014, the sample size was increased to 1450 individuals and to 1500 individuals in 2018; in these two cohorts, older individuals between the ages of 72 and 80 were sampled as well. Prospective participants were sent a postal questionnaire consisting of the Swedish version of the 10-item AUDIT questionnaire [15,21]. In 2018, a 13-item DSM-5 short screener was added, in a Swedish translation approved by the originator [22]; however, the results from this screener are not reported herein but have been reported in a parallel paper [23]. The postal envelope also contained a letter briefly describing the study, indicating that participation was anonymous and voluntary, in addition to a stamped response envelope. As noted above, survey results from 1997 to 2014 have been reported elsewhere [18–20,24]. The Regional Ethical Review Board in Stockholm, Sweden approved the project (no 2018/184-32).

During the period of time between 1997 and 2018, the response rate decreased from 80% in 1997 to 71% in 2001, 72% in 2005, 50% in 2009, 60% 2014, and 47% in 2018. The proportion of men and women was approximately constant throughout all measurement points; 45–47% men and 53–55% women.

Instrument

The AUDIT was developed to assess risky and harmful alcohol habits, and alcohol-related problems rather than as a diagnostic instrument [21]. The AUDIT has been shown to have good validity and reliability in several reviews [10–12]. Each of the 10 items is scored with response alternatives between 0 and 4 points, resulting in a total score between 0 and 40 points, where a higher score indicates more harmful alcohol habits. In the original AUDIT, a standard drink was defined as containing 10 g of alcohol but this amount varies across countries. Higgins-Biddle and Babor [25] later adapted the original AUDIT to US standard drinks containing 14 g of alcohol. In the Swedish version of the AUDIT, different beverages containing a standard drink of 12 g of alcohol are depicted.

Statistics

All calculations were made using SPSS 24.0. Descriptive statistics included calculated means and standard deviations and a three-way ANOVA used to evaluate main effects and interactions between assessment year, gender and six age groups. Cohen’s d was calculated to report the effect size of differences in total AUDIT scores.

Results

In Table 1 and Figure 1, the development of Swedish general populations’ alcohol habits operationalized as total AUDIT scores is shown. Clearly, young men have lower total AUDIT scores since 2009 but young women do not show this trend. Scores tended to decrease across years, according to Scheffe’s test of multiple comparisons between 2001 and 2009 (p = .061) and also between 2001 and 2018 (p = .066). A tendency towards a main effect for assessment year was identified (see Table 2). Regarding 2014 versus 2018, no difference in total scores and no significant interactions occurred. The change in total AUDIT score between 2014 and 2018 for men in the youngest age group was small as indicated by a Cohen’s d of 0.28. Inspection of item-total correlations revealed that the importance of the first question of AUDIT (i.e. How often do you have a drink containing alcohol?) decreased over time from 0.36 in 1997 to 0.21 in 2014, and to 0.28 in 2018. The variation was less for the other items. The mean values for each question showed that there were marginal changes across time; however, there were relatively larger changes for the first AUDIT question that increased slightly over time.

Table 1. Mean total AUDIT score (standard deviations; number of respondents, N) at each measurement year by gender and age group.

Year/variable	1997	2001	2005	2009	2014	2018
Men	4.94 (4.67;471)	5.15 (4.90;420)	4.87 (4.80;409)	4.18 (4.18;276)	4.31 (3.71;344)	4.06 (4.22;278)
Women	2.74 (2.90;526)	3.19 (2.96;472)	3.15 (3.31;495)	2.96 (2.65;344)	2.97 (2.99;422)	2.96 (3.51;341)
Age 17–27	6.06 (5.47;184)	6.32 (5.41;154)	6.54 (5.53;154)	5.30 (5.11;105)	6.10 (5.06;68)	4.94 (5.29;87)
Age 28–38	3.88 (3.74;219)	4.38 (3.40;203)	3.69 (3.50;175)	3.64 (2.82;96)	4.16 (3.10;133)	3.48 (3.64;94)
Age 39–49	3.38 (3.11;230)	3.82 (3.40;169)	3.99 (4.25;191)	3.48 (2.90;130)	3.79 (3.86;127)	3.66 (3.15;107)
Age 50–60	3.21 (3.83;218)	3.68 (3.61;218)	3.09 (3.15;207)	2.94 (2.28;126)	3.20 (2.98;164)	3.76 (4.54;117)
Age 61–71	2.25 (2.43;146)	2.35 (2.54;137)	2.95 (3.38;152)	2.58 (3.11;151)	2.89 (2.65;189)	3.07 (3.44;129)
Age 72–80	N/Aa	N/A	N/A	N/A	2.43 (2.41;90)	2.17 (2.18;90)

^a Data for this age group was not collected in 1997–2009.

Figure 1. Mean total AUDIT scores for men (A) and women (B) over the years 1997 through 2018 separated by each age category, diamond (♦) 17–27 year olds, box (▪) 28–38 year olds, triangle (Δ) 39–49 year olds, cross (×) 50–60 year olds, plus (+) 61–71 year olds, and circle (•) 72–80 year olds.

Table 2. Results of three-way ANOVA showing main and interaction effects for assessment year, gender and age group, with F-statistics, degrees of freedom (Df) and p-values.

Variable	F	Df	p
Year	2.18	54,696	.054
Gender	175.46	14,696	<.001
Age group	62.90	54,696	<.001
Year * Gender	1.95	54,696	.083
Year * Age group	1.26	22,4696	.189
Gender * Age group	.987	54,696	.424
Gender * Year * Age group	1.14	21,4696	.297

Table 2 shows main effects for assessment year, gender and age group indicating that men, on average across the entire measurement period, score higher than women, with an overall average difference between age groups. Scheffe’s test of multiple comparisons shows that participants in the 17–27 age group score higher on average than the other age groups and those above 60 years old score lower on average than other groups (all p < .001). Additionally, participants aged 28–38 scored higher compared to 50–60 year olds (p = 0.012).

Discussion

This study described measurements of alcohol habits among Swedish men and women 17–80 years old between 1997 and 2018, with a focus on the years between 2014 and 2018. In short, the results showed no difference between 2014 and 2018 in terms of total scores but did show a slight decrease in AUDIT scores for men since 2001, while women remained on the same level. Our results also showed that for each question of the AUDIT, there were only marginal mean value changes across time apart from the first question that increased slightly over time. This could be interpreted such that people in recent years generally consume alcohol more often but in smaller quantities. Decreases also occurred among younger age groups (aged 17–27 and 28–38) and in the oldest age group (aged 72–80). A recent study found similar trends when measuring alcohol use disorder among the Finnish general population in an 11-year follow-up between 2000 and 2011 [26]. Here, it was found that alcohol use disorder had decreased the most among the younger age groups (i.e. 30–44 year old) included in the study [26]. For both gender and age, this could be described as a form of convergence where the groups with highest traditional consumption (men in younger age groups) reduce their consumption in the direction of women’s consumption levels. However, the original convergence hypothesis proposed women to increase their consumption to the level of men. As shown in Figure 1, changes in alcohol habits seemed to differ across gender and age groups. These changes across time were not significant and can therefore be taken as evidence for the predictions of the theory of collectivity of drinking cultures that changes in alcohol consumption appear in concert across subgroups. However, this mechanism could not be described statistically as a two- or three-way interaction and hence deserves a forthcoming study on its own. Since the propensity to respond has decreased over time in young men, the weight of the heaviest consumers in the average has decreased, and so has the overall average. As a consequence, the reduction in AUDIT scores may possibly be due to selective dropout of high consumers.

This study has some strengths and limitations. A major strength is that the design of the repeated surveys was in principle identical over the years. Another strength is the use of the psychometrically validated AUDIT instrument. However, the increasing problem with missing data due to low response rates is a significant factor that complicates interpretation of the results. It would have been desirable for nonparticipation to be examined in more detail, however, due to restrictions in data availability we could not address this issue. Further, for ethical reasons, we had no information about the respondents’ addresses, and hence, no information on the sampling distribution between urban and rural areas. Nevertheless, the missing data in this study and other studies in this field constitute a complication that will come to require some creative and novel methodological approaches in the future. This could include different combinations of data from wastewater, internet searches and traditional surveys such as this one. Finally, the tendency to interaction between year and gender support the ‘convergence’-hypothesis but this is mainly due to decreasing scores across years by men but not by women.

Notes on contributors

All authors helped to plan and design the study. HK and TE made the first draft of the manuscript. All authors iterated the manuscript and approved final version.

Disclosure statement

H.K., A.H.B. and P.W. are the authors of a Swedish-language book (manual) for the AUDIT and DUDIT. P.W. coordinated, and A.H.B. participated in, evaluation of assessment instruments including the AUDIT for the 2015 National Board of Health and Welfare revision of National Guidelines. The authors have no other conflicts of interest to declare. The research also complies with ethical guidelines in Sweden for studies including humans. Such a statement is mentioned in the Methods section above.

Additional information

Funding

This work was supported by grants from the Swedish Research Council for Health, Working life and Welfare (grant number 2009-1705), and the Alcohol Research Council of the Swedish Alcohol Retailing Monopoly and Forskningsrådet om Hälsa, Arbetsliv och Välfärd.