The relationship between socioeconomic status and risky drinking in Denmark: a cross-sectional general population study

With regard to the specific health behaviour of alcohol use, research has shown that individual SES (measured by all components: education, income, and occupation) is associated in various ways with risky alcohol consumption [13‐15]. In the same manner that individual health risk factors such as smoking and sedentary lifestyles are more prevalent among lower SES groups [14, 16, 17], those with low SES are also more likely to suffer from alcohol-related morbidity and mortality [18‐20]. However, regarding drinking patterns, people with low income have been found to be more likely to be either abstainers or heavy drinkers, and less likely to be moderate drinkers [21] or to drink frequently [22]. Thus, research has shown that the social gradient does not always follow in the same direction. Some findings from higher income Western countries indicate that those with higher education and/or income can drink at harmful levels (e.g. [23]).

One way to further investigate the link between SES and drinking is by assessing the different potential relationships from the various components of SES to drinking patterns. So far, research that attempts to do this is rare, but one such study that used a variety of data sets from UK and USA found a strong mediation effect of income on the relationship between education and health behaviours [24]. Specifically, income was found to reduce the effect of education on current smoking by 26% and on heavy drinking by 12%, indicating that income mediated the effect of education on smoking and drinking.

In general, people aged 25 years or older have completed their education, thus this adult age group is an appropriate population in which to study the effects of education as a stable component of SES [3, 24]. However, the links between education and employment status are more complicated, as type of employment may influence health through the prestige associated with the type of job, which, in turn, may again vary between countries and may have changed over time [3]. Education is considered in this study to be an independent component of adult SES (i.e., influencing income or job status), and income or job status are then tested as potential mediators that more directly influence drinking behaviour.

To shed new light on the links between SES and alcohol indicators, the present study specifically addresses the following research questions: (a) Is education associated with risky single occasion drinking (RSOD) in Denmark? (b) Does income or employment mediate the effects of education on RSOD? We analyse general population survey data from Denmark to address this gap in the alcohol research literature. We examine a recognised indicator of possible problematic drinking, risky single occasion drinking (RSOD), as our outcome variable [25].

Data came from the 2011 national survey of alcohol and drugs conducted by Statistics Denmark for the Centre for Alcohol and Drug Research of Aarhus University. A representative sample of 8000 persons between the ages of 15–79 years old was randomly drawn from the central person registration (CPR) numbers. Upon birth or immigration to the country, each resident is assigned a unique registration number, which is used for official record keeping. Potential respondents (i.e., the 8000 potential cases) were invited by postal letter to complete a web questionnaire during September and October 2011. Telephone interviews were conducted with those individuals who had not responded after two reminders. The interviewed sample consisted of 5133 respondents representing a response rate of 64%. The Danish Data Protection Agency approved the survey and all confidentiality and privacy requirements were met. Registry data from Statistics Denmark were used to include information about respondents’ years of schooling and disposable income. The survey’s age range was restricted to correspond roughly to the end of tertiary education and at the high-end age for the start of retirement (26–67 years) [24]. This resulted in a final sample of 3600 individuals for the analyses (see Fig. 1).

To measure alcohol consumption respondents were first asked whether they had drunk any alcohol in the past 12 months or not. In a separate question, current drinkers were asked about their frequency of RSOD during the previous 12 months. RSOD was defined as drinking five or more units of alcohol (equivalent to 60 g of alcohol or more) on a single occasion. Respondents had a choice of different frequencies, which we coded as one if respondents reported any RSOD at least once per month in the last 12 months or otherwise as zero.

The main independent variable of interest was SES, specifically its three components education, income, and employment status. As a measure of education, we used years of schooling. We measured income using personal disposable income, which is the summation of pre-tax income and imputed rent minus interest expenses, tax, and paid alimonies. In the analysis, its logarithmic form was designed to reduce skewness. Data on education and personal disposable income were obtained from Statistics Denmark and merged with the survey data. The survey asked respondents to provide their employment status by choosing from given categories in the questionnaire. For the analyses, we coded one when respondents were employed and zero otherwise.

The sociodemographic variables used in the analysis include age groupings (26-35, 36-45, 46-55, and 56 years and older), gender and civil status (in a relationship i.e. married and/or living with a partner, not in a relationship i.e. divorced/separated/widowed, and never married). We also created a binary indicator of whether the respondent was living with children under the age of 18 years. Our indicator of religiosity equals one if the respondent reported attending worship or religious ceremonies more than four times in the last 12 months. Finally, we included area of residence (coded as capital, other large cities, rural).

The pairwise correlation between RSOD and each covariate was examined using the Kruskal-Wallis equality-of-populations rank test and Spearman’s correlation test. To examine the association between components of SES and RSOD, we used multiple logistic regression. To avoid small cell frequencies, we dichotomized RSOD (at least monthly RSOD in the past year versus all other). In the analyses, we followed three steps. In the first step, the relationship between education and RSOD was analysed without including the other components of SES while controlling sociodemographic variables. In the second step, we added income to the model. In the third step, we added employment status.

In all analyses, we controlled age, gender, marital status, living with children under the age of 18, religiosity, and area of residence, as these variables appear from the literature to be key sociodemographic factors affecting drinking. We used weighted data, using weights, which were created by Statistics Denmark and are based on national distributions of age, sex, family structure, education, income, and country of origin. We conducted all analyses with STATA 14.0 software [26].