Study Design and Sample

We used a longitudinal design based on data from waves 2 and 4 (2006–07 and 2011–12, hereinafter called 2006 and 2012) of the Survey of Health, Ageing and Retirement in Europe (SHARE) project [33,34], for 11 European countries that participated in both waves: Austria, Belgium, Czech Republic, Denmark, France, Germany, Italy, Netherlands, Spain, Sweden and Switzerland. Each wave of the SHARE survey consisted on a face-to-face computer-assisted interview (CAPI), supplemented by a self-administered questionnaire. The survey of each wave covered economic, social and health factors that accompany and influence ageing processes (for further details on the survey methodology, see www.share-project.org [35–37]). Data from waves 1 and 3 were not included in the study as the alcohol questions in those two waves were different and not comparable. Poland participated in the SHARE project, but its data were excluded from the study to avoid that its extreme values on both prevalence of hazardous drinking and contextual variables could drive the results. Sampling was performed independently in each country; all countries obtained a probabilistic sample, although the exact sample design differed slightly between countries. The inclusion criteria for this study were being 50- to 64-years old and not being a hazardous drinker at baseline (2006–2007). People with no data on alcohol consumption at baseline were excluded (1.5%). The final sample size was of 7,615 individuals.

Dependent variable

The dependent variable was the incidence of hazardous drinking. To calculate individual daily alcohol consumption, we combined individual’s responses to questions on drinking frequency during the previous 3 months (“How often did you drink any alcoholic beverages, like beer, cider, wine, spirits or cocktails?”) and typical quantity per occasion (“On the days you drank during the past three months, about how many drinks do you have?”). Hazardous drinking was defined as an average daily consumption of ≥40g and ≥20g of pure alcohol during the previous 3 months, in men and women, respectively [38,39]. The incidence of hazardous drinking was calculated using answers given by participants during visits at baseline (2006–07) and follow-up (2011–12). Participants who said, at baseline, that they had not consumed alcohol during the three months prior to the interview were classified as abstainers, and the ones who drank but were not hazardous drinkers were classified as low-risk drinkers.

Independent variables

The individual-level independent variables were: sex, age, country of residence, employment status [employed; unemployed; other (retired, homemaker, sick or disabled)], changes in employment status (employed at baseline and follow-up; employed at baseline, unemployed at follow-up; unemployed at baseline, employed at follow-up; other status), educational level (lower secondary education or less; upper secondary education or more) and self-perceived health (good, very good or excellent; fair or poor).

The country-level independent variables were: a) changes in unemployment rate between 2006 and 2010 (dichotomous: decrease, increase) [40]; b) percentage of increase in Gross Domestic Product (GDP) between 2006 and 2010 (continuous) [41]; c) percentage of increase in social protection expenditure as a percentage of GDP between 2006 and 2010 (continuous) [41]; d) percentage of increase in household disposable income between 2006 and 2010 (continuous) [40]; e) standardized Gini index for 2007 (continuous) [41]; f) changes in Gender Inequality Index between 2005 and 2010 (dichotomous: decrease, increase) [42]; g) standardized degree of alcohol advertising restriction, 2008 (continuous) (S1 Table) [43,44]; h) other standardized alcohol control policies, 2006 (continuous) (S2 Table) [45]; i) drinking patterns score, 2005 (dichotomous: low-risk drinking patterns, risk drinking patterns) [43]. For country-level variables, we used data from 2006 and 2010 where available for all participating countries. However, where no information was available for either year, we used data from the nearest year.

Ethics Statement

During waves 1 to 4, SHARE has been repeatedly reviewed and approved by the Ethics Committee of the University of Mannheim. In addition wave 4 was reviewed and approved by the Ethics Committee of the Max Planck Society in 2012. All information in SHARE is pseudo-anonymised and therefore the identification of individual persons is not possible. All respondents have been informed about the storage and use of the data and about their right to withdraw their consent. Written consent was given by the respondents for their information to be stored in the database and used for research when required by national or regional data protection laws.

Statistical analysis

The sample distribution at the beginning of the follow-up period was calculated for each individual variable. Follow-up bias was assessed among the 7,615 individuals at baseline. The cumulative incidence of hazardous drinking between 2006 and 2012 was estimated for each individual-level independent variable. To test for association between the incidence of hazardous drinking and changes in employment status, while accounting for individual- and country-level variables, we fitted multilevel Poisson regression models with robust variance, which yielded risk ratios (RR) and 95% Confidence Intervals (95%CI). All analyses were adjusted for length of follow-up (i.e. time between the two interviews), which was measured in years. First, we estimated the variability of the incidence of hazardous drinking between countries (empty model: Model 0). Second, we fitted several bivariate Poisson regression models with robust variance, including in each model the dependent variable (incidence of hazardous drinking) and one independent variable (Model 1). The final adjusted model (Model 2) was fitted by including all individual- and country-level variables and eliminating, one at a time, each of the contextual variables that were not statistically significant. Thus, the final adjusted model included all individual variables, as well as the country-level variables that were statistically significant in the multivariate analysis. Statistical analyses were conducted using STATA 13.0 and HLM6.

Strengths and Limitations

Both, individual (e.g. changes in participants´ employment status) and country level factors (e.g. unemployment rate) may be affected by an economic crisis and both should be taken into account to better analyze changes in alcohol consumption during an economic downturn. The main strength of our study is the consideration of the two types of factors in the analysis, which is not usual in this research field [46]. Our study presents some limitations. Firstly, we could not determine whether changes in employment status lead to hazardous drinking or vice versa, since there were no data on the date of changes in employment status, and data on drinking only capture the three months prior to the follow-up interview. Secondly, 36.5% of individuals did not have a follow-up interview and were excluded from the study, which indicates a potential differential follow-up bias. In the follow-up bias assessment, we found some differences between those who were lost to follow-up and those who were not for some individual-level variables, but not for employment status (Table 1). We analysed each cross-sectional survey separately (incl. all participants, regardless of follow-up status), and observed no statistically significant difference between baseline and follow-up in the prevalence of hazardous drinking in these variables. Moreover, our conclusions are further supported by the fact that the risk of hazardous drinking among unemployed compared to employed participants was 1.42 (95%CI: 1.13–1.80) in the entire 2011–12 sample, and just 1.19 (95%CI: 0.88–1.60) in the 2006–07 sample. Thirdly, the use of individual self-reported data might have led to underreporting bias in some population groups, such as unemployed hazardous drinkers who wish to avoid a stigma that would hinder job searching [19]. Fourthly, the incidence of hazardous drinking could be underestimated because of survival bias: individuals with a history of hazardous drinking but who gave up drinking following development of a serious alcohol-related illness are more likely to die during follow-up than those without serious illnesses. Fifthly, while we controlled for various individual-level factors related to alcohol consumption, we were not able to account for participants’ previous consumption of alcohol or other substances due to lack of data. We only know that the participants were not hazardous drinkers in the three months prior to the baseline interview. Individuals with history of alcohol abuse or dependence may be at higher risk of relapse when dealing with stressful situations [47]. The risk of relapse among former hazardous drinkers is potentially higher after losing a job during economically challenging times. Predictably, the availability of data on previous risk patterns among participants would have reinforced our results. Finally, Poland was excluded from the analysis because it had a very low incidence of hazardous drinking with respect to other countries and its contextual variables’ values were also quite different. However, a sensitivity analysis including Poland provided similar results at the individual level.

Changes in alcohol consumption during the late 2000s economic crisis

During the 6-year follow-up period, cumulative incidence of hazardous drinking among middle-aged people was 6.6 per 100 persons, which may correspond to an annual incidence of 1.1 cases per 100 person-years assuming that the incidence was constant over the period of study. This incidence was similar to the one found by Kalousova et al. [19] in a study of 19-64-years-old people from the U.S.: after 2 years of follow-up, 4% of individuals engaged in hazardous drinking (i.e. an annual incidence of 2 cases per 100 persons). Future studies should focus on the risk of initiating a risky pattern of drinking, as it is related to adverse health consequences [48,49]. The evolution towards risky patterns among middle-aged deserve the scaling up of prevention policies specifically in this age group [49]. Our findings show that almost 7% of people became hazardous drinkers during the follow-up period, which is of concern not only because of the harmful health effects, but also because negative health behaviours established or re-established during economic difficulties can persist afterwards and become a life-long habit [19].

We observed that job loss in middle-aged people is a risk factor for hazardous alcohol consumption when comparing to employed people, which is consistent with several studies in active population of all ages [17,18,23]. The analysis including country employment rates and other variables, suggest that this individual factor is not confounded by important variables acting at contextual level. Our finding implies that losing the job may also have an effect on health of the group of economically active people over 50.

Individuals aged 50–65 years face a double vulnerability: social and physical. Firstly, a lower probability of being reemployed than younger ones [30,31] would make individual job loss a relevant factor leading to very important levels of strain or hardships for affected middle-aged individuals during economically challenging times in a context of generalized job insecurity. The diminished expectations of finding a new job would increase their stress, which might raise the risk of increasing the intake of alcohol. Studies on mental health found that job loss is also associated with worsening of mental health in all age groups [4,50]. The stress hypothesis is a likely explanation for our findings [23,51,52], although from our data no causal relationship can be established between being deprived of employment and the increase in alcohol intake. Our finding that the risk of becoming a hazardous drinker increases with age, such that older individuals who become unemployed may be most likely to start to drink hazardously is in line with the decreasing possibilities of finding a job as age grows [30,31]. However, this hypothesis requires validation, and caution is needed when interpreting this finding because our results on the role of age may be sensitive to a differential loss to follow-up.

Secondly, when considering the same intake of alcohol, middle-aged and older individuals are more vulnerable to adverse health effects than younger individuals due to age-related changes [26,27]. Middle-aged people and older adults are more likely than younger groups to suffer health problems that can get worsened by risky drinking patterns, as stroke, high blood pressure, diabetes, or mood disorders [28]. In addition, the increased likelihood of health problems among people over 50, also multiplies the possibility of interaction with medication. Non-fatal injuries have been found to be associated with frequency of drinking and problem drinking in a wide sample of 45–69 year old European men and women [53] and over 2/3 of all alcohol-attributable deaths occurring among the 20–64 year old population of the European Union occur in the 45–64 years olds [49]. There is an increasing public health concern regarding the harmful consequences of alcohol among people from middle age onwards because of the rapid ageing population in many countries worldwide [49,54]. Our finding support the conclusion by Rehm et al. (2010) of having middle-aged as a priority group for intervention and policy [49].

Finally, we observed differences in the cumulative incidence of hazardous drinking among countries. Although these differences were not statistically significant, they varied from 5.1% in Germany to 8.2% in Belgium. The highest absolute incidence observed in Belgium could be related to the fact that prevalence of alcohol consumption in this country, and especially of hazardous drinking, is among the highest of the participating countries, as observed in previous studies [44,55].

Although the economic crisis did not have the worst effects on Belgium, other participating countries which suffered the most severe effects of the economic crisis such as Spain or Italy were among those with the highest incidence of hazardous drinking. In this sense, as we observed in the multilevel analysis, there was an association between the incidence of hazardous drinking and the national unemployment rate and the household disposable income. Both variables were correlated with hazardous drinking (changes in unemployment rate: r = 0.420; increases in household disposable income: r = -0.561). While associations found at contextual level should be interpreted with caution due to the low number of countries in the second level of the multilevel analyses, our results are consistent with previous studies that highlight the relationship between declining macroeconomic conditions and alcohol-related risky behaviours [11]. A possible explanation may be that increased national unemployment rate could generate real or perceived economic insecurity among workers, increasing stress and possibly alcohol consumption [9]. We also found that a subtle increase in household disposable income was associated with higher cumulative incidence of hazardous drinking. This supports the hypothesis that poor macroeconomic conditions might promote risky drinking patterns.