Predictors and prevalence of hazardous alcohol use in middle-late to late adulthood in Europe

Abstract

Objectives: Even low to moderate levels of alcohol consumption can have detrimental health consequences, especially in older adults (OA). Although many studies report an increase in the proportion of drinkers among OA, there are regional variations. Therefore, we examined alcohol consumption and the prevalence of hazardous alcohol use (HAU) among men and women aged 50+ years in four European regions and investigated predictors of HAU.

Methods: We analyzed data of N = 35,042 participants of the European SHARE study. We investigated differences in alcohol consumption (units last week) according to gender, age and EU-region using ANOVAs. Furthermore, logistic regression models were used to examine the effect of income, education, marital status, history of a low-quality parent–child relationship and smoking on HAU, also stratified for gender and EU-region. HAU was operationalized as binge drinking or risky drinking (<12.5 units of 10 ml alcohol/week).

Results: Overall, past week alcohol consumption was 5.0 units (±7.8), prevalence of HAU was 25.4% within our sample of European adults aged 50+ years. Male gender, younger age and living in Western Europe were linked to both higher alcohol consumption and higher risks of HAU. Income, education, smoking, a low-quality parent–child relationship, living in Northern and especially Eastern Europe were positively associated with HAU. Stratified analyses revealed differences by region and gender.

Conclusions: HAU was highly prevalent within this European sample of OA. Alcohol consumption and determinants of HAU differed between EU-regions, hinting to a necessity of risk-stratified population-level strategies to prevent HAU and subsequent alcohol use disorders.

Keywords:

• Hazardous alcohol use
• older adults
• middle-aged adults
• Europe
• alcohol
• drug and alcohol abuse
• cross-national/international studies
• environmental factors/housing/rural-urban factors
• epidemiology (mental health)

Introduction

Although the proportion of drinkers in people aged 15+ years slightly decreased worldwide, overall per capita consumption (including abstainers) of pure alcohol rose from 5.7 l in 2000 to 6.4 l in 2010 and 2016 (World Health Organization, 2018). With each unit containing 10 ml of pure alcohol, this equals an increase from 570 to 640 units a year and results in one glass of wine (125 ml) or a bottle of beer (330 ml) each day on average. This is concerning, because even low to moderate levels of alcohol consumption were linked to serious health consequences like cardiovascular diseases (Wood et al., 2018), cancer (Bagnardi et al., 2015) and all-cause mortality (Wood et al., 2018), and heavy drinking increases the risk for aggressive behavior (Heinz et al., 2011) as well as cognitive decline and dementia (Livingston et al., 2020; Rao et al., 2021).

Alcohol use changes over the life course and frequent drinking becomes especially prevalent in mid to older age (Britton et al., 2015). For adults aged 50+ years, studies report increasing prevalences of drinking over the past years (Bye & Moan, 2020; Calvo et al., 2020; Grucza et al., 2018). Calvo et al. (2020) report a decrease in the proportion of abstainers aged 50+ years in 13 of 21 examined countries from 1998 to 2016. Although the picture is less clear, findings also indicate an increase in the prevalence of harmful alcohol use among older adults (OA) in several countries, like binge drinking or heavy episodic drinking (Calvo et al., 2020; Han & Moore, 2018). For example, Grucza et al. (2018) observed an increase of 3.4% per year (2000–2016) in binge drinking in the US among OA (65+), and in Norway, the proportion of OA (60+) reporting any heavy episodic drinking within the last year rose from 17% in 1985 to 30% in 2017 (Bye & Moan, 2020). Calvo et al. (2020) report an increase of heavy drinking among OA in Austria, China, Czech Republic, Denmark, Germany and the United States, but at the same time, prevalence decreased in Belgium, France, Italy, Mexico, Netherlands, Poland and Spain; for the remaining eight countries included in this study, rates of heavy drinking remained relatively stable. This underlines the wide variations of alcohol use among OA found across countries, which might be due to religious or cultural differences, alcohol policies (Angus et al., 2019; Calvo et al., 2021; Sudhinaraset et al., 2016) or other factors that affect alcohol consumption according to previous studies. For example, male gender (Bloomfield et al., 2006; Devaux & Sassi, 2016) and smoking (Paula et al., 2022; Wang et al., 2022) are associated with higher alcohol intake or risky drinking among OA, while for marital status findings are more complex and vary by gender or region (Gell et al., 2015; Iparraguirre, 2015). Higher education is associated with higher likelihood of current drinking, irrespective of gender (Bloomfield et al., 2006; Grittner et al., 2013). For binge drinking or drinking larger amounts of alcohol on a regular basis, the pattern is less clear and differs between genders and regions (Bloomfield et al., 2006; Devaux & Sassi, 2016; Grittner et al., 2013). Additionally, there is some evidence for a protective effect of a good parent–child relationship on adolescent (Visser et al., 2012) and (early) adulthood alcohol use (Boden et al., 2021; Mak & Iacovou, 2019), but less is known about this link in OA. Since adolescent drinking behavior has been associated with alcohol use in adulthood (Cable & Sacker, 2008; Jefferis et al., 2005), such an effect of parent–child relationship could potentially extend to later ages.

The trend of increasing prevalences of alcohol consumption in OA is concerning, since drinking can be especially harmful for older people. First, OA react more sensitive to alcohol due to age-related physiological changes (Barry & Blow, 2016; Oneta et al., 2001). Second, medication use and polypharmacy are common within OA and the combination of drugs and alcohol can be particularly harmful (Cousins et al., 2014; Han & Moore, 2018). Third, the effect of alcohol related injuries and falls can be detrimental for OA, as falls tend to have worse consequences for older compared to younger people (Chatha et al., 2018) and the additional influence of alcohol can lead to worse injuries (Shakya et al., 2020).

Europe is the region with the highest proportion of current drinkers and the highest per capita consumption worldwide (World Health Organization, 2018). However, since alcohol consumption and the influence of its contributing factors vary regionally, partly because of differences in alcohol policies, taxation, country-level SES or social and cultural contexts (Angus et al., 2019; Grittner et al., 2013; Sudhinaraset et al., 2016), it is likely that prevalences and predictors of alcohol use differ regionally. Therefore, we aimed to examine and compare alcohol consumption and the prevalence of hazardous alcohol use (HAU) as well as its predictors among men and women aged 50+ years in four European regions (Northern, Eastern, Southern, Western Europe).

Materials & methods

Data were drawn from the longitudinal Survey of Health, Aging and Retirement in Europe (SHARE) (Börsch-Supan et al., 2013) in the sense of a retrospective cohort study. SHARE collects data on health, SES, and social networks of individuals aged 50+ years and their partners. Nearly 140,000 participants from 28 European countries and Israel participated between 2004 (Wave 1) and 2017 (Wave 7) (Bergmann et al., 2019). SHARE applies probability-based sampling. Where possible, sampling is based on an official population register. If such a register is not available, the best alternative sampling frame resource is used (i.e. telephone directory and screening procedure). Our analysis is based on Wave 6 of SHARE and uses an additional predictor from Wave 7 (see Figure 1; for more details, see next section). We grouped countries to EU-regions according to the United Nations Statistics Divisions’ list of geographic regions (United Nations Statistics Division, n.d.): Northern Europe (Denmark, Estonia, Sweden), Eastern Europe (Czech Republic, Poland), Southern Europe (Croatia, Greece, Italy, Portugal, Slovenia, Spain) and Western Europe (Austria, Belgium, France, Germany, Luxembourg, Switzerland). We did not include participants from Israel.

Figure 1. Flow chart of the study sample. Note: * Some participants’ partners were additionally interviewed, regardless of their age.

Measurements

Most variables, including information on alcohol consumption within the last week and binge drinking, were taken from Wave 6 of SHARE in 2015 (Börsch-Supan, 2018). Parent–child relationship quality (concerning participants’ own childhood) was obtained from SHARElife Wave 7 (2017) (Börsch-Supan, 2019).

Outcome variables

Current alcohol consumption was indicated by alcohol units (10 ml of pure alcohol) consumed in the past seven days. It was assessed via self-reported consumption of standard drinks in the past week with approximately 1.4 units of pure alcohol per drink (Alcohol Units, 2018).

We defined risky alcohol consumption, according to Wood et al. (2018), as the consumption of >12.5 units alcohol per week (7.5 bottles of beer (330 ml) or six glasses of wine (175 ml).

Binge drinking was defined as consuming six or more standard drinks on at least one occasion within the last three months (World Health Organisation, 2014).

We compared the original data on number of drinks during the past week as well as binge drinking, and checked for plausibility and missing data. Where possible, we derived the information from the other variable if either was missing. If both items were answered, but substantially contradicted each other, we used the higher value, because social desirability more likely leads to underreporting of alcohol consumption.

We defined hazardous use of alcohol (HAU) as either binge drinking at least once in the past 3 months or risky alcohol consumption in the past week (>12.5 units/week), as both drinking habits carry the risk of adverse health consequences in adults (Wood et al., 2018). Since Wave 6 interviews contained no global question on current drinking behavior, but only asked for binge drinking and alcohol consumption during the past week, it was impossible to identify abstainers.

Other variables

We categorized age into four categories (50–59, 60–60, 70–79 and 80+ years).

Income per capita (€) was calculated by dividing the total annual household net income by the number of household members. Marital status, Educational years (full-time) and Smoking status are self-reported information. A positive smoking status was present when participants ever smoked cigarettes, cigars, cigarillos or a pipe daily for a period of at least one year.

We extracted information on the relationship quality to participants’ parents or guardian during childhood and adolescence. We dichotomized the available categories in order to contrast the lowest quality relationships (‘poor’ relationship with mother/father/both) against higher quality relationships (‘excellent’ to ‘fair’ relationships).

Statistical analyses

We replaced missing values in marital status, income, years of education and smoking status by multiple imputations provided within the SHARE dataset (de Luca, 2018; de Luca et al., 2015). Participants with an alcohol consumption (units last week) that exaggerated the group mean by more than three standard deviations were considered outliers and were excluded. All analyses were performed using SPSS27. All tests were performed two-sided and considered significant at p < .05. Income was transformed using a natural logarithm to improve the distribution characteristics.

We conducted a 2 × 4 × 4 factorial analysis of variance (ANOVA) to analyze main effects of gender, age and EU-region on alcohol consumption. Due to variance heterogeneity in our ANOVA data, we compared our results to robust parameters based on a regression model (Hayes & Cai, 2007). Post hoc tests were conducted via Fisher’s Least Significant Difference (LSD). A hierarchical binary logistic regression analysis was used to investigate predictors of HAU. All full regression models were stratified by gender and EU-region, stratified analyses were corrected using the Bonferroni–Holm method (Holm, 1979).

We performed collinearity diagnostics for the regression models by analyzing the Variance Inflation Factors (VIF) and the tolerance statistics, which did not reveal any multicollinearity for our variables (all VIF < 1.16, average VIF = 1.084, all tolerance scores > .86).

Results

The distribution of participants and their age, gender, living status, income and education by EU-regions are presented in Table 1.

Table 1. Sample characteristics.

	Age (years)	N (%)	Female (%)	Married, living with spouse (%)	Total income per capita (€; M (SD))	Years of education (M (SD))
Overall sample	50 − 59	10,707 (30.6)	59.1	72.1	15,436.50 (17,476.12)	12.20 (4.10)
	60 − 69	12,122 (34.6)	55.4	73.0	16,044.74 (26,082.30)	11.43 (4.21)
	70 − 79	8490 (24.2)	56.9	65.9	13,559.83 (17,616.37)	10.38 (4.40)
	80+	3723 (10.6)	60.4	46.9	12,916.60 (49,260.35)	9.19 (4.49)
	overall	35,042	57.4	68.2	14,924.50 (25,747.88)	11.17 (4.36)
Southern Europe	50 − 59	3588 (30.0)	60.5	79.2	6906.98 (6977.21)	11.09 (4.00)
	60 − 69	4180 (34.9)	54.1	80.5	8013.71 (7892.30)	10.30 (4.14)
	70 − 79	2867 (24.2)	55.2	71.3	7644.20 (6855.50)	8.63 (4.40)
	80+	1327 (10.6)	61.6	50.3	8884.13 (77,903.57)	7.37 (4.45)
	overall	11,962	57.1	74.6	7689.74 (26,847.90)	9.81 (4.38)
Western Europe	50 − 59	3978 (33.8)	58.1	68.3	24,499.93 (21,828.36)	12.38 (4.46)
	60 − 69	4012 (34.1)	54.9	69.8	28,658.39 (39,995.21)	11.48 (4.70)
	70 − 79	2634 (22.4)	54.6	66.7	25,034.92 (25,670.31)	10.78 (4.53)
	80+	1154 (9.8)	59.2	45.8	23,359.81 (25,348.84)	9.90 (4.40)
	overall	11,778	56.3	66.3	25,924.38 (30,331.28)	11.47 (4.62)
Northern Europe	50 − 59	2310 (28.4)	57.7	66.6	16,969.43 (14,716.57)	13.53 (3.41)
	60 − 69	2695 (33.1)	56.0	68.6	14,771.50 (13,586.84)	12.65 (3.45)
	70 − 79	2182 (26.8)	60.0	61.7	10,792.79 (10,345.09)	11.57 (3.79)
	80+	948 (11.7)	61.0	44.1	8425.11 (7659.97)	10.04 (3.92)
	overall	8135	58.1	63.3	13,588.87 (12,925.23)	12.31 (3.75)
Eastern Europe	50 − 59	831 (26.6)	62.3	74.7	4616.51 (3399.93)	12.43 (3.04)
	60 − 69	1235 (39.0)	60.4	67.4	5028.61 (2832.40)	12.49 (3.05)
	70 − 79	807 (25.5)	61.6	55.9	4603.67 (1822.50)	12.06 (3.30)
	80+	294 (9.3)	57.8	44.2	4609.03 (1846.87)	11.83 (3.75)
	overall	3167	61.0	64.2	4773.25 (2712.99)	12.30 (3.19)

Effects of EU-region, age and gender on alcohol consumption

Mean units of alcohol consumption in the last week by EU-region, age and gender are displayed in Figures 2 and 3b and Table 2. There was a statistically significant main effect of age (brackets) on the units of alcohol consumed (F_3.35042 = 95.88, p < .001, $\eta$_p² = .008). Post hoc tests revealed significantly less units of alcohol in older compared to younger age brackets (all p < .001), except for higher alcohol consumption in the age bracket 60–69 years compared to age bracket 50–59 years (p = .126). However, this effect is due to a differential gender distribution in the age brackets and disappears in gender-stratified analyses (see Table 2). EU-regions also showed a significant main effect on units of alcohol consumed (F_3.35042 = 294.64, p < .001, $\eta$_p² = .025). Post hoc tests revealed differences in alcohol consumption between all four EU-regions (all p < .001), except for the comparison between Southern Europe and Northern Europe (p = .279) and highest alcohol consumption in Western Europe. Furthermore, there was a significant main effect for gender on consumed alcohol units (F_1.35042 = 3577.30, p < .001, $\eta$_p² = .093) with lower alcohol consumption in female participants.

Figure 2. Alcohol consumption in men and women by EU-region and age (Overall sample: N = 35,042).

Figure 3. (a) Alcohol consumption by EU-region and age in women (N = 20,125). (b) Alcohol consumption by EU-region and age in men (N = 14,917).

Table 2. Alcohol consumption in middle-late to late adulthood in Europe.

	Overall sample	Gender		Southern Europe		Western Europe		Northern Europe		Eastern Europe
	Women	Men	Women	Men	Women	Men	Women	Men	Women	Men
N	35,042	20,125	14,917	6833	5129	6634	5144	4727	3408	1931	1236
50–59 years.	5.4 (8.0)	3.4 (5.8)	8.3 (9.7)	2.1 (4.7)	7.5 (9.3)	4.6 (6.6)	9.3 (10.1)	3.6 (5.6)	7.6 (8.9)	3.5 (5.9)	8.5 (9.5)
60–69 years.	5.6 (8.2)	3.3 (5.9)	8.4 (9.7)	2.1 (4.8)	6.9 (9.2)	4.8 (7.1)	10.4 (10.4)	3.0 (5.2)	6.6 (8.1)	3.0 (5.2)	10.9 (10.8)
70–79 years.	4.6 (7.6)	2.6 (5.2)	7.2 (9.4)	1.7 (4.2)	6.5 (9.3)	4.0 (6.2)	9.1 (9.7)	1.9 (4.5)	5.1 (7.5)	2.9 (5.5)	9.1 (10.1)
80+	3.2 (6.3)	1.7 (4.4)	5.5 (7.9)	1.1 (3.4)	4.4 (7.0)	3.1 (6.2)	8.0 (9.0)	0.8 (2.5)	3.4 (6.6)	1.5 (3.5)	6.1 (8.4)
Overall sample	5.0 (7.8)	3.0 (5.6)	7.8 (9.5)	1.9 (4.5)	6.7 (9.1)	4.4 (6.7)	9.5 (10.0)	2.6 (5.0)	6.2 (8.2)	3.0 (5.4)	9.4 (10.2)

Note. Alcohol units last week, mean (standard deviation).

Prevalence of hazardous alcohol use

Prevalences of HAU (binge drinking in the past three months or risky alcohol consumption (>12.5 units/week) by gender, age and EU-region are shown in Table 3. All categories (gender, age, EU-region) showed significant differences at p < .001 (chi-squared tests).

Table 3. Hazardous alcohol use in middle-late to late adulthood in Europe.

	Overall sample	Gender		Southern Europe		Western Europe		Northern Europe		Eastern Europe
	Women	Men	Women	Men	Women	Men	Women	Men	Women	Men
N	35,042	20,125	14,917	6833	5129	6634	5144	4727	3408	1931	1236
50–59 years.	30.1	20.1	44.5	9.7	33.1	24.5	48.0	27.7	52.5	25.1	53.0
60–69 years.	28.0	16.6	42.1	9.4	31.4	21.0	48.5	19.8	43.4	19.3	57.5
70–79 years.	21.1	12.2	32.8	6.8	26.0	16.6	40.1	13.4	29.2	13.7	43.2
80+ years.	13.1	6.8	22.6	2.7	15.7	11.6	31.0	6.1	18.6	10.0	32.5
Overall sample	25.4	15.6	38.6	8.1	29.0	20.3	44.8	18.6	39.7	18.6	50.2

Note. Binging in past 3 months or risky alcohol consumption (>12.5 units/week), in %.

Predictors of hazardous alcohol use

The hierarchical binary logistic regression models for hazardous versus non-hazardous alcohol use revealed statistically significant odds ratios (OR) for the following variables or categories (see Table 4, Model 1): female gender (OR = .318, CI = .301–.336, p < .001), age categories 50–59, 60–69, 70–79 (reference category 80+, all ps < .001), education (OR = 1.035, CI = 1.029–1.041, p < .001), annual household income (OR = 1.137, CI = 1.116–1.158, p = .008), ‘married, not living with spouse’ and ‘divorced’ (reference category ‘married, living with spouse’; OR=.1.392, CI = 1.113–1.741, p = .004 and OR = 1.093, CI = 1.001–1.194, p = .046, respectively), history of a low-quality parent–child relationship (OR = .894, CI = .804–.995, p = .041), and smoking (OR = 1.487, CI = 1.411–1.567, p < .001).

Table 4. Hierarchical binary logistic regression: Predictors of hazardous versus nonhazardous alcohol use.

	Model 1		Model 2		Women – Model 2		Men – Model 2
Nagelkerkes R^2	.145		.164		.084		.071
Variable	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
Female gender	0.318***	0.301 − 0.336	0.309***	0.292 − 0.326	–	–	–	–
Age (Ref. 80+ years.)
50–59 years.	2.469***	2.201 − 2.770	2.553***	2.274 − 2.867	2.329***	1.922 − 2.823	2.688***	2.322 − 3.111
60–69 years.	2.155***	1.926 − 2.411	2.232***	1.993 − 2.499	1.952***	1.615 − 2.358	2.435***	2.113 − 2.806
70–79 years.	1.600***	1.425 − 1.797	1.627***	1.448 − 1.829	1.562***	1.289 − 1.894	1.673***	1.445 − 1.937
Education (in years)	1.035***	1.029 − 1.041	1.025***	1.018 − 1.031	1.033***	1.022 − 1.043	1.017***	1.009 − 1.026
Annual household income (per capita)	1.137***	1.116 − 1.158	1.068***	1.047 − 1.091	1.098***	1.059 − 1.138	1.053***	1.026 − 1.080
MS (Ref. married, living with spouse)
Registered partnership	0.949	0.774 − 1.163	1.073	0.873 − 1.318	1.220	0.883 − 1.685	0.980	0.754 − 1.274
Married, not living with spouse	1.392**	1.113 − 1.741	1.273*	1.016 − 1.594	1.085	0.763 − 1.543	1.449*	1.072 − 1.960
Never married	0.950	0.852 − 1.058	0.902	0.808 − 1.006	0.917	0.772 − 1.089	0.875	0.760 − 1.009
Divorced	1.093*	1.001 − 1.194	0.978	0.895 − 1.068	0.974	0.863 − 1.100	0.962	0.845 − 1.094
Widowed	0.936	0.852 − 1.028	0.885*	0.805 − 0.973	0.807**	0.713 − 0.912	1.086	0.928 − 1.273
Low-quality parent–child relationship	0.894*	0.804 − 0.995	0.846**	0.759 − 0.942	0.833*	0.715 − 0.970	0.853*	0.732 − 0.993
Smoking	1.487***	1.411 − 1.567	1.467***	1.392 − 1.547	1.576***	1.454 − 1.709	1.372***	1.279 − 1.472
Region (Ref. Southern Europe)
Western Europe			1.990***	1.849 − 2.141	2.327***	2.064 − 2.623	1.808***	1.645 − 1.987
Northern Europe			1.774***	1.644 − 1.914	2.223***	1.969 − 2.511	1.525***	1.382 − 1.683
Eastern Europe			2.347***	2.131 − 2.585	2.408***	2.075 − 2.794	2.476***	2.170 − 2.826

Note. N = 34,295. MS = Marital status; OR = Odds Ratio; 95% CI = 95% Confidential Interval. ^*p < .05, ^**p < .01, ^***p < .001.

Apart from being ‘divorced’, all these predictors stayed significant when entering EU-regions as additional predictors in model 2. In this model, also ‘widowed’ (OR = .885, CI = .805–.973, p = .012) predicted HAU. Compared to living in Southern Europe (reference category), living in Western Europe, in Northern or Eastern Europe (all p < .001) were a significant risk factors for HAU.

Gender-stratified predictors of hazardous alcohol use

In the gender stratified full model (Table 4), age brackets remained significant predictors of HAU. ORs for age brackets were slightly higher in men. Furthermore, higher educational years increased the risk for HAU slightly stronger in women (OR = 1.033, CI = 1.022–1.043, p < .001) than in men (OR = 1.017, CI = 1.009–1.026, p < .001). Similarly, higher income increased the risk of HAU stronger in women (OR = 1.098 CI = 1.059–1.138, p < .001) than in men (OR = 1.053, CI = 1.026–1.080, p < .001). Regarding marital status, the category ‘married, not living with spouse’ was only associated with a significantly higher risk of HAU in men (OR = 1.449, CI = 1.072–1.960, p = .032), whereas being widowed only was significantly and negatively associated in women (OR = .807, CI = .713–.912, p = .001). A low-quality parent–child relationship significantly predicted HAU in women (OR = .833, CI = .715–.970, p = .019) and men (OR = .853, CI = .732–.993, p = .040). Female smokers showed a slightly stronger increased risk of HAU (OR = 1.576, CI = 1.454–1.709, p < .001) compared to male smokers (OR = 1.372, CI = 1.279–1.472, p < .001).

Regarding EU-regions, again Western, Northern and Eastern Europe showed higher risks of HAU compared to Southern Europe, with the highest OR in Eastern European men and women. For Western and Northern Europe, the risks were stronger increased in women (OR = 2.327, CI = 2.064–2.623, p < .001; OR = 2.223, CI = 1.969–2.511, p < .001) than in men (OR = 1.808, CI = 1.645–1.987, p < .001; OR = 1.525, CI = 1.382–1.683, p < .001).

Predictors of hazardous alcohol use stratified by EU-region

Stratified by EU-region, analyses revealed gender, age and education as significant predictors in all four regions in Europe (see Table 5), except for 70–79 years. compared to 80+ years. in Eastern Europe. Female gender was associated with a stronger decreased risk for HAU in Southern and Eastern Europe (OR = .239, CI = .213–.268, p < .001; OR = .243, CI = .204–.290, p < .001) than in Western and Northern Europe (OR = .345, CI = .317–.376, p < .001; OR = .377, CI = .338–.420, p < .001). Years of education were positively and stronger associated with an increased risk of HAU in Eastern Europe (OR = 1.065, CI = 1.037–1.095, p < .001) than in all other EU-regions. The increased risk in the youngest compared to the oldest age bracket was stronger in Northern Europe compared to all other European regions. Additionally, higher household income was associated with an increased risk for HAU in Western and Northern Europe (OR = 1.177, CI = 1.121–1.236, p < .001; OR = 1.278, CI = 1.214–1.346, p < .001), but not in Southern and Eastern Europe. Categories of marital status were not significantly associated with HAU in the EU region stratified analyses after Bonferroni–Holm correction (see Table 5).

Table 5. Binary logistic regression: Predictors of hazardous alcohol use stratified by EU-region.

	Southern Europe		Western Europe		Northern Europe		Eastern Europe
Nagelkerkes R^2	.143		.128		.171		.186
Variable	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
Female gender	0.239***	0.213–0.268	0.345***	0.317–0.376	0.377***	0.338–0.420	0.243***	0.204–0.290
Age (Ref. 80+ years.)
50–59 years.	2.470***	1.935–3.151	1.901***	1.592–2.270	4.141***	3.248–5.280	2.426***	1.681–3.502
60–69 years.	2.419***	1.910–3.063	1.779***	1.494–2.119	2.789***	2.199–3.537	2.314***	1.630–3.284
70–79 years.	1.957***	1.537–2.493	1.383***	1.154–1.657	1.836***	1.440–2.340	1.489	1.036–2.140
Education (in years)	1.013*	1.001–1.025	1.023***	1.014–1.032	1.025***	1.010–1.041	1.065***	1.037–1.095
Annual household income (per capita)	0.983	0.960–1.008	1.177***	1.121–1.236	1.278***	1.214–1.346	1.023	0.925–1.132
MS (Ref. married, living with spouse)
Registered partnership	1.167	0.853–1.596	1.090	0.778–1.527	1.069	0.662–1.725	0.000	0.000–0.000
Married, not living with spouse	1.652	1.001–2.726	1.007	0.718–1.410	1.788	1.120–2.854	1.282	0.677–2.429
Never married	1.072	0.848–1.356	0.913	0.772–1.080	0.782	0.638–0.958	1.092	0.694–1.718
Divorced	1.047	0.815–1.345	0.998	0.876–1.136	0.944	0.798–1.116	0.926	0.716–1.199
Widowed	1.004	0.824–1.224	0.875	0.751–1.018	0.940	0.776–1.137	0.782	0.600–1.020
Low-quality parent–child relationship	0.938	0.716–1.230	0.772**	0.667–0.893	0.951	0.765–1.182	0.734	0.449–1.200
Smoking	1.434***	1.292–1.592	1.495***	1.375–1.627	1.621***	1.455–1.807	1.197*	1.011–1.416
N	11,616		11,617		7948		3114

Note. MS = Marital status; OR = Odds Ratio; 95% CI = 95% Confidential Interval. *p < .05, **p < .01, ^***p < .001.

A low-quality parent–child relationship only significantly decreased the risk in Western Europe (OR = .772, CI = .667–.893, p = .002). Lastly, smoking was associated with in increased risk for HAU in all EU-regions. Analyses stratified for gender and EU-region are displayed in Supporting Information Tables S1 and S2.

Discussion

The present study investigated determinants of alcohol consumption and HAU in a large European sample. Regarding alcohol consumption (alcohol units last week), age showed an effect narrowly below a small effect, EU-Region had a small effect, whereas gender had a medium effect size. Alcohol consumption decreased with increasing age and was substantially lower in women. We found a higher alcohol consumption in Western Europe compared to all other EU regions. Northern Europe showed lower alcohol consumption data compared to Western and Eastern Europe, but no difference in alcohol consumption compared to Southern Europe.

HAU was present in 25.4% of our overall sample of European OA. We found differences for gender, age and EU-region in prevalences of HAU, which was more than twice as prevalent in men compared to women and more than twice as prevalent in the youngest age bracket compared to the oldest.

The logistic regression model confirmed these effects of gender and age on HAU. The risks for HAU in younger age brackets compared to the oldest were slightly higher in men than in women, hinting to a slightly stronger difference across age brackets in men versus women. This could in part be due to the fact that overall fewer women engage in HAU compared to men (Britton et al., 2015), which means that there is less scope for reduction over the years. Stratification by region and gender (see Supporting Information Tables S1 and S2) confirmed this pattern for all EU-regions but Southern Europe, where the ORs for women are slightly higher than among men across all age groups. Also, the effect of age differed between EU-regions and was largest in Northern Europe, with higher risks for HAU among younger participants (see Table 5).

Additionally, SES variables like education and income were associated with HAU: Years of education were positively associated with a slightly increasing risk for HAU. This was also true for both genders and all EU-regions. In the gender- and region-stratified analyses this effect only remained significant in Western and Eastern European women. While this association of higher education and higher prevalence of HAU among women has been reported for several countries (Bloomfield et al., 2006; Grittner et al., 2013), for men, this observation is interesting, since usually the pattern is described as being reversed in men (Devaux & Sassi, 2016). This might be explained by our sample of adults aged 50+ years, a population in which alcohol use has previously been associated with higher SES (Bonevski et al., 2014). Since the harmful consequences of heavy alcohol consumption more strongly affect people of lower SES (Katikireddi et al., 2017; Probst et al., 2020), lower educated people might reduce or stop drinking earlier due to health consequences, while persons with higher education may tend to stay relatively healthy and maintain their drinking habits. Within the region-stratified analysis, income was significantly and positively associated with HAU in Northern and Western Europe, but had no significant effect on HAU in Southern and Eastern Europe. The significant positive associations of income and HAU in Northern and Western Europe likely reflect the higher affordability of alcohol for people with higher income, especially in regions with higher alcohol prices (Wagenaar et al., 2009): where alcohol prices are high (e.g. because of higher alcohol duty rates (Angus et al., 2019)), alcohol is more affordable for people with higher income. Although alcohol policies and taxation differ significantly between countries, most Northern and some Western European countries have higher alcohol duty rates than Eastern and Southern European countries (Angus et al., 2019), possibly leading to a higher alcohol consumption among wealthier people in these countries.

Regarding marital status, being ‘widowed’ decreased the risk of HAU, while being ‘married, not living with spouse’ and being ‘divorced’ increased the risk. Gender-stratified analyses revealed that being widowed only decreased the risk in women, while being married, but not living with the spouse significantly increased the risk of HAU in men. Previous studies have shown that single living arrangements often increased the risk of hazardous drinking in older men, while there was mostly no or even a protective effect in older women (Bosque-Prous et al., 2017; Iparraguirre, 2015; Watt et al., 2014). A possible explanation are fewer social contacts (Iliffe et al., 2007; Kharicha et al., 2007; Shimada et al., 2014) and a different influence on drinking habits for men and women. Indeed, a study from Sweden found that women but not men aged 65+ with lower levels of social contacts were less likely to regularly consume alcohol (Agahi et al., 2019). This pattern could be explained by conventional gender norms that potentially prevent older women from drinking alone (Agahi et al., 2019).

Interestingly, a low-quality parent–child relationship slightly decreased the risk for HAU only for Western European participants. Further region- and gender-stratified analyses revealed that this effect was significant only in Western European men. Again, since a poor parent–child relationship can lead to an early and rather harmful use of alcohol in younger years (Boden et al., 2021), such individuals may stop or reduce drinking in later life due to health consequences.

The risk for HAU in smokers was higher in women than men, which is in line with findings of a recent review (Wilsnack et al., 2018). The authors assume that changing gender stereotypes might lead to gender differences in the association of binge drinking and tobacco use. Interestingly, smoking showed a weaker association with HAU in Eastern Europe.

EU-regions had an additional predictive value on HAU in our study. All other EU-regions had higher risks of HAU compared to Southern Europe, and Eastern Europe showed the highest risk for HAU. Interestingly, the risk increased in Western and Northern Europe compared to Southern Europe higher for women compared to men. Likewise, the effect of gender was stronger in Western and Northern Europe than in Southern and Eastern Europe.

Strengths & limitations

The large sample of middle- to older-aged participants from all over Europe represents a strength of this study as it allows comparisons throughout large parts of the European continent and it allows stratification for gender and European regions. Therefore, the large sample enabled us to analyze important gender- and EU-region-sensitive predictors of HAU as a crucial public health issue.

Due to a lack of assessing abstinence in Wave 6, we were not able to identify abstainers. Since differences in the proportion of abstainers between regions and age groups are likely (Calvo et al., 2020), it is important to notice that the mean units of alcohol last week reflect the consumption per capita rather than the average amount of alcohol consumed within the group of drinkers, which may vary heavily. However, this shortage does not affect the outcomes of the prediction of HAU. Since mental health variables like anxiety or depression are potential confounders of the association between risk or protective factors and alcohol consumption, further research should consider statistical adjustments for mental health. Furthermore, we chose 12.5 units of alcohol per week as the cut-off point for risky alcohol use irrespective of gender, since a higher consumption is associated with increased risk of morbidity and lower life-expectancy in both, men and women (Wood et al., 2018). Compared to most drinking guidelines, this is a rather low threshold (Kalinowski & Humphreys, 2016), which might contribute to the high proportion of HAU within our data. But since it is associated to harmful effects on health in men and women (Wood et al., 2018) and was proposed as an threshold especially for older people (Royal College of Psychiatrists, 2018), this seems to be a suitable definition for HAU in OA. Conversely, the commonly used cut-off of five drinks for men and four drinks for women on one occasion for binge drinking (Han et al., 2019) would possibly lead to higher HAU prevalences in our sample and could also impact findings on our HAU predictors.

Conclusion

Regional differences for HAU in European OA were stronger in women than in men and were present beyond established risk factors for alcohol misuse, and therefore, not solely explained by social and socioeconomic differences between EU-regions.

Our data showed that HAU is highly prevalent in European OA. Markers of higher SES were associated with a higher risk of HAU. We observed an additional regional effect on HAU in European OA, with the highest risk of HAU in Eastern Europe.

Because of these regional differences in HAU and its predictors in European OA, population-level strategies need to be adjusted to vulnerable groups in the targeted region and not just adopted from other regions. Future research should focus on a deeper understanding of those regional differences to further inform risk-stratified prevention.

Acknowledgement

This paper uses data from SHARE Waves 6 and 7 (DOIs: 10.6103/SHARE.w6.710, 10.6103/SHARE.w7.711), see Börsch-Supan et al. (2013) for methodological details.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the German Ministry of Education and Research (Forschungsnetz AERIAL 01EE1406A, 01EE1406B, 01EE1406I) and the Deutsche Forschungsgemeinschaft DFG (SFB 940). The SHARE data collection has been funded by the European Commission, DG RTD through FP5 (QLK6-CT-2001-00360), FP6 (SHARE-I3: RII-CT-2006-062193, COMPARE: CIT5-CT-2005-028857, SHARELIFE: CIT4-CT-2006-028812), FP7 (SHARE-PREP: GA No. 211909, SHARE-LEAP: GA No. 227822, SHARE M4: GA No. 261982, DASISH: GA No. 283646) and Horizon 2020 (SHARE-DEV3: GA No. 676536, SHARE-COHESION: GA No. 870628, SERISS: GA No. 654221, SSHOC: GA No. 823782) and by DG Employment, Social Affairs & Inclusion through VS 2015/0195, VS 2016/0135, VS 2018/0285, VS 2019/0332, and VS 2020/0313. Additional funding from the German Ministry of Education and Research, the Max Planck Society for the Advancement of Science, the U.S. National Institute on Aging (U01_AG09740-13S2, P01_AG005842, P01_AG08291, P30_AG12815, R21_AG025169, Y1-AG-4553-01, IAG_BSR06-11, OGHA_04-064, HHSN271201300071C, RAG052527A) and from various national funding sources is gratefully acknowledged (see www.share-project.org).