Open access Original research Educational differences in alcohol-­ related mortality and their impact on life expectancy and lifespan variation in Spain (2016–2018): a cross-­sectional analysis using multiple causes of death Sergi Trias-­Llimós ‍ ‍, Jeroen J. A. Spijker ‍ ‍ To cite: Trias-­Llimós S, ABSTRACT Spijker JJA. Educational Strengths and limitations of this study Background Socioeconomic inequalities in alcohol-­ differences in alcohol-­related related mortality in Spain exists, and are postulated to mortality and their impact ► We examined socioeconomic inequalities in alcohol-­ contribute to inequalities in all-­cause mortality. We aim on life expectancy and related mortality in Spain using individual-­level mul- to assess absolute and relative educational inequalities lifespan variation in Spain tiple causes-­of-­death (MCOD) mortality data linked (2016–2018): a cross-­sectional in alcohol-­related mortality, and to estimate the role of with educational attainment as key socioeconomic analysis using multiple alcohol in educational inequalities in both life expectancy variable. causes of death. BMJ Open and lifespan variation in Spain. ► This is the first study that examines population-­level 2022;12:e053205. doi:10.1136/ Methods We used multiple cause-­of-­death (MCOD) socioeconomic inequalities in alcohol-­related mor- bmjopen-2021-053205 mortality data for individuals aged 30 and over for Spain tality in Spain using MCOD data. (2016–2018) by educational attainment. We estimated by ► Prepublication history and ► The observed inequalities are sensitive to the group- additional supplemental material sex and educational attainment age-­standardised alcohol-­ ing of the socioeconomic variable, education and its for this paper are available attributable mortality rates, relative and absolute indices composition, which may complicate comparisons online. To view these files, of educational inequalities; and total life expectancy and with future studies of the same group. please visit the journal online lifespan variation at age 30 for all-­cause mortality and ► MCOD data may still not be capturing all deaths due (http://dx.doi.org/10.1136/​ after eliminating alcohol-­attributable mortality. bmjopen-2021-053205). to alcohol consumption, although this underestima- Results The use of MCOD resulted in an additional 2543 tion is lower compared with commonly used under- annual alcohol-­related deaths (+75% among men and Received 10 May 2021 lying cause-­of-­death approaches. +50% among women) compared with estimates derived Accepted 23 November 2021 from underlying causes of death. In absolute terms, educational inequalities were the highest among men aged 45–84 and among women aged 45–64. In relative to country contexts and how alcohol-­related terms, higher inequalities raised in working ages, whereas mortality and socioeconomic groups are at older ages inequalities tended to be lower, although still defined and measured, according to previous important among men. Alcohol contributed to educational research low socioeconomic groups have inequalities in life expectancy (men: 0.13 years (3.2%); women 0.02 years (0.7%)) and lifespan variation (2.1% between 2 and 10 times higher mortality due and 1.4% for men and women, respectively). to alcohol consumption compared with their Conclusion Alcohol consumption remains an high socioeconomic counterparts.2 4 6 These important lifestyle habit to be tackled in order to reduce inequalities in alcohol-­related mortality make socioeconomic inequalities in mortality in Spain, alcohol a net contributor to all-­cause socio- © Author(s) (or their particularly among men. economic inequalities in mortality.2 7 employer(s)) 2022. Re-­use Understanding why these socioeconomic permitted under CC BY-­NC. No differentials in alcohol-­related mortality exist commercial re-­use. See rights INTRODUCTION is, however, not that straightforward as it is and permissions. Published by BMJ. Alcohol consumption is a major contrib- not only associated with similar social gradi- Centre d'Estudis Demografics, utor to the burden of disease and mortality.1 ents in consumption patterns. Although it is Centres de Recerca de Despite the important role of alcohol in true that, as for other lifestyles, individuals Catalunya (CERCA), Universitat health outcomes, overall estimates mask from low socioeconomic groups tend to have Autònoma de Barcelona, notable socioeconomic differences, as low higher prevalence of risky alcohol consump- Bellaterra, Spain socioeconomic groups tend to suffer a greater tion such as heavy drinking,8 9 the impact of Correspondence to health burden from alcohol consumption alcohol on health and mortality is known Dr Sergi Trias-­Llimós; than those from a higher socioeconomic to be disproportionally higher among low ​strias@​ced.​uab.​cat strata.2–6 Although the estimates are sensitive socioeconomic groups. In other words, when Trias-­Llimós S, Spijker JJA. BMJ Open 2022;12:e053205. doi:10.1136/bmjopen-2021-053205 1 Open access consumption levels are controlled for, low socioeconomic on alcohol-­related mortality using MCOD has also been groups suffer a higher health burden due to alcohol. This conducted in France,20 the USA23 24 and the UK,25 but paradox is known as the alcohol-­harm paradox, which is none of these studies focused on socioeconomic differen- partly explained by other unhealthy lifestyles such as poor tials therein. In other contexts, including Southern Euro- diet, smoking or insufficient physical activity.10 11 pean countries, lie Spain, there is still a lack of studies on While most recent European studies on socioeconomic inequalities in alcohol-­attributable mortality using more differences in alcohol-­ related mortality have focused on refined approaches compared with the UCOD. Nordic or Eastern European countries,6 12 13 Spain and other Using Spanish MCOD mortality data for time period southern European countries have received little attention, 2016–2018 our objective is therefore twofold: despite there being distinct reasons to do so. First, alcohol 1. To assess absolute and relative socioeconomic differ- drinking is considered a cultural aspect of social life in ences in alcohol-­related mortality. Spain, and alcohol consumption is widespread.14 Second, 2. To assess the role of educational differences in alcohol-­ after continuous declines in alcohol consumption since related mortality to all-­cause inequalities in life expec- the late 1970s, its levels increased moderately in the period tancy and lifespan variation. 2013–2016,15 and are currently at 10.9 L of pure alcohol consumption per capita.15 Third, earlier studies reported METHODS a socioeconomic gradient in alcohol-­related mortality in a Data sample of 22 million residents aged 35 and over in the period Detailed sex-­specific and age-­specific mortality data for indi- 2001–2011,4 as well as in three Spanish subpopulations viduals aged 30 and over containing all causes listed in the (Barcelona, Basque Country and Madrid) during the late death certificate were retrieved for the years 2016–2018 from 1990s and 2000s.2 In both studies, alcohol-­related mortality the Spanish National Statistics Institute (INE). These data was found to be 4–6 times higher among low socioeconomic contain information on educational attainment retrieved groups compared with the high socioeconomic groups. Yet, from a data linkage with a source feed from multiple data- these previous studies are based on old or regional data and sets and carried out by INE. The latest includes municipal not on national-­level data. Finally, beyond alcohol-­related population registers (Padrón), the 2001 and 2011 censuses, mortality inequalities, all-­cause mortality inequalities have data on official degrees issued by the Ministry of education persisted (or even increased) over the last several decades or data on enrolment at Spanish universities, register of in Spain.16 17 enrolled and graduated at non-­university education. In case Socioeconomic inequalities in alcohol-­related mortality of multiple educational attainment information, the highest have been mostly examined using underlying causes-­of-­ level of education was chosen. More details about this data death (UCOD).2 3 18 UCOD approaches tend to underes- linkage can be found elsewhere.26 27 Population estimates timate the impact of alcohol on mortality as they are only from 2017 by age, sex and educational attainment, also capturing the share of mortality attributable to alcohol retrieved from INE, were set as denominators to produce from diseases clearly related to alcohol (eg, alcoholic our mortality indicators. Based on both data sources, liver cirrhosis), and therefore these methods are not able we created three educational attainment categories: low to capture the burden of alcohol to other diseases and (primary education or less, ISCED-­2011 0–2) middle (lower causes of death in which alcohol is not the risk factor and upper secondary education, ISCED-­2011 3) and high (eg, ischaemic heart diseases or several neoplasms). The (postsecondary vocational and university education, ISCED-­ complex interactions between alcohol consumption 2011 4+). The population distribution by age, sex and educa- (levels and patterns of drinking) and health outcomes tional attainment is shown in online supplemental figure S1. represent an important challenge as regards to accurately estimate the impact of alcohol on mortality at the popula- Patient and public involvement tion level.1 19–21 One of the possibilities to partly overcome No patients were involved in this study. some of the limitations of UCOD consists in defining alcohol-­related mortality as the sum of deaths with a cause of death wholly attributable to alcohol (100% attributable to alcohol) listed in the death certificate (eg, alcoholic METHODS liver cirrhosis, mental and behavioural disorders due to Alcohol-­related mortality was defined as the sum of deaths use of alcohol), either as underlying or contributory, also with a cause of death 100% attributable to alcohol mentioned known as the multiple cause-­of-­death (MCOD) approach. anywhere in the death certificate (MCOD). In line with The use of MCOD approaches is a promising tool for previous studies,20 28 we used the following causes of death population-­level alcohol research studies. However, there that are wholly related to alcohol: alcoholic dependence are only a handful of studies that used MCOD to examine (International Classification of Diseases version 10 (ICD-­10): socioeconomic inequalities in alcohol-­related mortality. F10), alcoholic liver cirrhosis (K70), external causes due to These mainly come from the Nordic countries,13 22 alcohol use (X45, X65 and Y15) and other causes wholly where alcohol-­related mortality was found to account for attributable to alcohol (G312, G621, G721, I426, K292 and 15%–35% of the income gradient in life expectancy K852; causes with zero deaths were not included, eg, E244, among men and 10%–25% among women.13 Research 0354, Q860 and R780). Therefore, deaths with one of the 2 Trias-­Llimós S, Spijker JJA. BMJ Open 2022;12:e053205. doi:10.1136/bmjopen-2021-053205 Open access abovementioned causes either listed as underlying or as The role of alcohol (MCOD approach) in all-cause mortality contributory (at any other section in the death certificate) To analyse the role of alcohol in all-­cause mortality, we were considered to be related with alcohol.22 estimated life expectancies at age 30 (e30) for each sex and educational group for the whole population and Educational inequalities in alcohol-related mortality (MCOD eliminating alcohol-­related mortality. We did so by using approach) 5-­year age group data starting at age 30 and by applying We calculated age-­standardised alcohol-­related mortality standard life table techniques.30 Lifespan variation for the rates by age (30–44, 45–64, 65–84, 85+ and 30+), sex and same population groups was measured as the SD in the educational category using the Spanish population in age at death from age 30 (SD30) using life table data. To 2016–2018 as the standard. 95% CIs were estimated using estimate it from age 30, we adapted the formula provided conventional approaches: in previous research.5 31 ( ) We estimated the contribution of alcohol to life expec- asdr 95% CI = asdr ± 1.96 √ asdr tancy and lifespan variation by sex and educational group ‍ n deaths ‍ by comparing the corresponding values for the whole where, asdr refers to age-­standardised death rates and n population and excluding alcohol-­related deaths, defined deaths to the observed number of deaths. as potential gains in life expectancy (PGLE).32 In other Educational inequalities in alcohol-­ related mortality words, PGLE measures the potential increases in life were estimated by means of both absolute and relative expectancy if alcohol-­related mortality was eliminated. inequality measures. The relative index of inequality Finally, we estimated the inequality gap in life expectancy (RII) was used for measuring relative inequalities. The RII and lifespan variation between the highest and the lowest assumes the relative position of each educational group educated group, both in absolute and in relative terms. within the population, and it can be interpreted as the mortality rate ratio comparing those with low education RESULTS to those with high education. The RII was estimated from Descriptive results a multiplicative Poisson model adjusting by age following We estimated nearly 8000 alcohol-­related deaths in Spain the work by Moreno-­Betancur et al.29 Absolute inequalities for the period 2016–2018 (table 1). Within the UCOD, were measured by means of the slope index of inequality alcohol-­related mortality was mostly composed of deaths (SII),2 which measures the rate differences between those from alcoholic liver cirrhosis (77.4%) and alcohol depen- with low education to those with high education: ( ) dence (15.4%). MCOD deaths represented an increase SII = 2×asdr× RII−1 ( ) in alcohol-­related mortality of 75.1% among men and ‍ RII+1 ‍ 50.2% among women. Table 1 Alcohol-­related death counts by age and cause distinguishing between main underlying causes, alcohol listed in the contributory causes of death and total derived from either underlying or contributory causes: MCOD, Spain 2016–2018 Deaths with alcohol-­related causes as UCOD Alcoholic External causes liver Alcoholic due to alcohol Other causes wholly-­ All Alcohol listed in the cirrhosis dependence use (X45, X65, attributable to alcohol as contributory causes (K70) (F10) Y15) alcohol* underlying of death† MCOD MCOD/UCOD Men  30–44 156 18 28 6 208 66 274 1.32  45–64 1728 276 62 94 2160 1293 3453 1.60  65–84 1057 294 15 68 1434 1444 2878 2.01  85+ 74 27 0 3 104 132 236 2.27  All ages 30+ 3015 615 105 171 3906 2935 6841 1.75 Women  30–44 42 3 6 3 54 15 69 1.28  45–64 340 47 28 10 425 163 588 1.38  65–84 168 38 2 9 217 160 377 1.74  85+ 13 9 1 0 23 23 46 2.00  All ages 30+ 563 97 37 22 719 361 1080 1.50 *Corresponds to the cause-­of-­death categories and ICD-­10 codes: degeneration of nervous system due to alcohol (G31.2), alcoholic polyneuropathy (G62.1), alcoholic myopathy (G721), alcoholic cardiomyopathy (I42.6), alcoholic gastritis (K29.2) and alcohol-­induced acute pancreatitis (K85.2). †Refers to any alcohol-­related cause of death listed in columns 1–3 and the above note. MCOD, multiple causes-­of-­death; UCOD, underlying causes-­of-­death. Trias-­Llimós S, Spijker JJA. BMJ Open 2022;12:e053205. doi:10.1136/bmjopen-2021-053205 3 Open access Table 2 Age-­standardised alcohol-­related mortality rates (per 100 000 person-­years) by age groups and educational attainment, RII and SII, Spain 2016–18 Age-­standardised mortality rates by educational group (95% CI) Inequalities All Low Middle High RII (95% CI) SII Men  30–44 1.5 1.3 to 1.7 2.5 2.1 to 2.9 1.6 1.2 to 2.0 0.4 0.2 to 0.5 5.17 (3.13 to 8.55) 2.0  45–64 16.9 16.4 to 17.5 24.1 23.2 to 25.1 13.7 12.5 to 14.8 7.0 6.3 to 7.7 3.42 (2.78 to 4.22) 18.5  65–84 27.5 26.5 to 28.6 33.2 31.8 to 34.5 17.7 15.4 to 20.0 12.7 11.1 to 14.4 2.75 (2.42 to 3.12) 25.7  85+ 15.3 13.3 to 17.3 16.6 14.4 to 18.9 10.5 4.3 to 16.7 7.0 2.6 to 11.3 2.38 (1.34 to 4.24) 12.5  All ages 30+ 14.3 14.0 to 14.7 18.9 18.4 to 19.4 10.5 9.8 to 11.2 6.2 5.7 to 6.6 3.08 (1.82 to 5.02) 8.0 Women  30–44 0.4 0.3 to 0.5 0.7 0.5 to 1.0 0.5 0.2 to 0.7 0.1 0.0 to 0.2 6.05 (2.80 to 13.04) 0.6  45–64 2.7 2.5 to 3.0 3.8 3.4 to 4.2 2.3 1.9 to 2.8 1.5 1.2 to 1.8 2.62 (1.95 to 3.52) 2.4  65–84 2.9 2.6 to 3.2 2.9 2.5 to 3.2 2.9 1.9 to 3.9 3.7 2.6 to 4.9 0.86 (0.60 to 1.22) −0.4  85+ 1.6 1.1 to 2.0 1.5 1.0 to 1.9 1.4 −0.5 to 3.3 4.3 0.5 to 8.0 0.39 (0.15 to 1.00) −1.4  All ages 30+ 2.0 1.8 to 2.1 2.5 2.3 to 2.7 1.8 1.5 to 2.1 1.7 1.4 to 1.9 2.11 (1.21 to 3.68) 1.4 *Bold indicates statistically significant (95%) RII values. RII, relative index of inequality; SII, slope index of inequality. Alcohol-related mortality by age group, sex and educational men it ranged from 5.17 (3.13–8.55) at ages 30–44 to 2.38 attainment and relative inequalities (1.34–4.24) at ages 85 and over, while among women it Among men, alcohol-­related mortality rates ranged from ranged from 6.05 (2.80–13.04) to 0.39 (0.15–1.00) in the 6.2 (5.7–6.6) per 100 000 person-­years among the highest youngest and older age group, respectively. The SII peaked educated group to 18.9 (18.4–19.4) per 100 000 person-­ at ages 65–84 among men with 25.7 deaths per 100 000 years among the lowest educated group (table 2). Among person-­ years; and at ages 45–64 among women with 2.4 women, alcohol-­related mortality rates ranged between deaths per 100 000 person-­years. 1.7 (1.4–1.9) and 2.5 (2.3–2.7) per 100 000 person-­years, among the highest and the lowest educated groups, Impact of alcohol on overall mortality and on inequalities respectively. When stratified by educational groups, a therein clear socioeconomic gradient can be visually observed In Spain, the PGLE was 0.15 years among men and 0.03 across age groups, particularly among men and among years among women (table 3). Mortality indicators were working-­age women (figure 1). more favourable among highest educated groups: higher The relative index of inequality (RII) was 3.08 (1.82–5.22) life expectancies at age 30 (e30) and lower lifespan variation among men and 2.11 (1.21–3.68) among women. Among (SD30). When alcohol-­related mortality was eliminated, life Figure 1 Alcohol-­related mortality rates by age and educational attainment in Spain 2016–2018. 4 Trias-­Llimós S, Spijker JJA. BMJ Open 2022;12:e053205. doi:10.1136/bmjopen-2021-053205 Open access Table 3 Contribution of alcohol to all-­cause life expectancy and lifespan educational inequalities at age 30 by sex, Spain 2016–2018 Educational groups % Reduction in inequality by All Low Middle High Inequality gap eliminating alcohol Men  e30 50.66 49.19 52.49 53.36 4.17  e30 eliminating alcohol 50.81 49.39 52.62 53.43 4.04  PGLE alcohol 0.15 0.20 0.12 0.07 0.13 3.2  SD30 12.86 13.63 12.89 11.66 −1.98  SD30 eliminating alcohol 12.79 13.55 12.82 11.61 −1.93  Difference −0.07 −0.09 −0.07 −0.04 −0.04 2.1 Women  e30 55.92 55.16 57.61 57.73 2.56  e30 eliminating alcohol 55.95 55.20 57.64 57.75 2.55  PGLE alcohol 0.03 0.04 0.03 0.02 0.02 0.7  SD30 11.46 12.19 11.39 10.69 −1.50  SD30 eliminating alcohol 11.43 12.15 11.36 10.67 −1.48  Difference −0.03 −0.04 −0.03 −0.02 −0.02 1.4 PGLE, potential gains in life expectancy. expectancy at age 30 increased by 0.20, 0.12. and 0.07 years groups. Indeed, at ages 85 and over it seems that lower among low, middle and high educated groups in men; and educated women have lower alcohol-­ related mortality by 0.04, 0.03 and 0.02 years among the corresponding groups (RII: 0.39, 95% CI: 0.15 to 1.00). Weaker inequalities in women. Additionally, alcohol-­related mortality contrib- among women were also previously found elsewhere,2 but uted to lifespan variation. By eliminating alcohol-­ related unfortunately age-­specific results were not reported. mortality, all-­ cause inequalities in men would decline by Interestingly, our results on educational inequalities 3.2% in terms of life expectancy at age 30 and 2.1% in terms in alcohol-­related mortality can be compared with those of lifespan variation. Among women, the corresponding from smoking-­ related mortality from a recent study declines would be 0.7% and 1.4%. that used Spanish data from 2016.33 In relative terms, our results for the relative inequalities seem larger as compared with those from smoking-­related mortality,33 DISCUSSION for men (our estimates: RII=3.08, 95% CI: 1.82 to 5.22; Summary of results smoking estimates: RII=1.61, 95% CI: 1.55 to 1.67) and We have assessed educational inequalities in alcohol-­ women (our estimates: RII=2.11, 95% CI: 1.21 to 3.68; related mortality and their role in all-­ cause mortality smoking estimates: RII=0.39, 95% CI: 0.35 to 0.42), inequalities using multiple causes of death data for Spain, suggesting a higher relative importance of alcohol in and life tables and regression-­based inequality measures. educational inequalities in health. In both alcohol-­ Educational inequalities in alcohol-­related mortality were related and smoking-­related mortality an inverse gradient larger in working age groups (RII around or above 3), with higher lifestyle-­ related mortality among women whereas they decreased at older ages, particularly among aged 65 and over from high social classes was found. In women. When eliminating alcohol-­ related mortality, absolute terms, this comparison suggests higher absolute inequalities in life expectancy at age 30 diminished by inequalities in smoking-­ related mortality as compared 3.2% in men and 0.7% in women; and lifespan variation with alcohol-­related mortality among men, which seems diminished by 2.1% in men and 1.4% in women. to be explained by the higher levels of smoking-­related Comparison of results mortality compared with alcohol-­related mortality. Our results of the observed educational inequalities in Alcohol was found to contribute to all-­cause life expec- alcohol-­related mortality among men are in line with tancy and lifespan variation. For life expectancy and previous research from Mackenbach et al who assessed for the age group 30–74 these contributions were 1.8% alcohol-­related mortality from UCOD in three Spanish for women and 4.2% for men (online supplemental regional populations in the period 2001–2010.2 Among table S1). As expected, these values are slightly higher women, we found relative inequalities at ages 30 and over, compared with previous estimates from the early 2000s and particularly at ages 30–64, but not among 65 or over that used UCOD data and the SII as inequality measure.2 Trias-­Llimós S, Spijker JJA. BMJ Open 2022;12:e053205. doi:10.1136/bmjopen-2021-053205 5 Open access In the present study, we found that the role of alcohol in increase with age as a result of an accumulation of expo- all-­cause mortality inequalities among women was slightly sures across the life course that have been affected by the larger for lifespan variation than for life expectancy. This social and economic experiences of the individual.39 40 finding implies that alcohol seems to play a major role However, at old age variables such as income and educa- in the distribution of deaths over age (lifespan variation) tion become less decisive determinants for health as age rather than in its age-­ standardised mean (life expec- becomes a leveller mechanism, in part due to selective tancy), in line with previous findings from England and mortality as frailer individuals, who are more likely to Wales.5 This stresses the importance of complementing pertain to a lower socioeconomic class, die earlier and conventional average indicators widely used in the litera- surviving members may then be less sensitive to income ture with indicators of age-­at-­death dispersion. gains and losses.41 42 Our results on the contribution of alcohol to educa- The results of our study should be also framed in the tional differences in life expectancy are difficult to be context of the alcohol consumption situation in Spain. compared with other studies as the only studies assessing According to data from the early 2010s, high socio- socioeconomic inequalities in alcohol-­related mortality economic groups tend to have higher binge drinking using multiple causes-­ of-­ death stratified the popula- consumption prevalence compared with their lower tion by income quintiles. These studies, which focused educated counterparts.43 Given that binge drinking or on Nordic countries, found that alcohol accounted for heavy episodic drinking patterns are the highest contrib- 16%–36% and 6%–27% of the gap in life expectancy at utor to morbidity and mortality outcomes,8 our results ages 25–79 between top and bottom income quintiles may therefore seem counterintuitive as we found higher among men and women, respectively.12 13 Despite that alcohol-­related mortality among the lower educated. their study population pertained to a slightly smaller age Yet, this so-­called alcohol-­harm paradox is observed and range and they used a different socioeconomic variable documented elsewhere.10 11 Our results, could be partly and compared outer quintiles rather than terciles, the explained by the interaction between binge drinking contribution of alcohol to socioeconomic differences in and overall alcohol consumption. For example, the prev- mortality in Nordic countries does seem to be somewhat alence of high risk drinker (an average of >40 g/day in larger than in Spain. This would not be surprising given the last 12 months among men and >24 g/day among the overall higher alcohol-­related mortality there.2 women) does not show educational differences in men, while a positive educational gradient was apparent in Interpretation of results binge drinking.43 In the case of women, educational The observed age-­specific socioeconomic differentials in differences in heavy drinking were even larger as alcohol-­related mortality among women are particularly compared with binge drinking.43 Thus, the prevalence worth noting. The observed age-­effects do not seem to be of high-­risk drinkers seems to partly explain the alcohol-­ driven by age factors given the substantial differences in harm paradox. Yet, further studies are needed, preferably the adoption of alcohol consumption across women from that include dietary patterns44 with as well as morbidity different socioeconomic statuses and cohorts.34 A plau- and mortality follow-­ups, to disentangle the mechanisms sible explanation may be found in the theory of diffusion that contribute the most to explaining the socioeconomic of innovations35 regarding alcohol use. Similar to what has differences in alcohol consumption and related health been well-­documented for smoking in Spain,36 alcohol outcomes. use first became widespread among men, followed by women from high socioeconomic classes before women Evaluation of data and methods from low socioeconomic classes also began consuming This study relied on MCOD data to estimate alcohol-­ alcohol in larger numbers. In the context of our study, related mortality. This is the first time that these the oldest age group (85 and over) was born before 1943 estimates are produced for Spain. Using MCOD, we esti- and represents a generation of women characterised by mated around 2640 annual deaths (2016–2018), which large socioeconomic differentials in alcohol use. This represents increases of 75% and 50% among men and is the case in Spain as well as in most European coun- women, respectively, compared with UCOD estimates. tries.37 38 Conversely, current Spanish middle-­aged women However, we only included those causes of death 100% were born in a context of a more widespread alcohol use attributable to alcohol, meaning that they would not have among all socioeconomic classes, and as current drinking occurred without alcohol consumption. We therefore patterns suggest, they will continue to drink at older ages. left out major alcohol-­related chronic diseases that are It is therefore possible that alcohol consumption and not always or entirely attributable to alcohol, including alcohol-­related problems in older women will increase in liver cancer, non-­alcoholic liver cirrhosis, ischaemic heart the near future.34 38 diseases and stroke.1 45 46 Recent research using attribut- Our findings on the increasing relative educational able fraction (AF) approaches estimated around 15 000 inequalities over age (except for the older age groups) are annual deaths in Spain in the period 2010–201728 and in line with the literature on socioeconomic inequalities around 31 500 deaths in 2017 according to the Global in health and were observed as well for smoking-­related Burden of Disease estimates.47 Our estimate of the mortality in Spain.33 Yet, in absolute terms, inequalities impact of alcohol on male life expectancy (0.15 years in 6 Trias-­Llimós S, Spijker JJA. BMJ Open 2022;12:e053205. doi:10.1136/bmjopen-2021-053205 Open access 2016–2018) is also lower when compared with those from lower compared with AF-­derived estimates. Yet, in other another GBD data-­ based study (0.95 years in 2013).48 types of MCOD studies cause-­deleted life table should be These differences could be due to several factors. While used carefully as they may overestimate the final outcome. AF-­driven approaches capture a larger proportion of Finally, as in all studies examining socioeconomic mortality that stems from alcohol consumption (and inequalities, the results are sensitive to the socioeconomic thus educational differences), these methods have been variables chosen and its distribution. In our case educa- criticised for its potential biases in combining different tion was the only possibility in our data. However, the rapid sources of data. First, it is the difficult to obtain accu- educational expansion in the Spanish society represents a rate estimates of alcohol consumption and relative risks challenge in the ability to compare educational inequal- for alcohol-­attributable health outcomes that may differ ities across age. We therefore grouped the educational across countries and socioeconomic status.21 Second, attainment categories (less than or equal to primary capturing old-­age mortality patterns is particularly chal- school, secondary school, postsecondary vocational and lenging. Indeed, in Spain, almost half of the estimated university education) in order to ensure enough cases at 15 000 annual deaths were actually at ages 75 and over both extremes in the youngest and oldest ages. Despite (nearly 7000),28 but discrepancies across estimates may be the abovementioned limitations and given that our aim important.20 21 49 was to estimate socioeconomic inequalities, the MCOD Overall, we must acknowledge that MCOD methods in approach is preferred above the UCOD approach, even Spain may not be accurately capturing deaths from causes if the impact of alcohol on mortality is likely to be under- partly related to alcohol, as recently documented and estimated as the AF approach also suffers from uncertain- discussed in the US context.24 50 For instance, alcohol-­ ties in its estimates and limitations of its use for studying related cancer mortality was only very rarely accounted socioeconomic differentials. for in the MCOD approach, compared with the 6500 deaths estimated using AF approach.28 Furthermore, we did not include non-­ alcoholic liver cirrhosis (ICD-­10: K74), even though an important share of deaths from CONCLUSIONS liver cirrhosis is estimated to be attributable to alcohol: Alcohol-­related mortality inequalities according to socio- over 80% for men aged 45–7428 ; or around 60%–80% in economic status remain important in Spain. The exis- high-­income countries.51 Consequently, liver cirrhosis is tence of distinct age-­specific patterns, with inequalities one of the most important contributors to overall alcohol-­ being stronger among younger age groups suggest socio- related mortality in Spain.28 52 These elevated shares may economic inequalities to persist in the coming decades. be partly due to the high stigmatisation in reporting These inequalities had an undesirable impact on overall causes of death where alcohol is explicitly mentioned.24 53 mortality inequalities, both in terms of mean indicators We therefore performed a sensitivity analysis by including (life expectancy) and dispersion of deaths (lifespan non-­alcoholic liver cirrhosis (K74) in our alcohol-­related inequalities). Our estimates based on MCOD raised the mortality estimates. Results showed that the impact of death counts substantially compared to when only UCOD alcohol on life expectancy increased by up to 0.5 years would have been used. Finally, monitoring recent trends for lower educated men. The contribution of alcohol to in alcohol use and related mortality is necessary to obtain overall life expectancy inequalities increased to 6.5% and a deeper understanding of its role on all-­cause mortality. 2.6% among men and women, respectively (compared Public health policymakers should consider tackling with our original estimates of 3.2% and 0.7%); and these alcohol abuse in order to reduce the harmful effects of estimates were larger when restricting the population to alcohol on health outcomes and monitor inequalities ages 30 to 74 (9.0% among men and 4.8% among women) across population groups. (see online supplemental tables S2-­S4). In addition, as in all cause-­specific mortality studies, Contributors ST-­L contributed to the conception and design of the work; the acquisition, analysis and interpretation of data for the work; drafted and revised the we must acknowledge that death certificates may not work critically for important intellectual content; approved the final version of the always be accurately filled out.54 Although the quality of work to be published; and agrees to be accountable for all aspects of the work in vital registration systems is considered to be ‘very high’ ensuring that questions related to the accuracy or integrity of any part of the work in Spain,55 further research is needed to more compre- are appropriately investigated and resolved. JS contributed to the conception and design of the work; contributed to the interpretation of data for the work; revised hensively assess the biases that exist in the death certif- the work critically for important intellectual content; approved the final version of icate data.56 One possible way to improve the quality of the work to be published. ST-­L is acting as guarantor. mortality data would be to increase the share of autop- Funding The study is supported with funding provided by the European Research sied bodies (currently around 5.6%) to, for example, Council (ERC-­2019-­COG agreement No 864616, HEALIN). ST-­L acknowledges the levels of Finland (21%),57 and to assess the extent to research funding from the Juan de la Cierva-­Formación program of the Spanish Ministry of Science and Innovation (FJC-­2019-­039314-­I). JS also receives funding which certificates may be differently completed according under the ‘Ramón y Cajal' I3 program (RYC-­2013-­14851) and the R&D project to socioeconomic status. COMORHEALTHSES (PID2020-­113934RB-­I00) from the same ministry. The funders Furthermore, regarding the methodological details we had no role in study design, data analysis, decision to publish or preparation of the have used cause-­deleted life table methods for estimating manuscript. PGLE. We did so as our aggregated results seemed much Competing interests None declared. Trias-­Llimós S, Spijker JJA. BMJ Open 2022;12:e053205. doi:10.1136/bmjopen-2021-053205 7 Open access Patient consent for publication Not applicable. 13 Östergren O, Martikainen P, Tarkiainen L, et al. Contribution of smoking and alcohol consumption to income differences in Ethics approval This study involved secondary data analysis of public sources, life expectancy: evidence using Danish, Finnish, Norwegian which did not have any individual identifiers. As such, ethical approval for human and Swedish register data. J Epidemiol Community Health subject research from the Institutional Review Board of the respective institutions 2019;73:334–9. was exempted. 14 Sureda X, Villalbí JR, Espelt A, et al. Living under the influence: normalisation of alcohol consumption in our cities. Gac Sanit Provenance and peer review Not commissioned; externally peer reviewed. 2017;31:66–8. Data availability statement Data may be obtained from a third party and are 15 Global Health Observatory data repository. Global information not publicly available. Multiple causes of death data and population exposures by system on alcohol and health (GISAH). levels of consumption, 2020. educational attainment was obtained from the Instituto Nacional de Estadística Available: https://apps.who.int/gho/data/node.main.A1022?lang=en 16 Permanyer I, Spijker J, Blanes A, et al. Longevity and lifespan (INE). Interested researchers have the possibility to obtain data access by variation by educational attainment in Spain: 1960-­2015. contacting INE. Demography 2018;55:2045–70. Supplemental material This content has been supplied by the author(s). It has 17 Mackenbach JP, Valverde JR, Artnik B, et al. Trends in health not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been inequalities in 27 European countries. Proc Natl Acad Sci U S A 2018;115:6440–5. peer-­reviewed. Any opinions or recommendations discussed are solely those 18 Kraus L, Østhus S, Amundsen EJ, et al. Changes in mortality due to of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and major alcohol-­related diseases in four Nordic countries, France and responsibility arising from any reliance placed on the content. Where the content Germany between 1980 and 2009: a comparative age-­period-­cohort includes any translated material, BMJ does not warrant the accuracy and reliability analysis. Addiction 2015;110:1443–52. of the translations (including but not limited to local regulations, clinical guidelines, 19 Kehoe T, Gmel G, Shield KD, et al. Determining the best population-­ terminology, drug names and drug dosages), and is not responsible for any error level alcohol consumption model and its impact on estimates of and/or omissions arising from translation and adaptation or otherwise. alcohol-­attributable harms. Popul Health Metr 2012;10:6. 20 Trias-­Llimós S, Martikainen P, Mäkelä P, et al. Comparison of different Open access This is an open access article distributed in accordance with the approaches for estimating age-­specific alcohol-­attributable mortality: Creative Commons Attribution Non Commercial (CC BY-­NC 4.0) license, which the cases of France and Finland. PLoS One 2018;13:e0194478. permits others to distribute, remix, adapt, build upon this work non-­commercially, 21 Chrystoja BR, Rehm J, Manthey J, et al. A systematic comparison and license their derivative works on different terms, provided the original work is of the global comparative risk assessments for alcohol. Addiction properly cited, appropriate credit is given, any changes made indicated, and the use 2021;116:2026–38. is non-­commercial. See: http://creativecommons.org/licenses/by-nc/4.0/. 22 Martikainen P, Mäkelä P, Peltonen R, et al. Income differences in life expectancy: the changing contribution of harmful consumption of alcohol and smoking. Epidemiology 2014;25:182–90. ORCID iDs 23 Polednak AP. Surveillance of US death rates from chronic diseases Sergi Trias-­Llimós http://orcid.org/0000-0002-8052-6736 related to excessive alcohol use. Alcohol Alcohol 2016;51:54–62. Jeroen J. A. Spijker http://orcid.org/0000-0002-3957-9553 24 White AM, Castle I-­JP, Hingson RW, et al. Using death certificates to explore changes in alcohol-­related mortality in the United States, 1999 to 2017. Alcohol Clin Exp Res 2020;44:178–87. 25 Durkin A, Connolly S, O'Reilly D. Quantifying alcohol-­related REFERENCES mortality: should alcohol-­related contributory causes of death be 1 Rehm J, Gmel GE, Gmel G, et al. The relationship between different included? Alcohol Alcohol 2010;45:374–8. dimensions of alcohol use and the burden of disease-­an update. 26 INE. Estadística de defunciones. Asignación de nivel educativo Addiction 2017;112:968–1001. a ficheros de defunciones de 2017. Método de obtención y 2 Mackenbach JP, Kulhánová I, Bopp M, et al. Inequalities in alcohol-­ advertencias a usuarios, 2018. related mortality in 17 European countries: a retrospective analysis of 27 INE. Asignación de nivel educativo, relación Con La actividad laboral mortality registers. PLoS Med 2015;12:e1001909. Y ocupación a ficheros de Movimiento Natural de la Población 3 Alonso I, Vallejo F, Regidor E, et al. Changes in directly alcohol-­ (MNP). Método de obtención y advertencias a usuarios, 2020. attributable mortality during the great recession by employment 28 Donat M, Sordo L, Belza MJ. Mortalidad atribuible al alcohol en status in Spain: a population cohort of 22 million people. J Epidemiol España (2001-­2017). Metodología y resultados [online]. Ministerio Community Health 2017;71:736–44. de Sanidad, 2020. https://pnsd.sanidad.gob.es/profesionales/​ 4 Mateo-­Urdiales A, Barrio Anta G, José Belza M, et al. Changes in publicaciones/catalogo/catalogoPNSD/publicaciones/pdf/2020_​ drug and alcohol-­related mortality by educational status during Mortalidad_atribuible_al_alcohol_en_Espana_2001-2017.pdf the 2008-­2011 economic crisis: results from a Spanish longitudinal 29 Moreno-­Betancur M, Latouche A, Menvielle G, et al. Relative index study. Addict Behav 2020;104:106255. of inequality and slope index of inequality: a structured regression 5 Angus C, Pryce R, Holmes J, et al. Assessing the contribution of framework for estimation. Epidemiology 2015;26:518–27. alcohol‐specific causes to socio‐economic inequalities in mortality in 30 Preston S, Patrick H, Michel G. Demography: measuring and England and Wales 2001–16. Addiction 2020;115:2268–79. modeling population processes. Blackwell Publications, 2001. 6 Pechholdová M, Jasilionis D. Contrasts in alcohol-­related mortality 31 Aburto JM, Kashyap R, Scholey J. Estimating the burden of in Czechia and Lithuania: analysis of time trends and educational COVID-­19 on mortality, life expectancy and lifespan inequality in differences. Drug Alcohol Rev 2020;39:846–56. England and Wales: a population-­level study. medRxiv 2020. 7 Probst C, Roerecke M, Behrendt S, et al. Socioeconomic 32 Tsai SP, Lee ES, Hardy RJ. The effect of a reduction in leading differences in alcohol-­attributable mortality compared with all-­cause causes of death: potential gains in life expectancy. Am J Public mortality: a systematic review and meta-­analysis. Int J Epidemiol Health 1978;68:966–71. 2014;43:1314–27. 33 Haeberer M, León-­Gómez I, Pérez-­Gómez B, et al. Social inequalities 8 Probst C, Kilian C, Sanchez S, et al. The role of alcohol use and in tobacco-­attributable mortality in Spain. The intersection between drinking patterns in socioeconomic inequalities in mortality: a age, sex and educational level. PLoS One 2020;15:e0239866. systematic review. Lancet Public Health 2020;5:e324–32. 34 Soler-­Vila H, Ortolá R, García-­Esquinas E, et al. Changes in alcohol 9 Wood S, Bellis M. Socio-­economic inequalities in alcohol consumption and associated variables among older adults in Spain: consumption and harm: evidence for effective interventions and a population-­based cohort study. Sci Rep 2019;9:1–11. policy across EU countries. Bruss Eur Comm 2017. 35 Rogers EM. Diffusion of preventive innovations. Addict Behav 10 Bellis MA, Hughes K, Nicholls J, et al. The alcohol harm 2002;27:989–93. paradox: using a national survey to explore how alcohol may 36 Schiaffino A, Fernandez E, Borrell C, et al. Gender and educational disproportionately impact health in deprived individuals. BMC Public differences in smoking initiation rates in Spain from 1948 to 1992. Health 2016;16:111. Eur J Public Health 2003;13:56–60. 11 Lewer D, Meier P, Beard E, et al. Unravelling the alcohol harm 37 Trias-­Llimós S, Bosque-­Prous M, Obradors-­Rial N, et al. Alcohol paradox: a population-­based study of social gradients across very and educational inequalities: hazardous drinking prevalence and all-­ heavy drinking thresholds. BMC Public Health 2016;16:1–11. cause mortality by hazardous drinking group in people aged 50 and 12 Östergren O, Martikainen P, Lundberg O. The contribution of older in Europe. Subst Abus 2020:1–9. alcohol consumption and smoking to educational inequalities in life 38 León-­Muñoz LM, Galán I, Donado-­Campos J, et al. Patterns of expectancy among Swedish men and women during 1991-­2008. Int alcohol consumption in the older population of Spain, 2008-­2010. J J Public Health 2018;63:41–8. Acad Nutr Diet 2015;115:213–24. 8 Trias-­Llimós S, Spijker JJA. BMJ Open 2022;12:e053205. doi:10.1136/bmjopen-2021-053205 Open access 39 Kuh D, Ben-­Shlomo Y, Lynch J, et al. Life course epidemiology. J 48 Trias-­Llimós S, Kunst AE, Jasilionis D, et al. The contribution of Epidemiol Community Health 2003;57:778–83. alcohol to the East-­West life expectancy gap in Europe from 1990 40 Ross CE, Wu CL. Education, age, and the cumulative advantage in onward. Int J Epidemiol 2018;47:731–9. health. J Health Soc Behav 1996;37:104–20. 49 Manthey J, Rehm J. Mortality from alcoholic cardiomyopathy: 41 Dupre ME. Educational differences in age-­related patterns of exploring the gap between estimated and civil registry data. J Clin disease: reconsidering the cumulative disadvantage and age-­as-­ Med 2019;8:1137. leveler hypotheses. J Health Soc Behav 2007;48:1–15. 50 Rehm J. Alcohol use is a key factor in recent decreases in life expectancy in the United States. Alcohol Clin Exp Res 42 Mishra GD, Ball K, Dobson AJ, et al. Do socioeconomic gradients 2020;44:404–6. in women’s health widen over time and with age? Soc Sci Med 51 Sheron N. Alcohol and liver disease in Europe – Simple measures 2004;58:1585–95. have the potential to prevent tens of thousands of premature deaths. 43 Galán I, González MJ, Valencia-­Martín JL. Patrones de consumo J Hepatol 2016;64:957–67. de alcohol en España: un país en transición. Rev Esp Salud Publica 52 Fierro I, Yáñez JL, Álvarez FJ. Mortalidad prematura Y años 2014;88:529–40. potenciales de vida perdidos relacionados Con El consumo de 44 León-­Muñoz LM, Guallar-­Castillón P, Graciani A, et al. Adherence to alcohol en España Y en Las comunidades autónomas en El año the Mediterranean diet pattern has declined in Spanish adults. J Nutr 2004. Atención Primaria 2010;42:95–101. 2012;142:1843–50. 53 Room R, Stigma RR. Stigma, social inequality and alcohol and drug 45 Patra J, Taylor B, Irving H, et al. Alcohol consumption and the risk use. Drug Alcohol Rev 2005;24:143–55. of morbidity and mortality for different stroke types - a systematic 54 Flagg LA, Anderson RN. Unsuitable underlying causes of death for review and meta-­analysis. BMC Public Health 2010;10:1–12. assessing the quality of cause-­of-­death reporting. Natl Vital Stat Rep 46 Shield K, Manthey J, Rylett M, et al. National, regional, and global 2021;69:1-­25. 55 Mikkelsen L, Phillips DE, AbouZahr C, et al. A global assessment of burdens of disease from 2000 to 2016 attributable to alcohol civil registration and vital statistics systems: monitoring data quality use: a comparative risk assessment study. Lancet Public Health and progress. Lancet 2015;386:1395–406. 2020;5:e51–61. 56 Pinto Pastor P, Santiago-­Saéz A, Guijarro-­Castro C, et al. Completion 47 GBD 2017 Risk Factor Collaborators. Global, regional, and national of the medical certificate of cause of death in Madrid: a descriptive comparative risk assessment of 84 behavioural, environmental cross-­sectional study. Rev Clin Esp 2020;220:215–27. and occupational, and metabolic risks or clusters of risks for 195 57 Barbería E, Xifró A, Arimany-­Manso J. Beneficial impact of forensic countries and territories, 1990-­2017: a systematic analysis for the sources to cause-­of-­deaths statistics. Spanish Journal of Legal global burden of disease study 2017. Lancet 2018;392:1923–94. Medicine 2017;43:1–4. Trias-­Llimós S, Spijker JJA. BMJ Open 2022;12:e053205. doi:10.1136/bmjopen-2021-053205 9