4.2. Evaluation of Data and Methods
We carefully assessed current levels and past trends in alcohol-attributable mortality using an estimation approach that deals with important shortcomings of previous estimates. Firstly, compared to previous research that mostly adopted an underlying cause-of-death approach [20
], our estimates—which are largely based on the GBD estimates—include mortality due to alcohol from causes of death partly attributable to alcohol [22
]. Consequently, our estimates are higher than those by Rosén and Haglund (2019) and by Trias-Llimós et al. (2020)—which were only based on causes of death wholly related to alcohol [18
]—and are lower than estimates that include all deaths from causes of death partly attributable to alcohol (e.g., external causes of death), like the estimates by Pruckner et al. (2019) [17
]. Secondly, we adapted the GBD alcohol-attributable mortality rates for 65+ in response to quality concerns due to the limitations of applying their estimation technique at higher ages [20
]. Compared to the very steep increases (men) and steep declines (women) in alcohol-attributable mortality rates with age observed in the GBD data, we obtained an inverted U-shaped curve for both sexes (see Supplementary File 1—Appendix Figure A1
), which is considered more realistic [20
]. Consequently, compared to the GBD, our estimates tend to be lower for men and higher for women (see Supplementary File 1—Appendix Figure A2
), and are more likely to accurately represent the age pattern of alcohol-attributable mortality, and, in turn, its cohort patterns.
The use of a certain estimation technique affects not just past alcohol-attributable mortality levels [20
], but can also affect its past trends and consequently its future levels. For example, the past declines in alcohol-attributable mortality throughout Europe that Pruckner et al. [17
] reported are inconsistent with our current findings and with previous findings [9
] that showed different trends by country. This is likely because Pruckner et al. included mortality from all external causes [17
], which are not all attributable to alcohol, and which declined throughout Europe [3
The above illustrates that the estimation of alcohol-attributable mortality is not straightforward and could considerably affect the outcomes. Therefore, it is essential to critically assess the estimation technique before utilizing its outcomes. Despite our efforts to improve current alcohol-attributable mortality estimates, also our estimates remain estimates based on the current epidemiological evidence on the effects of alcohol on causes of death and age groups; and should be considered as such. Furthermore, we recommend investments in further improvements of estimates of alcohol-attributable mortality, particularly at older ages.
Our advanced approach to projecting alcohol-attributable mortality is, in our view, an important step forward compared to previous methodologies that mainly provided all-age estimates for the short-term future [15
]. That is, our age-period-cohort approach enabled us to take into account important trend breaks due to changes in alcohol consumption patterns and differences between birth cohorts in alcohol-attributable mortality, and to obtain realistic future estimates of age-specific alcohol-attributable mortality. Moreover, in contrast to the previous projections by Sheron et al. and Pruckner et al. [15
], which basically consisted of linear extrapolations of past trends, our projection approach is able to produce plausible long-term outcomes. That is, by transforming the outcome measure, implementing lower bounds, and assuming that increases will eventually turn into declines, we avoided not only generating long-term estimates below zero, but also projecting unlikely future crossovers and large differences in AAMF levels both between sexes and between country groups. Both kinds of outcome can easily occur when linearly extrapolating past trends for different countries with largely different past trends.
All in all, our projection provides society and health policy-makers not only with a more realistic estimate of future alcohol-attributable mortality by incorporating the experiences of other countries and different generations, but also with a more detailed outlook by providing estimates of future age-specific alcohol-attributable mortality. Thus, we have provided comparable estimates covering a longer future time horizon than previous projections. However, as is the case for any projection, our outcomes depend on the underlying assumptions.
First, our assumption that increases in AAMF will be followed by declines, which we made to avoid an unrealistic divergence in AAMF between countries, could be considered overly optimistic. Indeed, this assumption is based on the premise that there will be strong (continued) policy efforts and increased awareness of the harmful effects of alcohol in these countries (see as well Section 4.3
). However, our observations of (i) more favourable cohort patterns for countries with recent period increases, and (ii) of recent declines in or the stagnation of AAMFs for selected ages in most of these countries (e.g., United Kingdom, Lithuania, The Netherlands, Poland) (Supplementary File 1—Table S3; Supplementary Files 2 and 3
), are in line with our general assumption, which is also backed up by trends in other European countries, and by recent alcohol consumption patterns (see Section 2.4
Second, our long-run outcomes are, logically, dependent on the lower bounds we implemented. The implementation of these lower bounds in order to avoid crossovers between the historically higher alcohol-attributable mortality levels among men than among women, and similarly, the higher alcohol-attributable mortality levels among Eastern European men than among Western European men, could be considered conservative, particularly for those countries that currently have high and strongly declining levels of alcohol-attributable mortality.
Overall, however our projections of (eventual) declines with a lower bound seem to reflect the further (decelerating) decline in alcohol consumption for Europe as a whole that was recently projected by Manthey et al. [30
In addition, like all projections, our outcomes come with a degree of uncertainty that our projection intervals do not fully capture. First, our implementation of the lower bounds resulted in relatively small projection intervals that decrease with time. Second, more generally, projection intervals hardly ever fully reflect the full uncertainty of projection outcomes, which emerges from model uncertainty and parameter uncertainty, but also from uncertainty related to the underlying assumptions and choices [31
]. In fact, we consider our projections for men in Eastern European countries more uncertain than those for other populations. For men in Ukraine and Lithuania, in particular, the combination of the assumed lower bound value with the quadratic curve extrapolation resulted in large projected declines, leading to temporal crossovers with Western European countries.
Thus, although we devised a general methodology to project alcohol-attributable mortality realistically into the long-term future, our outcomes are dependent on our assumptions. For selected countries, further methodological advancements or refinements in assumptions based on additional national knowledge would be beneficial. Moreover, for the projection of alcohol-attributable mortality in individual countries, the inclusion in the model of country-specific information on the determinants of (trends in) alcohol-attributable mortality could have added value.
In addition to the issues with the estimation of alcohol-attributable mortality and the importance of our assumptions for our projection outcomes, a third important limitation of our study is that we could not include in our analysis the foreseen, but currently unknown, exact effects of the COVID-19 pandemic. In fact, both declines due to decreases in people’s ability to afford alcohol as a result of an economic downturn and increases due to increases in hazardous drinking to cope with increased unemployment and perceived stress, can be expected [32
4.3. Interpretation of Findings
We observed for the 26 European countries studied, that women have always exhibited lower AAMF levels than men. This gender gap may be related to higher alcohol consumption levels and riskier alcohol consumption patterns (binge drinking, drinking of spirits) among men than among women [33
]. These sex differences in alcohol consumption are generally explained by traditional gender differences in social roles (e.g., women are responsible for performing housework, while men are responsible for generating a labour market income) [34
], but also by sex differences in strategies for coping with stress, with men being more likely than women to deal with stressful situations by consuming large amounts of alcohol [35
]. Mainly for the latter reason, we do not regard it as likely that in the future AAMF levels for men will be lower than those projected for women.
In addition, we observed higher alcohol-attributable mortality in Eastern Europe than in Western Europe, particularly among men, which also reflects differences in alcohol consumption levels and patterns [11
]. Most Eastern European countries have a long tradition of binge drinking, of heavy consumption of vodka and other spirits, and of the homebrewing of spirits, which has resulted in high recorded and unrecorded levels of alcohol consumption, particularly among men, with very detrimental effects on health [38
]. During the economic crisis Eastern Europe experienced in the early 1990s the risky drinking patterns among adult men in particular were aggravated [40
]. Even though Eastern European countries are currently moving more and more towards beer consumption, because of their traditional patterns of heavy drinking, particularly when dealing with difficult socio-economic circumstances, we regard it as unlikely that the higher current age-standardised AAMF values in Eastern European countries for men will become lower than those in Western European countries in the future.
We observed a long-term declining trend in alcohol-attributable mortality among South-Western European countries, and recent declines or stagnating increases in most North-Western and Eastern European countries. The (decelerating) declines in Southern European countries can be linked to the move away from high levels of wine consumption, particularly during meals, and towards an increased consumption of beer, in line with wider societal changes [41
]. The recent declines in alcohol consumption in Eastern Europe have been attributed to the moderate shift away from drinking spirits and towards consuming beer in a context of economic stabilisation, and were reinforced by the implementation of stricter preventive health policies from the mid-2000s onwards (including controls on the production and sale of alcohol in Russia, and increased alcohol taxes in the Baltic countries) [42
]. For the North-Western European countries, the observed (past) increases in AAMF reflect (temporarily) increasing (Finland, Iceland, Ireland, Norway, Sweden, United Kingdom) or stagnating alcohol consumption patterns (Belgium, Denmark, The Netherlands) [7
], which have been attributed to the increased availability and affordability of alcohol [44
], combined with an expanding culture of heavy episodic drinking that is especially dangerous to health [45
]. The implementation of (successful) preventive policies targeting these recent unfavourable patterns (e.g., [27
]), has most likely resulted in the recent stagnation of the increase or the declines in alcohol consumption, and consequently alcohol-attributable mortality, in these countries. We projected that AAMF levels will converge across countries, but also that for men, AAMF levels will be higher in Eastern and South-western European countries than in North-western European countries. The high future levels for Eastern European men are mainly attributable to their high past levels. The high future levels for South-western European men, however, seem related to the deceleration in the decline in their alcohol-attributable mortality (Figure 1
), and their more unfavourable recent cohort patterns (Supplementary File 1—Table S3
). This could result from a levelling off of the decline in wine consumption [7
] and from a recent uptake of unfavourable drinking patterns among youth [46
In addition, we projected declines in all countries, even for selected Eastern and North-Western European populations for whom (stagnating) increasing trends have recently been observed. These projected declines are in line with our general projection approach (see Section 2.4
), with recent indications in the country-specific past trends (see Section 4.2
), but also rely on strong, ongoing efforts aimed at reducing excessive alcohol consumption and its negative health effects. Also, for the remainder of countries, our projections rely on the assumption that the recent favourable trends will continue. Given that these recent favourable trends are at least partly driven by effective public health efforts, continued public health action is required for these countries as well.
For selected Eastern and North-Western European countries in particular, we recommend the more extensive implementation of policies or restrictions that have proven to be effective in reducing alcohol consumption and, in turn, alcohol-attributable mortality. The aim of such policies is generally to reduce the affordability, availability and marketing of alcohol (see [27
] for further details). For example, there is evidence that that countries that have more restrictions on alcohol advertising also have the lowest prevalence of hazardous drinking among middle-aged people [48
], and that setting a minimum price for a unit of alcohol can reduce alcohol consumption among harmful drinkers [49
]. It has also been shown that setting limits on the availability of alcohol (e.g., by limiting the opening hours of shops where alcohol can be sold, or by establishing age restrictions for buying alcohol) are helpful, because the density and the opening hours of outlets that sell alcohol are positively associated with alcohol-related harms [50
However, given that previous declines in alcohol-attributable mortality seem to stem not only from the successful implementation of preventive actions, but from societal changes and shifts in the drinking culture [41
], it is also important to (further) improve social awareness of the health risks of (excessive) alcohol consumption. Among the specific actions that could increase this awareness include the lowering of the norms for alcohol consumption levels in national guidelines [2
], and the inclusion of health warning messages in alcohol labelling [47