Introduction: Alcohol is a major contributor to the burden of disease globally. In Europe, there is a long-standing tradition of drinking alcohol, with per capita consumption being the highest in the world, even if it has been declining in recent decades. Changes in per capita consumption are likely to be related to factors operating at societal level, of which globalisation has been one of the most important factors. Since early 2020, the global pandemic of the Coronavirus Disease 2019 (COVID-19) has caused serious disruptions to day-to-day and community life. As a consequence of the pandemic and the measures taken to contain the spread of the virus, patterns and levels of alcohol consumption are expected to change due to reduced alcohol availability, and according to income levels and distress experiences. Surveys are used to assess such short-term trends, but also to identify drinking patterns that cannot be derived from other sources such as alcohol purchases. However, general population surveys usually fail to capture the entire amount of alcohol consumed within a population. This inherently limiting factor of surveys is called underreporting and not well understood to date. Aims and Objectives: The overall aim of this dissertation was to provide a comprehensive and up-to-date assessment of alcohol consumption in Europe using survey data, as well as to compare survey-based annual consumption estimates with per capita consumption data, i.e., sales and other data derived. For this purpose, the following three objectives were of interest: First, to determine and to compare European drinking practices across 19 countries using survey data from 2015 (study I). Second, to examine changes in alcohol consumption during the first months of the COVID-19 pandemic in 21 European countries and whether these changes were associated with income and pandemic-related distress experiences (study II). Third, to estimate the degree of under-reporting of alcohol consumption in surveys and to study possible factors related to under-reporting based on 39 surveys from 23 European countries (study III). Design: All studies are based on individual-level data from cross-sectional pan-European surveys covering the general adult population. In study I, European drinking practices were determined by the means of latent class analysis using key indicators of alcohol consumption (past-year drinking frequency and quantity, risky single-occasion drinking, and preferred alcoholic beverage) derived from nationally representative data from 2015. In a second step, the identified drinking practices were compared across countries. In study II, self-reported changes in drinking frequency, drinking quantity, and the frequency of heavy episodic drinking over the past month were assessed in a large convenience sample collected between April and July 2020. Additionally, past-year alcohol consumption was recorded using the consumption items of the Alcohol Use Disorder Identification Test in order to obtain an estimate of alcohol consumption during the pre-pandemic period. Income and distress experiences were recorded as factors hypothesised to be associated with self-reported changes in consumption. In Study III, nationally representative surveys conducted between 2008 to 2015 were combined and compared with alcohol per capita consumption estimates for the same years and countries, in order to estimate survey-based under-reporting of alcohol consumption at the national level. Per capita consumption estimates were retrieved from sales statistics and corrected for unrecorded and tourist consumption. Data sources: Survey data were obtained from large-scale alcohol surveys covering the majority of European Union countries plus some non-European Union countries. Statistical analyses: To identify European drinking practices (study I), survey-weighted latent class models of key indicators of alcohol consumption, adjusted for respondent’s location, were conducted. The resulting class prevalence estimates were then contrasted across locations, and fractional response regression models were calculated for membership probabilities of each class (dependent variable) in explorative analysis, with sociodemographic data and individual alcohol harm experiences serving as independent variables. To identify changes in overall alcohol consumption over the past month during the COVID-19 pandemic (study II), a consumption-change score was calculated by combining the single change indicators (i.e., drinking frequency, drinking quantity, and frequency of heavy episodic drinking; range: -1 to +1). Multilevel linear regressions with the random intercept country of residence were used to test for statistical significance of the consumption-change score (dependent variable), adjusting for sample weights and past-year alcohol consumption, as well as for associations with income group and pandemic-related distress experiences. To quantify the extent to which national surveys assess alcohol per capita consumption (study III), coverage rates were calculated. Coverage is defined by dividing the survey-based annual alcohol consumption estimates by per capita consumption estimates. Using fractional response regressions, differences in coverage rates (dependent variable) across transnational regions, as well as the relative importance of heavy episodic drinking prevalence and indicators of survey methodology were tested. Results: Three latent classes of drinking practices were identified: a ‘light to moderate drinking without risky single-occasion drinking’ class (prevalence: 68.0%, 95% Confidence interval [CI]: 66.7, 69.3), a ‘infrequent heavy drinking’ class (prevalence: 12.6%, 95% CI: 11.5, 13.7), and a ‘regular drinking with at least monthly risk single-occasion drinking’ class (prevalence: 19.4%, 95% CI: 18.1, 20.9). In contrast to the former drinking practice of light to moderate drinking, the latter two were characterised by a high average amount of alcohol consumed per drink day, exceeding 100 grams pure alcohol and 60 grams pure alcohol per drink day, respectively. Regional variations in the distribution of drinking practices were observed, with the ‘light to moderate drinking without risky single-occasion drinking’ class prevailing particularly in southern European countries. In study II, an average decrease in overall alcohol consumption during the first months of the COVID-19 pandemic indicated by a negative consumption-change score was observed (-0.14, 95% CI: -0.18, -0.10). This average decline according to the consumption-change indicator was observed in all countries studied, except Ireland (no change) and the United Kingdom (increase). The most marked decrease was reported in the frequency of heavy episodic drinking (-0.17, 95% CI: -0.20, -0.14). Respondents with low- or average income, as well as those experiencing distress were more likely to increase their alcohol consumption than to decrease it. In study III, the average total alcohol consumption coverage was 36.5% (95% CI: 33.2, 39.8), with large variations in coverage rates between countries. While the prevalence of heavy episodic drinking explained up to 10% of the variance in coverage, there were no systematic variations across European regions or the survey methodologies studied, such as non-response rates. Conclusions: In many European countries, a combination of a light to moderate and risky drinking practices prevailed. With a combined prevalence of 32.0%, almost every third respondent engaged in some form of risky drinking. During the first months of the COVID-19 pandemic, overall alcohol consumption, and particularly heavy episodic drinking, appeared to have declined on average in the countries studied. On the other hand, people with low- or average incomes and those suffering from pandemic-related distress were at increased risk of escalating their alcohol consumption during this period. The results contribute both to an up-to-date assessment of alcohol consumption in Europe, including the identification of populations at increased risk of alcohol-related harm, and to the body of evidence on reducing alcohol availability as an effective measure to lower population-level consumption. Beyond consequences for alcohol policy, benefits and limitations of surveys in the field of alcohol epidemiology are discussed. In light of the substantial and apparently unsystematic under-reporting of alcohol consumption in surveys, a rethinking of this methodology is required. Presenting survey-based data and its limitations in an upfront manner, as well as applying sophisticated statistical methods are two options to address current challenges.:Statement for a publication-based dissertation I Table of contents III List of tables V List of figures VII Abbreviations IX Synopsis XI 1 Introduction 1 1.1 Determinants of alcohol consumption and alcohol-related harm 3 1.1.1 Individual vulnerability factors 4 1.1.2 Societal vulnerability factors 5 1.2 The impact of the COVID-19 pandemic on alcohol consumption 8 1.3 Strengths and limitations of survey research 11 2 Aims and objectives 13 3 Study design and methodology 14 3.1 The Standardised European Alcohol Survey 15 3.1.1 The RARHA SEAS questionnaire 15 3.1.2 Sampling procedure and data pre-processing 16 3.1.3 Calculation of annual alcohol consumption 17 3.1.4 Research ethics 17 3.2 Harmonising Alcohol-related Measures in European Surveys 17 3.2.1 Questionnaires 18 3.2.2 Sampling procedures and harmonising of the questionnaires 18 3.2.3 Research ethics 19 3.3 The Alcohol Use and COVID-19 Survey 19 3.3.1 Survey development and design 19 3.3.2 Survey dissemination 21 3.3.3 Calculation of survey and population weights 22 3.3.4 Research ethics 24 4 Study I – A new perspective on the European drinking culture: a model-based approach to determine variations in drinking practices across 19 European countries 25 4.1 Abstract 25 4.2 Introduction 26 4.3 Methods 27 4.4 Results 30 4.5 Discussion 35 5 Study II – Alcohol consumption during the COVID-19 pandemic in Europe: a large-scale cross-sectional study in 21 countries 38 5.1 Abstract 38 5.2 Introduction 39 5.3 Methods 40 5.4 Results 42 5.5 Discussion 49 6 Study III – Why is per capita consumption underestimated in alcohol surveys? Results from 39 surveys in 23 European countries 53 6.1 Abstract 53 6.2 Introduction 54 6.3 Methods 56 6.4 Results 62 6.5 Discussion 64 7 General discussion 69 7.1 Summary of findings 69 7.2 Strengths and limitations 70 7.3 Implications for future research 72 7.3.1 New impulses for drinking culture research 72 7.3.2 The realisation of multi-country online surveys 73 7.3.3 Consequences of low alcohol consumption coverage in alcohol surveys 74 7.4 Implications for alcohol policy 75 7.4.1 Altered availability of alcohol as a secondary outcome of the COVID-19 pandemic 76 7.4.2 Towards a comprehensive alcohol control policy for Europe 78 7.4.3 Alcohol screening and access to health care in high-risk drinking populations 7.5 Outlook 81 7.6 Conclusion 83 8 References 85 9 Appendix 106 9.1 Appendix A (study I) 106 9.2 Appendix B (study II) 126 9.3 Appendix C (study III) 152 10 Erklärung gemäß §5 der Promotionsordnung 160

Alcohol is the most widely used psychoactive drug on earth and continues to be responsible for a substantial burden of death and disability. Mitigating these harms is an important focus of any healthful society. Population-level alcohol policy strategies may be employed to decrease these harms and improve population health. To assist towards these goals, this dissertation has two research objectives relating to the estimation and mitigation of alcohol harms: (1) to complete a series of studies regarding the Alcohol Total Consumption Model (TCM) and (2) to specify and test a novel alcohol health harms estimator and alcohol policy scenario modeler, the International Model of Alcohol Harms and Policies (InterMAHP). The TCM is an important theory in alcohol studies and connects alcohol policies, per capita alcohol consumption and alcohol-attributable (AA) harms in a unified social theory. In brief, policies are expected to reflect on population-level consumption, which in turn is the most important predictor of alcohol harms. The TCM theorizes that change should flow directionally through the model – a policy expected to decrease consumption would be predicted to decrease alcohol harms. This theory has been critical towards informing alcohol control policies in the past five decades. In this dissertation, a series of studies were conducted to test the assumptions of the TCM, to test their continued viability. Study A is a comprehensive systematic review and series of meta-analyses that established the link between alcohol policies influencing day/hours of sale and outlet density and per capita consumption. Study B is a primary research study that examined the direct effect of a changed alcohol policy on alcohol-related ED visits, in the context of Saskatchewan. Studies C and D establish the link between alcohol consumption and AA mortality and morbidity through mathematical specification of InterMAHP. Next, the model was applied to the exemplar of AA mortality in Canada in 2016. Last, Study E extended InterMAHP functionalities to include modeling changes in AA harms expected from potential or realized per capita consumption changes resulting from policy change. An application was provided in the context of Québec. The results of this dissertation research provide some support, in a modern context, to the relationships defined in the TCM. The findings suggest that the TCM continues to be a largely appropriate conceptual model in consideration of alcohol policy-making. InterMAHP provides global alcohol researchers with a novel model towards estimating the health harms of alcohol. / Graduate / 2020-04-09

Aims: Trends in morbidity and mortality, fully or partially attributable to alcohol, for adults aged 18–64 were assessed for Germany. Methods: The underestimation of population exposure was corrected by triangulating survey data with per capita consumption. Alcohol-attributable fractions by sex and two age groups were estimated for major disease categories causally linked to alcohol. Absolute numbers, population rates and proportions relative to all hospitalizations and deaths were calculated. Results: Trends of 100% alcohol-attributable morbidity and mortality over thirteen and eighteen years, respectively, show an increase in rates of hospitalizations and a decrease in mortality rates. Comparisons of alcohol-attributable morbidity including diseases partially caused by alcohol revealed an increase in hospitalization rates between 2006 and 2012. The proportion of alcohol-attributable hospitalizations remained constant. Rates of alcohol-attributable mortality and the proportion among all deaths decreased. Conclusions: The increasing trend in mortality due to alcohol until the mid-1990s has reversed. The constant proportion of all hospitalizations that were attributable to alcohol indicates that factors such as improved treatment and easier health care access may have influenced the general increase in all-cause morbidity. To further reduce alcohol-related mortality, efforts in reducing consumption and increasing treatment utilization are needed.