Variation in Survey Coverage of Alcohol Consumption by Respondent Sociodemographic and Drinking Characteristics and by Beverage Characteristics: Findings from the CHEAP Study

Introduction: Although survey data is the most practical and common way to assess alcohol consumption in the population, survey ‘coverage’ of alcohol consumption, meaning the ratio of survey-based consumption to that based on sales data, is typically less than 50%. Furthermore, since sales data does not reflect consumption patterns within the population, it is not possible to understand the degree to which coverage varies based on sociodemographic or drinking characteristics. Methods: Alcohol consumption coverage will be assessed by comparing population-weighted product-level alcohol consumption from the Costs Harms, Expenditures and Alcohol Prices (CHEAP) survey (consisting of 1217 BC residents aged 19+) to complete product-level sales data from the British Columbia Liquor Distribution Branch. Specifically, for each of the741 products consumed by respondents, we calculated a coverage ratio based on per capita survey:sales consumption. We then estimated relative survey:sales coverage by socio-demographic and alcohol consumption characteristics based on the average coverage ratio based on the composition of products consumed by each group. Results: We predict that (i) estimates of alcohol per capita (APC) based on product-specific alcohol use measures from the CHEAP survey will be lower than those based on sales data from the BC Liquor Distribution Branch; (ii) there will be significant variability in survey:sales coverage on the basis of socio-demographic factors (e.g., by sex, age), alcohol consumption patterns (weekly consumption, reported binge alcohol use), and beverage characteristics (e.g., type of alcohol, price per standard drink); (iii) on a per-person basis, the greatest absolute under-coverage (expected consumption based on product-specific sales minus survey report) will be largest among men, younger persons, those with high consumption or who have higher AUDIT scores, and for products with lower prices; (iv) on a population level, a substantial fraction of under-coverage will be accounted for by those drinking moderate amounts of alcohol because they represent a large percentage of those who consume alcohol. Discussion: Results will be relevant to understanding patterns of alcohol consumption based on survey data and can help with adjustment of consumption data that is used to model the impacts of alcohol consumption, including analyses using population-attributable fraction methodology.