19
ABSTRACT
Background: Minimum price of alcohol is one of the proposed set of alcohol policies in many high income countries. However, the extent to which alcohol-related harm is associated with minimum prices across socioeconomic groups is not known.
Methods: Using Finnish national registers in 1988-2007, we investigated, by means of time-series analysis, the association between minimum prices for alcohol overall, as well as for various types of alcoholic beverages, and alcohol-related mortality, among men and women aged 30-79 years across three educational groups. We defined quarterly aggregations of alcohol-related deaths, based on a sample including 80% of all deaths, in accordance with information on both underlying and contributory causes of death.
Results: About 62,500 persons died from alcohol-related causes during the 20-year follow-up. The alcohol-related mortality rate was more than three-fold higher among those with a basic education than among those with a tertiary education. Among men with a basic education, an increase of 1% in the minimum price of alcohol was associated with a decrease of 0.03% (95% confidence interval = 0.01%–0.04%) in deaths per 100,000 person-years. Changes in the minimum prices of distilled spirits, intermediate products, and strong beer were also associated with changes in the opposite direction among men with a basic education and among women with a secondary education, whereas among the most highly educated there were no associations between the minimum prices of any beverages and mortality. Moreover, we found no evidence of an association between lower minimum prices for wine and higher rates of alcohol-related mortality in any of the population sub-groups.
Conclusions: The results reveal associations between higher minimum prices and lower alcohol-related mortality among men with a basic education and women with a secondary education for all beverage types except wine.
Systematic reviews indicate a negative association between the price of alcohol and rates of alcohol-related mortality and morbidity.1,2 They suggest, for example, that doubling the alcohol tax could reduce alcohol-related mortality by an average of 35%.1 In particular, setting the minimum price of alcohol is one of the policies proposed by governments and policy advocates in Europe and elsewhere to reduce alcohol-related harm.3 The major reasoning behind these proposals is a view that minimum unit pricing would reduce harm among the most harmful drinkers. 4
Research on the association between minimum prices and consequences related to harmful drinking is scarce. Research in British Columbia shows associations between an increase in the average minimum price of all alcoholic beverages and reductions in alcohol consumption,5 wholly alcohol-attributable deaths,6 and chronic and acute alcohol-attributable hospitalizations.7 In another Canadian province (Saskatchewan), raising the minimum price of products with a higher alcohol content within a particular type of beverage group brought a shift in consumption toward less strong products.8 However, these studies did not determine which population sub-groups were affected most by the minimum pricing.
There is documented evidence of marked differences in alcohol-related mortality among socio-economic groups.9,10 In addition, according to a study conducted in Finland, the greatest impact on mortality in response to a large reduction in alcohol prices resulting from a reduction in excise taxes, was on those with a low level of education and the non-employed,9 groups that already suffered the highest levels of alcohol-related harm before the legislative change. Similarly, a recent policy-appraisal modelling exercise carried out in the UK assessed that minimum unit pricing would have the largest impact on consumption and alcohol-related harms among harmful drinkers with the lowest incomes and those in routine or manual worker groups.11 We therefore hypothesize that those with the lowest socioeconomic status who have the lowest incomes and who suffer most from alcohol-related harm could benefit most from the use of minimum prices.
Studies focusing on specific beverages have shown that the association between alcohol and health may vary across beverage types. A study examining the association between beverage-specific per capita consumption and cirrhosis mortality in a group of primarily beer-drinking countries, for example, reported an association with the consumption of spirits, rather than beer or wine.12 A recent meta-analysis confirmed the J-shape association between wine consumption and vascular risk, and provided evidence of a similar relationship for beer but not for spirits.13 These studies do not necessarily imply causal effects and some variation between study contexts have been observed. Nevertheless, we expect that the association between minimum prices for different types of beverages and harm related to alcohol use varies so that it is weaker for wine and stronger for spirits. This is mainly because in Finland spirits are somewhat more often consumed by heavy drinkers and by middle-aged men, who are at higher risk of alcohol-related mortality, and spirits are also much more common in heavy-drinking situations.14
In order to investigate these matters further, we analysed data on the lowest prices for alcoholic beverages from the records of the national alcohol monopoly, and further data from the nationwide register of causes of death in Finland. Our aim was to determine, by means of time-series analysis, the association of minimum prices for alcohol, overall and for different types of alcoholic beverages, with alcohol-related mortality among men and women across three educational groups.
METHODS
Overview of the research design
We address the association between minimum prices for alcohol, overall and for several types of alcoholic beverages, and alcohol-related mortality among men and women aged between 30 and 79 across three educational groups. To do this we created quarterly (4-month period) time-series data on alcohol-related mortality using individual-level Finnish national registers and minimum price data obtained from alcohol monopoly price catalogues for the period 1988-2007. These data were analyzed by means of vector-autoregressive time-series modeling.
Study population
Our mortality data were based on Statistics Finland Labour Market data covering all Finns with a linkage to death records during the period 1 January 1988 to 31 December 2007. In accordance with regulations on data protection concerning living individuals, Statistics Finland provided only an 11% sample of the whole dataset. In order to maintain power in the mortality analyses, we further obtained an oversample of those who died in the period (for whom the regulations are less strict) and thus covered 80% of all deaths in that period. We restricted the sample we used in this study to persons aged 30-79 years, on the understanding that educational level is more stable after the age of 30, and because alcohol-related deaths are rare at younger and older ages.
Follow-up for alcohol-related mortality
We classified causes of death according to the Finnish edition of the International Statistical Classification of Diseases and Related Health Problems, 9th and 10th revision (ICD-9 and ICD-10). We further defined alcohol-related deaths as those for which there was a reference to alcohol on the death certificate as the underlying or one of the contributory causes. Estimating alcohol-related mortality on the basis of both the underlying and contributory causes yields more versatile and comprehensive data than the standard method based solely on the underlying cause, particularly in Finland, where death certificates record alcohol intoxication as a contributory cause more frequently and accurately than in most other countries.15 Frequent use of medico-legal autopsy is an important factor enabling the proper attribution of alcohol intoxication as a contributory cause of death. In Finland, medico-legal autopsies determined the cause of over 90% of accidental or violent deaths and over 60% of all deaths in people aged under 65 years during the period 1988–2007.16,17
The total pool of alcohol-related deaths we used consists of the following two main categories: (i) the underlying cause of death was a disease attributable to alcohol (listed below) or fatal alcohol poisoning; and (ii) the underlying cause was not related to alcohol, but a contributory cause was an alcohol-attributable disease or alcohol intoxication (ICD-10 code F100). The first group included alcoholic liver disease (K70), mental and behavioral disorders attributable to alcohol (F101–F109), alcoholic cardiomyopathy (I426), alcoholic diseases of the pancreas (K852 and K860), fatal alcohol poisoning (X45) and a few rarely occurring categories (K292, G312, G4051, G621, and G721). In 84% of the cases in the second group, the underlying cause was cardiovascular disease, or an accident or violence.
We used alcoholic liver diseases instead of all liver diseases because there seems not to be a strong tendency to underreport alcoholic cases in Finland: for example, in 2006, 98% of deaths attributable to liver cirrhosis among men aged under 65, and 83% among all men, were classified as alcohol-related on the death certificate.18
From these individual-level data we calculated quarterly mortality rates stratified by sex and education for the period 1988-2007. We based the three educational categories on the highest educational qualification achieved, according to the National Register of Completed Education and Degrees: basic education, secondary education, and tertiary education.
Minimum prices for alcoholic beverages
Time-series data for minimum prices were relatively easily available for Finland because alcoholic beverages containing more than 4.7% alcohol by volume are available only in outlets belonging to the state-owned retail alcohol monopoly, Alko. Until 1995 Alko also defined the allowed price range for mild alcoholic beverages sold in food shops, and in addition price competition was non-existent until the opening of a German supermarket chain in Finland in 2002. Thus the prices in the Alko outlets well depict the lowest prices of medium beer in Finland until 2002. The documented annual per capita consumption through retail outlets in 2007 was six litres for distilled spirits and other strong beverages, 69 litres for beer but only 10 litres for wine. Therefore, Finland could be considered a beer and spirits culture in terms of drinking habits.
We obtained quarterly data on the prices of alcoholic beverages from 1988 to 2007 from Alko’s price catalogs that are published 3 or 4 times a year. These data include the following information on the beverages: the type of beverage, alcohol by volume, the size of bottle or other container, and the price of beverage. We formed datasets of minimum prices for five categories of beverages: distilled spirits (over 30% alcohol by volume), intermediate products (15%-30%), wine (10%-14.9%), strong beer (over 4.7%), and beer of medium strength (2.9%- 4.7%). The data for beer of medium strength covers only the period up to the end of August 2002, the period for which reliable data were available. We defined the minimum price of alcoholic beverages as the lowest price per litre of 100% alcohol by volume for each beverage type in each quarter. Finally, we formed a series for the minimum price for alcohol overall, in other words the minimum value of the specific beverage type in each quarter. For the analysis we adjusted these minimum prices for inflation by dividing the nominal minimum prices by the consumer price index obtained from Statistics Finland.
To ensure that the changes in the lowest prices are not simply a proxy for prices in general, we ran sensitivity analyses using data on real retail price indices obtained from the national statistics. The series are compiled by the National Institute for Health and Welfare by combining alcoholic beverage price indices prepared by Alko Inc. and the grocery shop alcoholic beverage price indices calculated by Statistics Finland using annual shares of the value of consumption as weights.
Statistical analysis
For the time-series analyses we converted alcohol-related deaths to quarterly death rates per 100,000 person-years, separately for men and women, across the three educational groups. When using time-series analyses, standardizing for age is not necessary, as ageing produces trends in time series, and in time-series analysis trends are removed, often by simply differencing the time series. 19 We used vector-autoregressive modeling to analyze the association between the quarterly series of minimum prices and alcohol-related mortality. A vector-autoregressive model is a multivariate autoregressive model that consists of a set of regression equations for a system of two or more variables. All variables in the system are treated as endogenous, which means that they can be both exposures and outcomes. vector-autoregressive modeling makes it possible to describe the interaction of variables through time in a complex multivariate system.19 One of its main features is its ability to investigate causal relationships between variables.20,21
We applied a two-variable vector-autoregressive model in the study including the minimum price of alcohol and alcohol-related mortality in the following two simultaneous equations:
Pricet = β10 + β11 Mortality t–1 + … + β1p Mortality t–p + α11 Price t–1 + … + α1p Price t–p + ε1t
Mortalityt = β20 + β21 Mortality t–1 + … + β2p Mortality t–p + α21 Price t–1 + … + α2p Price t–p + ε2t
where α’s and β’s are unknown coefficients and ε1t and ε2t are error terms. We captured seasonality, when necessary, by including seasonal dummies as exogenous variables (see more details about the statistical model in the eAppendix). We also performed a sensitivity analysis, including an indicator variable for a large reduction in Finnish excise tax on alcohol in March 2004.22 However, the estimates for minimum price changed very little. We also repeated the analyses using real retail price indices of alcohol, instead of minimum prices, to examine whether the changes in the lowest prices are simply a proxy for prices in general.
Vector-autoregressive models require both series to be stationary, indicating that there is no systematic change in mean or variance over time in the series. We ran the Augmented Dickey-Fuller test in order to investigate this. When a series turned out to be non-stationary, we applied differencing to the series. An important preliminary step in vector-autoregressive modelling is to select the order of the autoregression based on the data. The most accurate criterion for the quarterly vector-autoregressive models with a sample size smaller than 120 turned out to be the Schwartz Bayesian information criterion.23 This criterion suggested a lag length of one as appropriate in almost every model, and therefore entered it into all the models. After estimating the vector-autoregressive models we used the Schwartz Bayesian information criterion to determine the choice between competing models. In addition, we calculated orthogonalized impulse response functions that show the percentage change over time in mortality in response to a 1% change in the minimum prices. We used Stata statistical software, version 11.2 MP for all calculations.