Total versus hazardous per capita alcohol consumption as predictors of acute and chronic alcohol-related harm

Abstract

This paper is a preliminary attempt to compare the power of different measures of alcohol consumption to predict variations in rates of alcohol-related harm across 130 adjacent geographical areas comprising the state of Western Australia (WA). It has been argued recently that there may be both scientific and political advantages to replacing, where possible, estimates of per capita total consumption of alcohol with estimates of hazardous consumption.' The arguments favoring this view include: (i) evidence that frequency of heavy drinking occasions predicts self-reported alcohol-related harm in population surveys to a greater degree than does average daily intake;2 (ii) evidence that prevention strategies that overtly target total population consumption are less well received by the voting public than those that target high-risk drinking or actual harm.3 To date this evidence has accumulated mainly from survey data. The present study attempts to examine the issue of predictive power in relation to rates of actual harm as indicated by traffic, health and police statistics for WA and of rates of alcohol consumption derived from both wholesale and survey data. Alcohol consumption is reported in a variety of ways to suit different purposes. For example, surveys frequently report the number of people whose average daily consumption exceeds levels considered to be harmful by Australia's peak medical research body' (e.g., 6% men and 3% women-). We are often told about per capita alcohol consumption-e.g., in 1993 the average Australian adult consumed 9.4 liters of absolute alcohol.6 What is not reported, however, is how much of this consumption is low risk or even beneficial (e.g., in terms of protecting the drinker against ischemic heart disease') and how much would be considered medium or high risk. A method for making such estimates from consumption survey data was described recently, which resulted in the calculation that almost one-third of all alcohol consumed in WA was at a harmful level.8 This method is adapted in this paper to calculate the proportion of total alcohol consumption in 130 regions of WA that is, to use a more conservative term, hazardous. Most other studies of drinking patterns have estimated these in terms of the frequency of heavy drinking occasions-e.g., drinking more than five standard drinks.2 The measure used here combines both frequency and the quantity consumed on such occasions and may thus combine the advantages of measures of both pattern and volume of drinking in one variable. This study is an early product of a research program entitled Measurement of Alcohol Problems for Policy (MAPP). This program brings together for the first time a number of statewide data sets that include measures or likely indicators of alcohol consumption and related harm. The relationships between these various measures and other sociodemographic variables have so far been explored for the financial year 1991-1992.9 The eventual aim of the MAPP project is to identify those indicators that are most sensitive to variations in consumption and harm across different geographical regions and over time. It is hoped this will enable prevention strategies to be better targeted, monitored and evaluated. In the road safety research area, a surrogate measure of alcohol-related crashes has been developed to overcome the problem of estimating alcohol-relatedness: single-vehicle nighttime accidents. These have been shown repeatedly to involve mostly drunk drivers and are now used routinely in the evaluation of major statewide prevention initiatives that target drinking and driving.'o It is likely that similar surrogate measures can be developed for other problem domains. For example, in a study of alcohol use and violence, Ireland and Thommeny reported that 91% of street offenses committed between 10 pm and 2 am involve prior consumption of alcohol. The first report of the MAPP project examined a number of potential harm indicators in the domains of road crashes, hospital morbidity data, drink driving, and assault offenses! …