Correspondence to: Michael Livingston, Turning Point Alcohol and Drug Centre, Fitzroy, Victoria. Fax: (03) 9416 3420; e-mail: email@example.com
Objective: To examine recent trends in alcohol-related harm and risky drinking in Victoria, Australia.
Methods: The study compiled eight measures of alcohol-related harm from published and unpublished sources, covering data relating to health, crime, alcohol treatment and traffic crashes for the financial years 1999/2000 to 2007/08. In addition, published estimates of short and long-term risky drinking from three-sets of surveys between 2001 and 2007 were examined.
Results: Six of the eight harm indicators substantially increased, while only alcohol-related mortality and single-vehicle night-time crashes remained relatively stable. In particular, rates of emergency presentations for intoxication and alcohol-related ambulance attendances increased dramatically. Contrastingly, survey-derived estimates of the rate of risky-drinking among Victorians were stable over the time-period examined.
Conclusions: Evidence across the data examined suggests significant increases in alcohol-related harm taking place during a period of relatively stable alcohol consumption levels. This disparity may be accounted for by changing drinking patterns among small, high-risk, subgroups of the population.
Implications: The sharply increasing rates of alcohol-related harm among Victorians suggest that changes to alcohol policies focusing on improving public health are necessary.
Alcohol consumption and alcohol-related harm are pressing political issues in Australia, with federal and state governments increasingly focused on addressing binge-drinking, alcohol-related violence and other problems. At the national level, the recently convened National Preventative Health Taskforce included alcohol as one of the three public health issues in need of action (along with tobacco and obesity),1 while the federal government has introduced a special tax on ‘alcopops’ to reduce drinking, particularly among young people.2 In Victoria, the state government recently released the Victorian Alcohol Action Plan,3 which laid out a series of strategies aimed to reduce the harm associated with alcohol in the state. As suggested in the plan, the government has set up a new unit within the Department of Justice (Responsible Alcohol Victoria) to focus on issues around liquor licensing and employed a substantial new force of 40 liquor licensing inspectors.
This political concern is warranted, with alcohol misuse a major health and social problem in Australia. Begg et al.4 estimated that in 2003, 3.2% of the burden of premature death, disease and injury in Australia was due to alcohol, while Collins and Lapsley5 assessed the cost of alcohol misuse to Australian society in 2004/05 as $15.3 billion. However, much of the recent attention to alcohol problems in Australia has been driven by perceptions that these problems are increasing. For example, former Australian Prime Minister, Kevin Rudd, discussed an ‘epidemic of binge-drinking’ and described alcohol problems as ‘getting out of control’,6 suggesting dramatic recent changes in drinking harm. This perception has been reinforced by media coverage suggesting that alcohol-related problems are becoming increasingly common (e.g. references 6–10). Despite the widespread reporting of worsening problems, little recent systematic work has been undertaken to assess the trends in alcohol consumption and related problems either nationally or in Victoria. Victoria is particularly worthy of attention, as in recent years there have been major changes in the availability of alcohol in the state, with numbers of alcohol outlets increasing by around 60% between 1999 and 2007 (unpublished data). This study examines trends between 1999/2000 and 2007/08 in eight indicators of alcohol-related harm and three population survey estimates of risky drinking rates.
There has been some previous work in this area, largely through the National Alcohol Indicators Project (NAIP), which provides estimates of national and state trends in consumption and harm in a series of bulletins and reports (e.g. references 11, 12). The most recent report13 finds an increase of almost 80% in alcohol-related hospitalisations in Victoria between 1996 and 2005, and almost no change in alcohol-related deaths over the same period. More recently, Livingston14 examined Victorian trends in risky drinking and alcohol-related harm among young Victorians (aged 12–24), using hospital and emergency department data and self-reported consumption data from four population surveys, finding no clear trends in risky consumption contrasted with substantial increases in alcohol-specific hospitalisations and emergency department presentations. This study provides a broader view of alcohol-related harm than either of these studies and includes measures of consumption across the entire Victorian population.
The measures for this study were selected based on the indicators outlined in a WHO guide to monitoring trends in alcohol consumption and harm.15 This guide suggests that the risks of identifying spurious trends can be minimised by making use of a variety of different indicators and by including some indicators with direct alcohol attribution (e.g. emergency department presentations for alcohol intoxication) as well as indicators based on proxy measures of alcohol involvement (e.g. single-vehicle night-time road crashes). The specific indicators were selected to cover a broad range of harms, including those related to intoxication (e.g. night-time assaults, emergency presentations for alcohol intoxication) and those related to long-term alcohol consumption (e.g. alcohol-related mortality and morbidity).
This study summarises trends in eight indicators of alcohol-related harm. Harms are presented as rates per 10,000 residents in Victoria (based on Estimated Residential Populations produced by the Australian Bureau of Statistics). Harm trends are examined for financial years beginning 1 July from 1999/2000 through to 2007/08 unless otherwise noted. The indicators used include measures based on proxies of alcohol involvement (e.g. single vehicle night-time motor vehicle crashes), unambiguous direct alcohol measures (e.g. emergency department presentations for alcohol intoxication) and measures based on some combination of the two (e.g. alcohol-related hospital admissions).
These data are based on all hospital separations (including both private and public hospitals) in Victoria. Hospital separations related to alcohol were extracted from all separations based on the ICD-10 codes of the principal diagnosis associated with the separation. This extraction was based on alcohol attributable fractions derived from previous reviews of the literature by English et al.16 and Ridolfo and Stevenson.17 Thus, the alcohol-related hospital admissions reported here include both alcohol-specific diagnoses (e.g. alcoholic liver cirrhosis, alcohol intoxication, etc.) and proportions of other diagnoses to which alcohol contributes (e.g. motor vehicle crashes, stroke, some cancers, etc).
These data are based on the Unit Record Mortality data collected by the Australian Bureau of Statistics. Deaths were attributed to alcohol based on the primary cause of death using the same method of attributable fractions described for hospitalisations. Mortality data are based on calendar year rather than financial year.
Emergency department presentations for intoxication
These data were extracted from the Victorian Emergency Minimum Dataset (VEMD), which includes data from all emergency data presentations in Victoria. Full details of the dataset are presented in a user manual.18 All presentations with an ICD-10 diagnosis code of F10.0 (acute intoxication due to alcohol) were extracted.
Alcohol-related ambulance attendances
Data from all alcohol- and drug-related ambulance attendances in greater Melbourne are collected as part of a broader surveillance project. These data include paramedic assessment of specific alcohol and drug involvement for all ambulance attendances. Data extracted for this study included cases where paramedics reported alcohol consumption by the patient. Due to industrial action, the ambulance attendance dataset is missing several months of data. These missing months were interpolated based on data from the preceding and subsequent years.
The data are based on police records of assault incidents recorded in Victoria Police's incident database (the Law Enforcement Assistance Program, or LEAP). As there is no reliable coding for alcohol involvement in these offences, assaults from Friday and Saturday nights (between 8 pm and 6 am) were used as a proxy for alcohol-related assaults. Data from a previous Victorian study19 suggest that around two-thirds of assaults during these periods are alcohol-related. Data from Victoria Police were not yet available for 2007/08.
Alcohol-related domestic violence
These data were provided by Victoria Police from their LEAP database. Data on ‘family incidents’ (predominantly incidents of domestic violence) are recorded by the investigating officer, with a specific field for alcohol involvement in the incident. While this recording of alcohol involvement is subjective, it provides a contrast to the data for general assault, where alcohol involvement is estimated based on the time of day and week of the incident. Family incidents where alcohol involvement was considered ‘possible’ or ‘definite’ (around 41% of all family incidents in 2006/07) were included as alcohol-related in this study. Data from Victoria Police were not yet available for 2007/08.
Courses of treatment for alcohol
Data on courses of alcohol treatment in Victoria are stored in the Alcohol and Drug Information System (ADIS), which collates information from all community-based specialist drug and alcohol treatment services in Victoria. The annual number of courses of treatment where the primary drug of concern was alcohol was extracted from ADIS to provide an estimate of trends in treatment utilisation for alcohol problems in Victoria.
Single-vehicle night-time motor vehicle crashes
Single-vehicle night-time crashes are widely used as a proxy measure of alcohol-related motor vehicle crashes (e.g. reference 15). In Victoria, data on all motor vehicle accidents resulting in either fatalities or hospital emergency department presentation are collected by police and provided to VicRoads for entry into their Road Network Database, which includes information on the number of vehicles involved in the accident and the time of day and day of week of the accident. Based on previous research in Victoria, slightly different hours are used to assign crashes as ‘night-time’ for rural and metropolitan areas and between weekends and weekdays.19 Crashes occurring within the times identified as high alcohol times and involving only one vehicle were extracted for this study as a proxy for alcohol-related crashes.
Trends in these harm indicators were tested for statistical significance using the ratio of the rate of each indicator in the most recent year to the rate in the earliest year for which data were available. Confidence intervals were calculated using the incidence rate-ratio calculator in Stata 11.0.
Survey estimates of risky drinking
Measures of risky alcohol consumption derived from three different sets of surveys are examined in this study. Survey data from the three studies cover 2001 through to 2007. Due to the difficulties in comparing results from different studies, survey results are only compared within waves of the same study and not across studies. All survey data were taken from published reports, with only the most recent report cited in full for the sake of brevity. Standard errors for survey estimates are not reproduced in this paper, but the statistical significance of any differences observed over time is discussed. Estimates of risky drinking were based on the 2001 NHMRC drinking guidelines20 and separate estimates for short and long-term risk are presented. Male respondents were classified as short-term risky drinkers if they drank in excess of six drinks in a day, 12 or more times a year. The cut-off for females was lower – more than four drinks in a day, 12 or more times a year. For long-term risk, male respondents were classified as risky drinkers if they drank an average of more than four drinks per day, while for females the cut-off was more than two drinks per day.
The National Drug Strategy Household Survey
Data from the National Drug Strategy Household Survey (NDSHS) were available for 2001, 2004 and 2007.21 The NDSHS is a national mixed-methods survey (using both drop-and-collect and telephone interviews for data collection) that has asked consistent questions on alcohol consumption across the three waves under consideration here. Response rates for the NDSHS across the three waves were low but comparable (47%, 46% and 49%), and the Victorian sample size for each wave was sufficiently large for reliable estimates (between 4,800 and 6,300). Data on alcohol consumption were collected using the graduated frequency method,22 which asks respondents to report how often they drank at various levels in the past 12 months. Risky drinking rates from the NDSHS are for Victorians aged 14 and over.
The Victorian Population Health Survey
Data from the Victorian Population Health Survey (VPHS) were available from 2001 through to 2007.23 The VPHS has used consistent alcohol questions from 2002 through to 2007, using simple usual drinking frequency and usual drinking quantity to measure total volume of consumption and then asking a specific question about how often respondents drank in excess of the 2001 NHMRC guidelines. In 2001 this question on drinking pattern was not included, so only long-term risky drinking can be assessed using the 2001 data. Across the seven waves, the participation rate of the VPHS has been reasonably high and consistent (between 61% and 69%) and the sample has been 7,500 every year. The VPHS includes only respondents aged 18 or over, so risky drinking rates for this survey are for adult Victorians.
The National Health Survey
Data from the National Health Survey (NHS) were available for 2001, 2004/05 and 2007/08.24 The NHS has used consistent alcohol questions across the three waves of data collection, based on three-day recent recall diaries of alcohol consumption. The nature of these questions means that only long-term risky drinking is reported for the NHS. Across the three waves examined, the NHS had a consistent and excellent response rate (92%, 89% and 91%) and a Victorian sample in excess of 3,000. Alcohol questions were only asked of respondents aged 18 or over, so risky drinking rates for this survey are for adult Victorians.
Trends in the eight harm indicators are presented in Table 1, along with rate ratios of the rates in the last and first year of data available. Thus, for example, the rate of hospitalisations in 2007/08 was 1.47 times larger than the rate in 1999/2000. Six of these indicators have seen statistically significant increases over the time-period examined, while single-vehicle night-time motor vehicle crashes and alcohol-related mortality have remained relatively stable (with mortality rates declining slightly). The largest increase occurred in alcohol-related ambulance attendances, which have increased by 167% over the past nine years. Emergency presentations for intoxication have almost doubled, increasing by 98%, while night-time assaults, domestic violence, hospitalisations and treatment episodes have all increased more slowly, but have still grown significantly (49% for assaults, 43% for domestic violence, 47% for hospitalisations and 55% for treatment episodes).
Table 1. Trends in the rates (per 10,000) of alcohol-related harm in Victoria, 1999/00 to 2007/08.
Rate ratio (last year/first year)
95% Confidence interval
ADIS treatment episodes
Ambulance presentations (metro Melbourne)
E.D. Presentations (intoxication)
Serious single vehicle night-time accidents
Prevalence of risky drinking
Data on short-term risky drinking are presented in Figure 1. The Victorian Population Health Survey shows a gradually declining trend between 2002 and 2007 (the estimate for 2007 is statistically significantly different from the 2002 estimate), while the National Drug Strategy Household Survey estimates suggest there has been no change in short-term risky drinking across the period.
The trends in long-term risky drinking in Victoria from the three relevant surveys are presented in Figure 2. The results from the National Drug Strategy Household Survey and the Victorian Population Health Survey both suggest that there has been little change in risky drinking over the past seven years, while the National Health Survey data shows a significant increase (from 9.3% to 12.1%) between 2001 and 2004, followed by a non-significant decline to 2007.
Overall, there is little evidence of major changes in risky drinking levels in Victoria over recent years. There is some indication from the VPHS that short-term risky drinking has declined, while the NHS estimates included a significant increase between 2001 and 2004.
It is worth noting that measuring trends in alcohol consumption and alcohol-related problems is inherently problematic. The difficulties involved are described in detail in a WHO guide,15 and largely relate to the unreliability of most measures of alcohol involvement in health and social problems. Thus, in most cases, indicators are based on surrogate measures of alcohol involvement. For example, alcohol-related hospital admissions are based on aetiological fractions, which are used to allocate certain proportions of admissions for various diagnoses as being caused by alcohol (e.g. all admissions for alcohol intoxication are considered alcohol-related, but only some proportion of admissions for assault injuries or breast cancer). Similarly, data recorded outside the health system (e.g. within the police system) seldom include accurate measures of alcohol involvement, so proxy measures such as assaults on Friday and Saturday evenings or single-vehicle night-time crashes are utilised. The main problem with these kinds of measures in monitoring trends is that factors other than alcohol can contribute to trends. For example, reductions in motor vehicle crashes due to campaigns to reduce driver fatigue will produce reductions in single-vehicle night-time crashes without any change in drink-driving behaviour. Despite these uncertainties, taking the eight indicators together, the consistency of the trends identified provides some reassurance that the overall level of alcohol-related harm is actually increasing. Considering the two indicators that showed little increase, single-vehicle night-time crashes are quite likely to have been held down by ongoing road safety campaigns, particularly around driver fatigue, while the slight decrease in rates of alcohol-related mortality may reflect improvements in healthcare over the period examined.
There are reasons to be cautious in interpreting the survey data presented here as well. In particular, response rates for the VPHS and the NDSHS indicate that these surveys are only covering around half the target population. In addition, all three surveys examined exclude people who are difficult to access for research (e.g. those who are homeless or in institutions like hospitals). Thus, there is the potential for the results presented here to reflect the trends only in a subset of the population and not to cover some subgroups at particular risk of harm. It is worth noting that the NHS, which has a response rate of around 90% across the three waves examined here, did estimate a small increase in risky drinking between 2001 and 2004/05.
There is some evidence from elsewhere that survey estimates of trends in alcohol consumption can point to erroneous conclusions, as demonstrated recently in Finland, where a 13% rise in a sales-based per-capita consumption measure over two years was undetected by surveys spanning the period, which found stable consumption.25 This evidence suggests a need for more robust measures of per capita consumption for Victoria, for which there are no viable estimates since sales data by state ceased to be routinely collected in 1997.26 At this point, per capita consumption is only available for Australia as a whole through estimates produced by the Australian Bureau of Statistics.27 These estimates have changed little in the past decade, suggesting relatively stable consumption, although it is possible that trends in Victoria differ from those for the whole country.
Despite the above concerns regarding the reliability of the trends presented, the weight of evidence suggests that Victoria has experienced sharply increasing rates of alcohol-related harm, occurring during a period of relatively stable alcohol consumption. This finding follows previous work examining youth in Victoria, which found increasing harms and generally stable survey estimates of alcohol consumption.14
These findings may reflect changes in drinking behaviour within small, high-risk subpopulations, who are either excluded from surveys or whose number is too small to have much impact on population-based consumption measures, but who are increasingly experiencing alcohol-related harm. This theory has been put forward by Mäkelä and Österberg28 to explain the relatively large impact on alcohol-related problems compared to per-capita consumption of a reduction in alcohol taxes in Finland, and is supported by a number of studies demonstrating higher impacts on harms than on consumption when policies change (e.g. reference 29). In addition, it is worth noting, with respect to the substantial increases in alcohol outlets in Victoria over recent years, that studies examining the impact of changes in alcohol outlet density have produced much more robust findings of an effect on alcohol-related problems than on measures of consumption.30
Thus, it is possible that the ongoing increases in alcohol availability in Victoria have particularly affected the drinking patterns of high-risk drinkers, resulting in a substantial rise in alcohol-related harm without substantially altering population level estimates of drinking behaviour. For example, according to the 2001 National Drug Strategy Household Survey, 3.0% of Victorians aged 14 or older were drinking at extremely risky levels (seven drinks for women, 11 for men) at least once a week. Even if this proportion had increased to 5.0% by 2007 (an increase of 97,000 extremely risky drinkers), the survey estimates would not be statistically significantly different. Thus, increases in particularly problematic behaviour among high-risk groups conceivably take place without being detected even with unbiased survey estimates. However, such changes in just the extreme end of the drinking distribution would not be in accordance with Skog's theory of the collectivity of drinking cultures, which predicts that, in the aggregate, the drinking of high-risk drinkers and of lower-risk drinkers rises and falls together.31 However, in a later comment, Skog acknowledges essentially that this theory is subject to a ceteris paribus qualification. For example, he notes that as a result of changes in economic inequality or in gender roles, “group means could change in different rates and in different directions as societies change”.32 On its face, the current pattern of trends in Victoria suggests that a societal change may have occurred. The long march towards ever greater availability of alcohol, both in terms of declining prices relative to income and in terms of the proliferation of drinking places and cut-price package stores, may be producing a change in the relationship between the drinking of lower-risk and high-risk drinkers, with a net result of considerably more problems per litre of alcohol. Clearly, this theory remains speculative and further research into the cause of these increases in alcohol-related harm needs to be undertaken.
While the precise mechanisms of the differing trends in alcohol-related harm and measures of consumption require further study, the significant and ongoing increases in indicators of harm in Victoria suggest the need for immediate policy responses. As discussed earlier, some initial attempts to reduce alcohol-related harm have been made in Australia and Victoria, but more substantial policy interventions are required to reverse the trends identified here.
Thanks to Paul McElwee foe his assistance with the ambulance attendance data.