Correspondence to: Michael Livingston, Turning Point Alcohol and Drug Centre, 54-62 Gertrude Street, Fitzroy, 3053. Australia. Fax: 03 9416 3420; e-mail: firstname.lastname@example.org
Objective: To examine recent trends in the proportion of young people who drink at risky levels and the rate of alcohol-related harms experienced by young people in Victoria, Australia.
Methods: The study uses published data from a series of surveys that ask questions relating to alcohol consumption to ascertain whether the proportion of young people drinking at risky levels has increased over the time period in which data are available. Alcohol-caused hospital admissions and emergency department presentations for young people are also examined over recent years to assess trends in alcohol-related harms.
Results: The survey data shows mixed results, with no clear trend in the rate of risky drinking among young people. The harms data suggests that rates of alcohol-related harm, particularly acute intoxication, have increased dramatically over recent years.
Conclusions: The relationship between survey-derived estimates of alcohol consumption and rates of alcohol-related harms is not as clear-cut as expected, and raise concerns about the sensitivity of population surveys in detecting changes in harmful drinking patterns.
Implications: The current increasing trends in alcohol-related harms for young people in Victoria suggest the need for immediate public health interventions.
Youth drinking has become an issue of increasing public concern in recent years, with media reports citing a binge drinking epidemic1 and suggesting “worsening under-age binge drinking”.2 This paper examines data from a range of population and sub-population surveys along with data on alcohol-related harms including alcohol-caused hospital admissions and emergency department presentations in an attempt to determine whether youth drinking patterns and related harms have changed significantly in recent years. There have been two recent studies that have explored trends in alcohol consumption and related harms amongst youth in Australia.3,4 Clemens et al.3 examined trends in the National Health Survey and the National Drug Strategy Household Survey, with only a small section exploring the results specifically for youth consumption. Their results highlighted the discontinuities in methodologies between survey waves and, when direct comparisons were available, found little change in youth risky drinking over recent years.3 Chikritzhs et al.4 focused on indicators with a strong alcohol component, including injuries and deaths stemming from traffic accidents, assaults and self-harm. Trends in these indicators were either stable or decreasing. However, these indicators do not strictly measure alcohol-caused harms and it is possible that trends in these kinds of injuries will be influenced by factors unrelated to alcohol.
The use of general population surveys to estimate alcohol consumption and drinking patterns has been widespread, using a diversity of methods and instruments.5 Surveys that attempt to estimate population levels of consumption have typically come up with estimates that are substantially smaller than estimates coming from aggregate data sources (e.g. alcohol sales data). For example, estimates of alcohol consumption from a large population survey of the Australian population covered between 50% and 75% of the consumption estimated from sales data.6 Two main reasons have been put forward for this: under-reporting by respondents (either intentionally or due to lapses of memory) and bias introduced by survey non-response (either due to an inadequate sampling frame or through bias introduced through survey refusals).5 In addition, there has been much debate about the most appropriate way to estimate drinking patterns using a survey6 and there remains variability in how drinking-patterns questions are asked. These differences have been shown to have substantial effects on survey estimates, making comparisons between surveys that use different questions unreliable.7 Furthermore, different surveys use different sampling schemes, different modalities and have widely varying response-rates, all of which make comparisons of data between surveys difficult.
Even when comparisons are restricted to within repeated waves of the same survey, it is possible that trends detected in a series of surveys do not reflect substantive changes in drinking behaviour, instead resulting from changes in the sampling frame (e.g. for telephone surveys, the increasing number of households without a landline telephone), changes in response rates or other, unaccounted for, changes. The few studies that have examined whether survey-detected trends in alcohol consumption are reflected in trends found in other data sources (such as sales data) have produced mixed results. Greenfield et al.8 demonstrated a similar pattern in data from the United States National Alcohol Surveys and trends in aggregate sales data in the USA. Contrastingly, recent studies in Finland have found declining or stable consumption in national surveys taking place while aggregate sales data increased significantly.9,10
Secondary data focusing on alcohol-related harms is also subject to a variety of possible biases. These are discussed in detail in Chikritzhs and Stockwell.11 Most of the problems are related to harm indicators which assume fixed proportions of certain events (e.g. motor vehicle accidents, night-time assaults) are alcohol-related, although even when using strictly alcohol-caused measures such as those used in this paper changes in rates of diagnoses due to social stigma associated with alcohol-caused illnesses can create misleading trends when the underlying rate of harm remains unchanged.
Thus, the trends discussed in this paper should be treated with caution, particularly where there are inconsistent results across the data sources examined.
This paper examines trends in youth risky drinking and alcohol-related harms in Victoria, using survey data for drinking patterns and administrative data for harms. Risky drinking was classified using the National Health and Medical Research Council's guidelines.12 Respondents were classified as risky drinkers if they exceed the short-term drinking guidelines (i.e. if females drink five or more drinks or males drink seven or more drinks). Where the survey data allow it, frequency of risky drinking has also been explored.
For the purposes of this paper, youth are defined as those aged between 10 and 24, broken down into three age categories (based on the data sources available): 12 to 15, 16 to 17 and 18 to 24. The duration of the trends examined also varies based on the source of the data, with the longest set of consistent data running from 1984 through to 2005.
The data sources used in this study are summarised in Table 1. Data from the Australian School Survey of Alcohol and Drugs (ASSAD) and the Victorian Population Health Survey (VPHS) were taken from published survey reports, while data from the National Drug Strategy Household Survey (NDSHS) and Victorian Youth Alcohol and Drug Survey (VYADS) were reanalysed. Where surveys collected data nationally, only results taken from the Victorian subsamples were examined. Due to the different modes of interview, questions used and sampling frames, only estimates from within the same series of surveys were examined for trends. Thus, data from the VYADS (youth, telephone sample) were not compared directly with data from the ASSAD (school students, classroom sample).
Table 1. Summary of data sources.
The National Health Survey was not included in this study as only long-term risky drinking was reported
Australian Secondary Students
Survey of secondary school students
1984 - 2005
Alcohol and Drug Survey
Victorian Youth Alcohol and Drug Survey
Victorian Population Health Survey
National Drug Strategy Household Survey
Combined telephone and drop and collect survey
Victorian Admitted Episodes Dataset
1998/99 - 2005/06
Victorian Emergency Minimum Dataset
Emergency department presentations
1999/00 - 2005/06
Survey estimates are presented with standard errors, either taken from the original published reports (ASSAD and VPHS) or estimated from tables provided in survey technical reports (NDSHS and VYADS). Differences between estimates were tested for statistical significance at the 0.05 level using the adjusted sample sizes suggested by the survey standard errors. It is worth noting that these error margins do not take into account non-response bias, variability in methodologies or issues with the accuracy of respondents’ answers. Therefore, even where two estimates from different surveys are reported to be statistically significantly different, it is possible that bias from these other sources is contributing to the estimated differences.
The alcohol-related harms data come from two sources: the Victorian Admitted Episodes Dataset (VAED) and the Victorian Emergency Minimum Dataset (VEMD). The harm indicators used in both cases relate to events (admissions or presentations) that were wholly caused by alcohol. In other words, the harms examined are those events for which the recorded diagnosis was entirely attributable to alcohol (e.g. alcoholic liver cirrhosis, acute alcohol intoxication etc). The two harms data sources are discussed in more detail in the results section.
Australian Secondary Students Alcohol and Drug Survey (ASSAD)
The ASSAD surveys have been undertaken every three years from 1984 to 2005 (8 waves), with school-level response rates around 70% (although response rates for all waves of the ASSAD surveys were not available). Data for the ASSAD surveys comes from interviews with a random sample of secondary students across Australia (aged between 12 and 17), and is undertaken in the classroom. This causes some complications when assessing trends over the entire period of the surveys, particularly amongst students aged sixteen or over. The proportion of students who continue in secondary schooling after turning 15 has risen sharply over the time frame of the survey, with the proportion of students who stay at school until the end of Year 12 almost doubling between the first and most recent surveys.13 This should be kept in mind when assessing trends in the drinking patterns of 16–17 year olds. The alcohol measures for this survey were derived from a retrospective weekly diary of the respondents drinking. Thus, risky drinking measures for this study are only measured within the week prior to the survey. All risky drinking data on the ASSAD surveys have been compiled from published reports.13 Sample size and response rate information was provided by Dr Victoria White (private correspondence).
Table 2 shows that the proportion of Victorian secondary school students drinking at levels that put them at risk of short-term harms has increased over the twenty years studied, particularly amongst 16 to 17 year olds. Between 1984 and 1993, the proportion of this age group drinking at risky levels increased from 15% to 21%. It has been reasonably stable since 1993, with only minor variation and no clear trend. There has been little change in risky drinking amongst 12 to 15 year olds, with the proportions varying between 3% and 6% over the eight studies.
Table 2. Proportion of Victorian secondary school students who drank 7 or more drinks (for males) or 5 or more drinks (for females) in the previous week - percentages, with standard errors in parentheses, Australian School Students' Alcohol and Drug Survey.
12-15 year olds
16-17 year olds
Victorian Youth Alcohol and Drug Survey (VYADS)
The Victorian Youth Alcohol and Drugs Survey (VYADS) is a telephone survey that focuses only on people aged between 16 and 24 and collects large Victorian samples. The VYADS has been undertaken three times: in 2002, 2003 and 2004. Response rates for the VYADS surveys have not been published, although each wave has reported cooperation rates of around 70%.14–16 The alcohol measures are collected using the standard graduated quantity-frequency questions and have remained unchanged across the three questionnaires, although the survey methodology has changed slightly. In 2002, the survey was run in three waves spread across the year (March, June and September), while the 2003 survey was conducted in two waves (March and December) and the 2004 survey in just one wave (between November and January). It is not entirely clear whether these changes to the timing of the data collection have influenced the survey results. However, some brief analyses (unpublished) of these survey data have not demonstrated substantial effects for the timing of the survey on estimates of risky drinking, suggesting that these methodological changes are not particularly problematic. The survey unit record data were reanalysed for this study.
The data presented in Table 3 show a significant increase in short-term risky drinking between 2002 and 2003 and then no significant change between 2003 and 2004. This increase is particularly marked for young age 16 to 17.
Table 3. Proportion of Victorians aged 16 to 24 who drank 7 or more drinks (for males) or 5 or more drinks (for females) monthly or more often, by sex and age - percentages, with standard errors in parentheses, Victorian Youth Alcohol and Drug Survey.
Victorian Population Health Survey (VPHS)
The Victorian Population Health Survey (VPHS) is a telephone survey of the Victorian adult population that has been conducted annually since 2001, with response rates between 61% and 69%. Since the 2002 survey, graduated quantity frequency measures have been included, allowing estimates for short-term risky drinking to be calculated. The sample for the VPHS includes only those aged 18 or older. All data presented in this section comes from published survey reports.17–21
The VPHS finds little change in the proportion of young people drinking at levels that put them at short-term risk of harm (Table 4). While the estimate of the proportion of females drinking at risky or high-risk levels on a monthly basis has increased from 37% to 43% between 2002 and 2005, this difference was not statistically significant. Similarly, the reduction in annual risky drinking for males from 77% in 2002 to 72% in 2005 is not statistically significant.
Table 4. Proportion of Victorians aged 18 to 24 who drank 7 or more drinks (for males) or 5 or more drinks (for females), by sex and frequency - percentages, with standard errors in parentheses, Victorian Population Health Survey.
Sample size (aged 18-24)
Monthly or more often
At least once a year
Monthly or more often
At least once a year
National Drug Strategy Household Survey
While the National Drug Strategy Household Survey has been running in some form since 1985, only the two most recent surveys (2001 and 2004) collected appropriate data from a large enough sample for this study. These surveys use mixed methods, including both telephone interviews and drop and collect survey forms. The 2001 survey does not include respondents younger than 14, so only the two higher age groups (16–17, 18–24) can be examined across time. Response rates for 2001 and 2004 were 47% and 44% respectively, and sample sizes (for Victorians aged 16–24) were 733 and 676. Data were reanalysed using the Victorian data for the age groups chosen for this study. Due to the relatively small samples of Victorian young people, these data were not disaggregated by gender. The NDSHS data show little change in monthly risky drinking rates, with the proportion of 16 to 17 year olds drinking riskily changing from 31.2% in 2001 to 28.2% in 2004 and the proportion of 18 to 24 year olds changing from 49.0% to 43.9%. Neither of these changes was statistically significant.
Victorian Admitted Episodes Dataset (VAED)
Recent trends in alcohol-related hospital admissions for young Victorians were examined to provide some context to the survey data presented in this report. Only hospital admissions where alcohol was likely to be the primary cause of admission were included in these trends. Specifically, admissions from the following diagnosis groups were counted:
•Mental and behavioural disorders due to use of alcohol
•Alcoholic liver cirrhosis
•Accidental poisoning by/exposure to alcohol
Hospital admissions were extracted from the Victorian Admitted Episodes Dataset (VAED) for the financial years 1998/99 to 2005/06 by gender for three age groups (12–15, 16–17 and 18–24 year olds) and are presented in Figure 1. The VAED is a database maintained by the Victorian Department of Human Services and contains details of all acute hospital separations in Victoria. An overview of the VAED is provided by The Health Data Standards & Systems Unit, Department of Human Services.24 Data are expressed as rates per 10,000 people, and the population data underlying the rates, based on the Australian Bureau of Statistics Estimated Residential Population data25 are presented in Table 5.*
Table 5. Estimated resident populations, Victoria, by age and sex, 1998/99 to 2005/06.
The rate of alcohol-caused hospital admissions for Victorians aged between 16 and 24 have increases substantially in the last eight years. This increase has occurred for both males and females, with the oldest female age group (18–24 years) increasing the most sharply, from a rate of 6.0 per 10,000 persons in 1998/99 to 14.6 per 10,000 persons in 2005/06. There has been little or no change in the rates of alcohol-caused hospital admissions for those aged between 12 and 15. Analysis of the detailed diagnosis codes showed that more than half of the alcohol caused admissions were for ‘Acute Intoxication’ (F10.0) or ‘Dependence Syndrome’ (F10.2), with the increase in admissions driven mainly by increases in these two diagnoses.
Victorian Emergency Minimum Dataset
To complement the analysis of trends in hospital admissions examined above, the trends in presentation to emergency departments were examined. Data for these trends came from the Victorian Emergency Minimum Dataset (VEMD), which keeps details on all presentations at Victorian public hospitals with 24-hour Emergency Departments. These data are summarised in Health Data Standards and Systems Unit.26
The trends presented in Figure 2 are based on presentations with alcohol-caused diagnoses – the same diagnoses that were used in the hospital admissions analysis above. Reliable diagnosis codes were only available from 1999–2000 onwards.
These results mirror the hospitalisation trends, with rapid increases in the alcohol-caused presentation rates of people between the ages of 16 and 24, and little change in the rates for 12 to 15 year olds. Again, the rates for females have increased the most markedly, with presentation rates more than doubling in the 16–17 and 18–24 age groups. An examination of the underlying data shows that this increase in emergency presentations has been driven primarily by an increase in young people being admitted with a diagnosis of ‘Acute Intoxication’ (ICD code ‘F10.0’). This diagnosis made up more than half the cases examined. Admissions for other diagnoses (e.g. ethanol toxicity) have not increased substantially over the same time period.
Only the ASSAD survey provides information for Victorians aged 12 to 15, suggesting little change in rates of risky drinking amongst this age group. Three surveys cover 16 and 17 year olds. The ASSAD survey found no substantial change in risky drinking amongst 16 to 17 year old school students since 1993, while the VYADS found a sharp increase between 2002 and 2003 and the NDSHS found no significant change. The VYADS, NDSHS and VPHS all provided data for 18 to 24 year olds, with only the VYADS finding any significant trend – an increase for both males and females between 2002 and 2003.
Thus, the surveys present a mixed picture of recent trends in risky drinking among Victorian young people. The largest youth survey found significant increases across both genders and age groups examined between 2002 and 2003, while all other data sources found non-significant differences across all age and gender groups in recent years.
Register harms data
The register data utilised in this study have potential weaknesses. Both the VAED and VEMD data may be subject to changes in coding practices. It is impossible to ascertain whether there have been any changes in how hospital staff record diagnoses for alcohol-related presentations/admissions, but the similar trends across the two data sources and the lack of any marked increase for 12 to 15 year olds in either data source suggest that coding practices are not responsible for the observed trends.
The register data examined almost uniformly suggests an upward trend in alcohol-related harm among young Victorians aged 16 and above. The hospital and emergency data show no marked changes among those aged 12 to 15, but substantial increases for young people (both male and female) between 16 and 24. Both data sources show particularly sharp increases amongst females aged 18 to 24. Detailed examination of both data sources highlighted the role of ‘acute intoxication’ diagnoses in the increases in alcohol-related problems observed.
The data summarised in this paper present a mixed view of trends in the risky drinking patterns of young people in Victoria. On the whole, the surveys do not provide a clear and consistent picture of trends in risky drinking amongst young people. There are few significant trends and almost no notable increases in risky drinking in recent years, with only the VYADS data suggesting increased risky drinking. Contrastingly, there is a clear increasing trend in alcohol-related harm among young people across three separate sources of secondary data. These trends are deeply concerning, suggesting that increasing numbers of young people are experiencing severe alcohol related problems such as hospitalisation. It is clear that, if these trends represent real increases in alcohol-related harm among young people, public health intervention is required, with increased alcohol taxes the best supported method of reducing alcohol-related harm.27 In addition, studies have found links between rates of binge-drinking amongst young people and alcohol availability, suggesting reductions in trading hours or premise numbers as worthwhile approaches.28
The contrast of the mixed survey results with the consistent and clear trend in the hospital admissions and emergency department presentations requires some explanation. There are a number of possible explanations that reflect problems with the data sources rather than substantive differences between the surveyed populations and the register data. For example, as mentioned previously, recording practices for hospital databases may have changed in recent years, resulting in increasing numbers of patients coded with alcohol-related diagnoses. This may be related to funding mechanisms based on the casemix being dealt with by hospitals, although the major changes to health funding processes in Victoria took place well before the increases seen in the register data examined here.29 Furthermore, the generally poor response rates of the surveys may have resulted in non-response bias in the estimates of risky-drinking, with non-responders responsible for the increases in the alcohol-related harms.
A substantive reason behind the disparate trends in the survey and register data has been suggested by Mustonen et al.9 following similar findings in Finland, where increases in alcohol sales and alcohol-related mortality were not matched by survey-based measures of alcohol consumption. They suggested that the overall increase in consumption was due to increased drinking amongst a small group of very heavy drinkers – including marginalised people excluded from survey sampling frames. In the context of the data presented in this study, this rationale can be applied in two ways. Firstly, it is possible that the increase in the number of young people who are drinking to the point that they are appearing in hospital records has taken place amongst people who are excluded from the survey samples discussed. For example, they may have left high school before turning 17 and thus be excluded from the ASSAD survey series, or they may live in situations that preclude their selection from the population-based surveys (e.g. they might live in a household with no phone, be homeless or live in an institution). A marked increase in heavy drinking amongst this group is a realistic possibility, particularly given the recent increases in alcohol availability in Victoria30 and the evidence found by Mäkelä et al31 that increased availability most affected the consumption of marginalised drinkers.
Secondly, it may be that the published survey data is hiding increasing rates of extremely heavy drinking, while rates of risky drinking remain stable. In other words, it is possible that, while the proportion of young people drinking at levels that exceed the NHMRC guidelines hasn’t changed markedly, more young people are drinking at extremely high levels and thus ending up in hospital. This is given modest support by slight (but non-significant) increases in the proportion of males drinking 20 or more drinks and females drinking 11 or more drinks over recent VYAD surveys.16 In addition, the VYAD surveys show increasing proportions of young people drinking until they can’t remember what happened.16 Unfortunately, published reports from the other surveys discussed in this paper do not provide comparable information on extremely high levels of alcohol consumption, making it difficult to determine whether this represents what is actually occurring.
If either of these explanations were verified, they would provide some evidence against Skog's theory of the collectivity of drinking cultures,32 which predicts that alcohol consumption changes uniformly across population subgroups. This would therefore imply that increases in alcohol consumption at the high end of the scale or amongst particular subgroups of drinkers, would be reflected in increasing proportions of young people drinking at risky levels, a finding that is not clearly reflected in the survey data discussed. The fact that these explanations cannot be confidently supported or rejected points to a clear need for a more rigorous examination of young people's drinking, particularly focussing on young people whose drinking results in serious health consequences.
The ABS population estimates are provided using five year age groups, which do not match the age groups used in this study. Where necessary, the ABS populations have been redistributed, assuming that the population for a single year of age is equal to one-fifth of the population of the five-year age group it is included in.