Undergraduate student drinking and related harms at an Australian university: web-based survey of a large random sample

Background There is considerable interest in university student hazardous drinking among the media and policy makers. However there have been no population-based studies in Australia to date. We sought to estimate the prevalence and correlates of hazardous drinking and secondhand effects among undergraduates at a Western Australian university. Method We invited 13,000 randomly selected undergraduate students from a commuter university in Australia to participate in an online survey of university drinking. Responses were received from 7,237 students (56%), who served as participants in this study. Results Ninety percent had consumed alcohol in the last 12 months and 34% met criteria for hazardous drinking (AUDIT score ≥ 8 and greater than 6 standard drinks in one sitting in the previous month). Men and Australian/New Zealand residents had significantly increased odds (OR: 2.1; 95% CI: 1.9-2.3; OR: 5.2; 95% CI: 4.4-6.2) of being categorised as dependent (AUDIT score 20 or over) than women and non-residents. In the previous 4 weeks, 13% of students had been insulted or humiliated and 6% had been pushed, hit or otherwise assaulted by others who were drinking. One percent of respondents had experienced sexual assault in this time period. Conclusions Half of men and over a third of women were drinking at hazardous levels and a relatively large proportion of students were negatively affected by their own and other students' drinking. There is a need for intervention to reduce hazardous drinking early in university participation. Trial registration ACTRN12608000104358


Background
A high prevalence of hazardous drinking by university students has been reported in many countries [1][2][3] with this population group often drinking more than their non-university/college student peers [4][5][6][7]. In large-scale national surveys in the United States, 37-44% of students report binge drinking (more than five standard drinks per occasion; each containing 12 g ethanol) in the previous two weeks [8][9][10] with men drinking more than women, although this difference has narrowed over time [10][11][12]. Among New Zealand (NZ) university students, 37% have reported one or more binge episodes in the previous week [13].
Factors within the university environment contribute to these high levels of consumption leading to a range of negative consequences [5,7,14]. These include: social, physical and psychological harms to the student e.g. academic impairment, blackouts, injury, suicide, unintended sexual activity and sexual coercion; harm to other people including interpersonal and sexual violence; and costs to the institution such as property damage and student attrition [13,[15][16][17][18][19][20][21]. The secondhand effects of people's drinking on others are also assuming greater importance for advocacy in alcohol control policy, both for the victims experiencing assaults, sexual violence Methods Participants A random sample of 13,000 undergraduate students aged 17-25 years, enrolled full-time and studying on campus at a Western Australian university, were invited to complete a web survey on alcohol consumption, secondhand effects, attitudes toward nutrition/ingredient labeling [68] and tobacco use [69]. Women made up 52.4% of the sample and 20.6% were non-residents. The term 'non-resident' refers to students enrolled at the university that are not permanent residents of Australia or New Zealand and includes those on student visas and humanitarian visas.

Procedure
We adopted a survey recruitment procedure described in detail elsewhere [59,65,67,70]. Four weeks after the start of the first semester of 2007 the University Surveys Office accessed the enrolment database to identify a random sample of 13,000 full-time undergraduates aged 17-25 years. A personally addressed letter from the research team was sent to each student, inviting them to participate in the survey. The letter explained that they would soon receive a hyperlink to the questionnaire in an email message, that responses would be confidential and that the research team was independent of the university administration. Students were offered the opportunity to win one of 40 AU$100 gift vouchers for participating. After 1 week, a reminder email was sent encouraging completion of the questionnaire to those who had not yet responded. A second reminder was sent 10 days later.

Measures
The questionnaire included items on: past alcohol use [71]; current alcohol use [Alcohol Use Disorders Identification Test (AUDIT) [72]]; peak consumption in the previous 4 weeks [73]; height and weight (in order to estimate Blood Alcohol Concentration); secondhand effects of drinking [22]; attitudes toward nutrition/ingredient labelling on alcohol packaging [68]; and tobacco use [74]. The use of standardised instruments for measuring personal use [Alcohol Use Disorders Identification Test (AUDIT) [72]] and secondhand effects [22] make it comparable to studies carried out in other countries. The complete wording and layout of all items can be seen at: http://lamp.health.curtin.edu.au/thrive/ baselinetest.php.
The representativeness of responders to the random sample was assessed using chi-squared tests. The association between participant demographics and being either early or late responder, or to having an AUDIT score ≥ 8, was assessed using chi-squared tests. T-tests were used to compare the mean AUDIT measure for the three subscales (alcohol consumption, dependence and problems) against participant age, sex, and citizenship and to compare total AUDIT score between early and late responders.
The association of secondhand effects experience to participant demographics and frequency of consuming six or more drinks on one occasion (item three in the AUDIT [72]) was also assessed using chi-squared tests. The association between frequencies of consuming six or more drinks (60 g ethanol) on secondhand effects was assessed using multivariable logistic regression after adjusting for gender, age and citizenship. Results are presented as odds-ratio and associated 99% confidence intervals.
The associations between age, sex and citizenship and hazardous drinking were analysed using binary logistic regression. To protect against small effects being considered as being statistically significant due to the large sample size in the study, p-values of < 0.01 will be considered as statistically significant. The assumptions behind the statistical models fitted were assessed to ensure validity of results. This
There was a higher representation of women, Australian/ NZ residents and those aged 17-19 years among the survey respondents compared to the sample (p < 0.001) (see Table 1). The small differences (less than 5%) between early (before the second reminder) and late (after the second reminder) survey responders in relation to age, citizenship and gender were not significant. There was also no significant difference in mean AUDIT score between early and late survey responders.

Consumption characteristics
Ninety percent of respondents had consumed alcohol in the last 12 months, with a mean volume per typical occasion of 5.1 (SD = 5.0) standard drinks for women and 8.7 (SD = 8.6) for men. The National Health and Medical Research Council (Australia) thresholds for acute harm (40 g/60 g ethanol for women/men) [76] were exceeded at least once in the last 4 weeks by 48% of respondents.
Men had higher odds of drinking at hazardous levels compared to women (OR: 2.0; 95% CI: 1.8-2.2). Australian/NZ residents had higher odds compared to nonresidents (OR: 5.1; 95% CI: 4.3-6.0) and the association with age was non-significant (p = 0.113). Significantly higher mean AUDIT scores (p < 0.001) were observed for men and Australian/NZ residents compared to women and non-residents in all AUDIT subscales (shown in Table 2). There were significant differences in relation to age in the AUDIT Consumption subscale with higher mean scores for [17][18][19] year olds compared to 20-24 year olds, but not the other subscales.

Secondhand effects
The 4-week prevalence of secondhand effects is shown in Table 3. The most commonly reported effects were having to 'baby-sit' inebriated students (27.2%); having studying or sleep interrupted (20.9%); being insulted or humiliated (12.9%); having a serious argument (12.5%); or experiencing an unwanted sexual advance (10.9%).
There was a significant difference based on citizenship for most secondhand effects with Australian/NZ residents more likely than non-residents to have had a serious argument or quarrel (13.2% vs. 9.4%; p < 0.001); had to baby-sit another student (28.8% vs. 19.0%; p < 0.001) or to have experienced an unwanted sexual advance (11.9% vs. 5.8%; p < 0.001). Non-residents on the other hand were more likely to have had their studying or sleep interrupted (25.0% vs. 20.1%; p < 0.001); been a victim of sexual assault (2.1% vs. 0.8%); been a victim of another crime on campus (2.2% vs. 0.6%; p < 0.001) and were almost twice as likely to have found vomit in the halls or bathroom of their residence (10.0% vs. 5.6%; p < 0.001). The odds of experiencing most secondhand effects increases with increasing frequency of consuming six or more drinks (60 g ethanol) on one occasion, after adjusting for gender, age and citizenship. Being a victim of sexual assault, and being a victim of another crime on and off campus are not significantly associated with the frequency of this level of alcohol consumption (see Table 4).

Discussion
This study is the first known prevalence study of student drinking completed in Australia with undergraduate students. The vast majority of students were current drinkers (90%) and there was a high prevalence of hazardous drinking (48%), with a higher prevalence among men compared with women, and in Australian/ NZ residents compared with non-residents. A relatively Table 2 AUDIT subscale scores and hazardous drinking (AUDIT Score ≥ 8) by demographic characteristics large proportion of students' experienced secondhand effects from other people's drinking. The survey had a response rate of 56%, which is higher than large college surveys in the early 2000s in the United States (52%) [10], but lower than online surveys in New Zealand using similar procedures (63-82%) [13,59]. Higher response rates for online surveys have been linked to pre-notification, personalised contacts and follow up reminders [77]. Both this and the New Zealand studies incorporated pre-notification, personalised emails and follow-up notices. However, the earlier New Zealand study used up to nine follow-up contacts (compared to five in this study) including a telephone reminder, which may explain some of the difference. Follow-up notices are likely to increase response rates though larger numbers of notices may not appreciably affect response if the contact develops resistance to participation [78]. It is also possible that the novelty factor  *Results are presented as odds-ratio and associated 99% confidence intervals after adjusting for gender, age and citizenship in multivariable logistic regression of online surveys may have reduced in the years since the New Zealand studies and factors such as proliferation of junk mail, bombardment with online questionnaires and demands on student time may also have impacted on response rates [47]. The level of alcohol consumption reported in this study is less than that reported in New Zealand, for both men and women [13]. Although gender convergence in drinking has been reported elsewhere [10][11][12]79] and a similar trend appears to be occurring in Australia [80], this study shows a large discrepancy between men and women. However, there are no older prevalence studies from which to assess attenuation trends.
Large numbers of people were affected by other students' drinking. Of particular note was the 0.9% (n = 36) of women and 1.1% (n = 35) of men who reported being a victim of sexual assault in the previous 4 weeks. This is slightly higher than that found in a New Zealand sample [81] though with overlapping confidence intervals. While the New Zealand sample was limited to those who had consumed alcohol in the previous 4 weeks, our sample included non-heavy drinkers and may highlight the impact that hazardous alcohol consumption can have on all students. Extrapolated to the entire student population this may mean approximately 140 students at this university experience sexual assault in this context each month.
A limitation of this study was the imprecision in the specificity of crimes listed in the secondhand effects questions and the reliance on respondents to attribute responsibility for the effect. As only yes or no responses were available, multiple experiences of the same effect were not captured and therefore the prevalence of these effects may be underestimated. Given the high prevalence of some of these effects further research in this area is warranted. Our estimates may be biased by selective non-response but conversely computerised questionnaires are known to increase reporting of high-risk behaviour [42,51].
Universities with large on-campus resident populations may have higher levels of drinking than commuter universities due to students' greater proximity to peers [82]. As this study is based on a single commuter university and has a high proportion of students on temporary visas, the findings may be limited in their generalisability.
This study highlights the need for university programs to target drinking in this population. With half of male, and over a third of female, respondents drinking at hazardous levels, population approaches are needed. The literature suggests that programs should also address environmental factors, particularly the availability and promotion of alcohol on and around campus [22,83].

Conclusions
Hazardous alcohol use among undergraduate students remains an issue of concern although there is a lack of prevalence data on this population's alcohol consumption in Australia. Some alcohol related harms such as sexual assault are only detected with large population samples. Web-based surveys are a costeffective approach for measuring health behaviours in student populations, with a relatively high response rate. It is suggested that this research is replicated in other Australian universities, particularly residential campuses. Such surveys are required to develop trend data which will facilitate intervention program development.