Background
Risky alcohol consumption is common among Australian young people and associated with elevated incidence of accidents, physical injury and other short-term harm, as well as contributing to long-term health conditions and alcohol dependence [
1‐
4]. In epidemiological research, monitoring alcohol consumption and related harm is important to identify high-risk groups, risk behaviours, and trends in alcohol use in order to inform, target and evaluate harm reduction strategies.
The Alcohol Use Disorders Identification Test (AUDIT) was developed by the World Health Organisation (WHO) as a simple clinical screening tool to detect hazardous and harmful alcohol use and facilitate early intervention in primary care [
5]. The AUDIT comprises ten questions covering the domains of alcohol consumption (questions 1–3), alcohol dependence (Q4-6), and harms (Q7-10) and is scored out of 40 [
6]. Although the AUDIT was developed as a tool for use in clinical practice, it has been used more broadly for web-based screening and epidemiological research [
7‐
9]. The AUDIT has also been adapted into shortened versions containing subsets of AUDIT items for use in settings subject to time constraints [
9]. The most common shortened versions are the AUDIT-C [
10], AUDIT-3[
11], AUDIT-4 [
12], and the Fast Alcohol Screening Test (FAST) [
13]). Previous research suggests that shortened versions may be sufficient proxies for the full AUDIT (AUDIT-10), but evaluations of both the AUDIT-10 and shortened AUDIT versions are generally limited to clinical [
10‐
12,
14‐
18] or adult general population samples [
7,
8,
19‐
21], with the exception of a number of studies of the AUDIT-10 in college students [
22‐
25]. Although some of these studies have reported results among young adults [
19,
26], there is a need for further research to determine the performance of shortened AUDIT versions in high-risk populations such as young people aged less than 18 years.
We have been conducting alcohol, other drug and sexual risk behaviour surveillance of young people attending a music festival in Melbourne, Australia since 2005 [
27,
28]. While the AUDIT-10 is a candidate instrument to classify alcohol consumption and risk, space constraints on our questionnaire make it difficult to conduct the full AUDIT-10; a shortened version that captures important information across the consumption, dependence and harms domains suited to our recruitment setting is required; to our knowledge, no such instrument exists. In this study we assess how established and novel shortened versions of the AUDIT perform in relation to the full AUDIT-10 in a community-based sample of young people attending a music festival. Our results will inform our subsequent community-based surveys and be useful for others conducting research with young non-clinical populations.
Discussion
This study highlights how shortened versions of the AUDIT, using just a few items, can capture much of the information available from the full AUDIT scale, thus suggesting they may be as effective a screening tool to measure hazardous or harmful alcohol consumption. A number of shortened versions of the AUDIT performed well in relation to the AUDIT-10 by means of internal consistency, variance explained, and concurrent validity. Although some three-item combinations scored highest for individual measures, four-item combinations scored more consistently high across all statistical tests. Novel AUDIT versions may be more effective than many established shortened versions as an alternative screening tool to the AUDIT-10 to measure hazardous or harmful alcohol consumption in community-based populations of young Australians.
While AUDIT-C was the only AUDIT version under our study criteria which met the Cronbach’s alpha cut-off of at least 0.7 for research purposes [
5,
35], it did not perform well by other measures. The AUDIT-C exists in a single AUDIT domain based on consumption patterns, and thus items are more likely to correlate with each other, leading to a higher Cronbach’s alpha [
36]. In contrast, the alternative novel versions crossed three domains that measured different types of alcohol misuse, as prioritised in the study protocol, so lower internal consistency may be reasonable.
Using just three of ten AUDIT questions was sufficient to explain over 80% of total variance in the AUDIT-10 score, with any additional items increasing this figure slightly. In general, correlation with AUDIT-10 was high, which may be due to the shortened AUDIT versions being subsets of the AUDIT-10 and not independent measures, as well as our study criteria for only considering novel combinations which represented all three AUDIT domains. In contrast, the AUDIT-C, which only covers one domain, explained considerably less variance and had lower correlation to the AUDIT-10 than novel combinations, which was also observed in a previous computerised survey of young people [
37].
Our results suggest that numerous variations of the AUDIT scale may be a suitable alternative to the AUDIT-10 for classifying harmful and hazardous drinking in this sample of young people. When considering performance of AUDIT variations across all statistical measures, the FAST scored highest overall. However, the FAST does not include item 9 on alcohol-related injury, which was prioritised due to its significance to young people [
33]. Nonetheless, there are advantages to using the FAST, given it has been validated in other settings, albeit primarily emergency departments [
13,
17,
21]. After FAST, the two top-performing variations were similar: 3, 4, 8 and 9 or 3, 4, 5, and 9. There is no sound basis for choosing one of these combinations over the other in this sample population, although item 8 on memory loss has been shown to be a useful measure of harm and predictor of future alcohol-related injury in young people [
37,
38].
A number of limitations to this study and interpretation should be considered. The survey did not include a gold-standard or comparative clinical diagnosis of harmful/hazardous alcohol use or alcohol-related problems, such as DSM-IV, meaning we were unable to assess predictive validity. Subsequently, novel AUDIT variations were compared to the AUDIT-10, which itself has not been validated in this particular population, and some prior research has reported that the AUDIT is less reliable in the general population compared to clinical settings [
8]. As such, we are only able to determine the performance of the shortened AUDIT variations in comparison to the full AUDIT, rather than as a predictor of harmful drinking
per se; agreement does not infer accuracy of either the shortened versions or AUDIT-10 as a predictor of hazardous drinking or high-level drinking problems. Furthermore, the shortened AUDIT models are not independent of the AUDIT-10, and thus undermine the assumption of independence for assessing linear regression (variance explained). While this method has been used in previous related studies, it should still be interpreted with caution (e.g. [
37]). A modified version of the FAST that was not gender-specific was used and may have overstated the performance of FAST in comparison to AUDIT-10.
This study population is based on a convenience sample and is not intended to be representative of all young Australians. In addition, due to the nature of the recruitment setting, we were unable to assess a response rate for participation. Prior surveys have demonstrated that participants recruited at the Melbourne Big Day Out music festival are more likely to engage in alcohol-, drug-, and sex-related risk behaviours than other young Australians [
39,
40]. While the high-risk nature of this sub-population makes them an important group for testing the performance of AUDIT variations, results may differ in other populations. Further research is needed to confirm findings in broader and more representative samples.
Conclusion
Among a sample of young people commonly reporting risky alcohol consumption, we identified a number of novel three and four-item AUDIT variations, as well as the established FAST scale, which were suitable proxies of the AUDIT-10. Although it is difficult to ascertain a single standout shortened AUDIT variation in this population, numerous variations performed better than the more widely used AUDIT-C according to multiple criteria. Four-item combinations scored more consistently high across evaluated statistical measures and are the preferred compromise for maximising the indication of alcohol misuse while ensuring a short and simple measure of hazardous drinking in a community-based sample of young people.
Acknowledgements
MH and PD received people support funding from the NH&MRC, and PD is the recipient of an ARC Future Fellowship. The 2010 survey was funded internally by the Burnet Institute. The authors gratefully acknowledge the contribution to this work of the Victorian Operational Infrastructure Support Program. The authors thank Bianca Fiebeger for allowing us to recruit at the Big Day Out music festival; the trained recruiters who explained the survey to study participants and collected written consent and data; and the study participants. Showbag contents were donated by the Victorian Department of Justice, Marie Stopes International, Youth Projects, Headspace, Department of Health and Ageing, and Cadbury Schweppes.
Competing interests
The authors report no competing interests relating to this manuscript.
Authors’ contributions
The following co-authors have contributed to the work: AB in data collection, data analysis, manuscript preparation and manuscript review; MG in data analysis and manuscript review; MH in study design, manuscript preparation and manuscript review; and PD in study design, data analysis, manuscript preparation and manuscript review. All authors read and approved the final manuscript.