Evaluating Shortened Versions of the AUDIT as Screeners for Alcohol Use Problems in a General Population Study

Abstract

Background: Efficient alcohol screening measures are important to prevent or treat alcohol use disorders (AUDs). Objectives: We studied different versions of the Alcohol Use Disorders Identification Test (AUDIT) comparing their performance to the full AUDIT and an AUD measure as screeners for alcohol use problems in Goa, India. Methods: Data from a general population study on 743 male drinkers aged 18–49 years are reported. Drinkers completed the AUDIT and an AUD measure. We created shorter versions of the AUDIT by (a) collapsing AUDIT item responses into three and two categories and (b) deleting two items with the lowest factor loadings. Each version was evaluated using factor, reliability and validity, and differential item functioning (DIF) analysis by age, education, standard of living index (SLI), and area of residence. Results: A single factor solution was found for each version with lower factor loadings for items on guilt and concern. There were no significant differences among the different AUDIT versions in predicting AUD. No significant DIF was found by education, SLI or area of residence. DIF was observed for the alcohol frequency item by age. Conclusions/Importance: The AUDIT may be used with dichotomized response options without loss of predictive validity. A shortened eight-item dichotomized scale can adequately screen for AUDs in Goa when brevity is of paramount importance, although with lower predictive validity. Although the frequency item was endorsed more by older men, there is no evidence that the AUDIT items perform differently in other groups of male drinkers in Goa.

Keywords:

• screener
• item analyses
• Alcohol Use Disorders Identification Test (AUDIT) versions
• alcohol use disorder (AUD)s
• India

THE AUTHORS

Madhabika B. Nayak received the Ph.D. degree in clinical psychology at Northern Illinois University, Dekalb, IL, USA. She is a licensed Psychologist and a Scientist at the Alcohol Research Group in Emeryville, CA, USA. Her areas of interest include alcohol assessment, interpersonal violence, global health, women's health, and brief interventions for problem drinking.

Jason C. Bond received the Ph.D. degree in statistics at the University of California, Los Angeles, CA, USA. He is a Biostatistician at the Alcohol Research Group in Emeryville, CA, USA.  His areas of interest include statistical methodology associated with the measurement of alcohol use, the design and analysis of clustered and longitudinal data, and latent variable methods. 

Thomas K. Greenfield received the Ph.D. degree in clinical psychology from the University of Michigan, Ann Arbor, MI, USA. He is the Scientific Director and Senior Scientist at the Alcohol Research Group in Emeryville, CA, USA and currently directs its National Alcohol Research Center and its National Alcohol Surveys.  He is also Core Faculty on the Clinical Services Research training program at the University of California San Francisco's Department of Psychiatry. His research interests include the epidemiology of alcohol use and problems including alcohol's harms to others, alcohol policy studies, consumer satisfaction, drinking patterns and mortality, psychometrics, and services research.

GLOSSARY

• Differential Item Functioning (DIF): Variability in the probability of endorsing an item by respondents from different groups with the same level of severity of the underlying trait. DI provides an indication of whether an item performs in different ways within a scale for different groups of respondents.

• Hazardous alcohol use: Drinking at levels likely to result in physical, social, and mental health problems.

• Item difficulty: A measure of the relative rate of endorsement of an item on the logit scale. For single group analyses, the latent problem severity at which the probability of a respondent endorsing an item is 0.5.

• Item discrimination: The degree to which the item aligns with an underlying latent trait (e.g., severity of alcohol problems).

• Item response theory (IRT) methods: Latent variable analysis methods for studying scaling properties of categorical items that form a scale.