Exploration of the (Interrater) Reliability and Latent Factor Structure of the Alcohol Use Disorders Identification Test (AUDIT) and the Drug Use Disorders Identification Test (DUDIT) in a Sample of Dutch Probationers

Abstract

Background: The use of brief, reliable, valid, and practical measures of substance use is critical for conducting individual (risk and need) assessments in probation practice. In this exploratory study, the basic psychometric properties of the Alcohol Use Disorders Identification Test (AUDIT) and the Drug Use Disorders Identification Test (DUDIT) are evaluated. Methods: The instruments were administered as an oral interview instead of a self-report questionnaire. The sample comprised 383 offenders (339 men, 44 women). A subset of 56 offenders (49 men, 7 women) participated in the interrater reliability study. Data collection took place between September 2011 and November 2012. Results: Overall, both instruments have acceptable levels of interrater reliability for total scores and acceptable to good interrater reliabilities for most of the individual items. Confirmatory factor analyses (CFA) indicated that the a priori one-, two- and three-factor solutions for the AUDIT did not fit the observed data very well. Principal axis factoring (PAF) supported a two-factor solution for the AUDIT that included a level of alcohol consumption/consequences factor (Factor 1) and a dependence factor (Factor 2), with both factors explaining substantial variance in AUDIT scores. For the DUDIT, CFA and PAF suggest that a one-factor solution is the preferred model (accounting for 62.61% of total variance). Conclusions: The Dutch language versions of the AUDIT and the DUDIT are reliable screening instruments for use with probationers and both instruments can be reliably administered by probation officers in probation practice. However, future research on concurrent and predictive validity is warranted.

Keywords:

• AUDIT
• DUDIT
• alcohol abuse
• drug abuse
• substance abuse
• reliability and factor structure
• probation
• screening

THE AUTHORS

Martin Hildebrand, Ph.D., LL.M works as a consultant forensic psychologist and researcher in private practice. He received his Ph.D. in clinical psychology from the University of Amsterdam, Amsterdam, the Netherlands, in 2004. Dr. Hildebrand's areas of professional interest include criminal responsibility, personality assessment, psychopathy, violence risk assessment, treatment evaluation, test development, and psychological assessment. He published rather extensively in the field of forensic psychology (journal articles, book chapters, and research reports). Furthermore, he has conducted numerous workshops in the Netherlands on risk assessment of child abuse, spousal assault, and general (sexual) violent behavior, as well as (PCL-R) psychopathy. Recently, by order of (and in collaboration with) the Dutch Probation Service, he developed the new version of the Risk Assessment Scales, the Dutch national diagnostic and risk assessment tool developed to help probation officers to accurately and consistently assess the risk of re-offending and dangerousness for each offender. For more information, see www.martinhildebrand.nl

Mirthe G. C. Noteborn, M.Sc., is a lecturer at the Department of Developmental Psychology of Tilburg University, Tilburg, the Netherlands, and researcher at forensic psychiatric hospital De Kijvelanden, Poortugaal, the Netherlands. Mirthe received a Bachelor's degree in Psychology from Maastricht University, Maastricht, the Netherlands, in 2009. In 2010, she received a Master's degree in Psychology and Law (cum laude). In 2012, she received her second Master's degree in Forensic Psychology (cum laude), also from Maastricht University. Mirthe's current research project involves the measurement and treatment of (distorted) cognitive schemas in sexual delinquents. Other professional and research interests include recidivism risk, desistance from crime, psychopathy, and personality disorders.

GLOSSARY

• Risk-need-responsivity (RNR) model: The RNR model has been widely regarded as the premier model for guiding offender assessment and treatment. The RNR model underlies some of the most widely used risk-needs offender assessment instruments, and, for a long time, it was the only theoretical model that has been used to interpret the offender treatment literature.

Notes

¹ Inconsistent answering was coded in the case of contradicting responses to the items, for example, if item 1 of the AUDIT (total number of drinking occasions) or item 1 of the DUDIT (how often drugs other than alcohol) = 0 and the score on item 3 of the AUDIT (frequency of binging occasions) or item 3 of the DUDIT (how many times taking drugs on a typical day when drugs are used) surpassed the total number of drinking or drug use occasions (item 1 = 0). It should be noted that in the present study, offenders who stated that they did not use alcohol (item 1 on AUDIT = 0) and left all other items unanswered, were assigned a total AUDIT score of 0 and were included in the analyses of the material (n = 88). Offenders who stated that they did not use drugs (item 1 on DUDIT = 0) and left all other items unanswered were assigned a total DUDIT score of 0 (n = 163).

² In most samples, especially in “low prevalence samples,” AUDIT and DUDIT scores are not normally distributed but positively skewed, that is, most respondents obtain relatively low scores and a few obtain high scores. Despite that, we have chosen to use parametric statistics and analyses procedures on raw scores to make our results comparable to previous AUDIT and DUDIT studies (e.g., Almarri et al., 2009; Berman et al., 2005; Mathew et al. 2010; Voluse et al., 2012).