Factorial Structure and Diagnostic Accuracy of a Digital Version of the Alcohol Use Disorders Identification Test (d-AUDIT) to Detect Hazardous and Problematic Drinking in Primary Care.

Abstract

We studied the factorial structure and diagnostic performance in primary care of a digital version of the Alcohol Use Disorders Identification Test (d-AUDIT) for screening for excessive drinking. In two primary care centers in Santiago, Chile, we conducted a cross-sectional study involving 330 people 18 years of age or older who had drunk alcohol six or more times in the last year. The d-AUDIT was developed from the paper version validated in Chile and was self-administered on 7-inch tablets. Trained psychologists evaluated the participants using a 1-year Timeline Followback and the alcohol use disorders section from the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition. We used confirmatory factorial analysis to examine the structure of the d-AUDIT and areas under the receiver operating characteristic curves (AUCs) to examine the diagnostic performance of the d-AUDIT. A two-factor model presented good overall fit, with item loads in the 0.53-0.88 range. The correlation among factors was .74, reflecting a good discriminant validity. The total score and the Fast Alcohol Screening Test (FAST) score (i.e., bingeing, role failure, blackouts, and others' concern items) obtained the best diagnostic performance for problematic drinking, with AUCs of 0.94 (CI [0.91, 0.97]) and 0.92 (CI [0.88, 0.96]), respectively. The FAST could differentiate hazardous drinking (cut point three in men and one in women) from problematic drinking (cut point four in men and two in women). We replicated prior factor analysis findings of a two-factor structure for the d-AUDIT with a good discriminant validity. The FAST obtained excellent diagnostic performance and retained some ability to discriminate between hazardous and problematic drinking.