Maximizing the use of the Alcohol Use Disorders Identification Test (AUDIT) as a two-step screening tool

Abstract

Few studies have examined Alcohol Use Disorders Identification Test (AUDIT) measurement invariance and no study has done so across clinically relevant drinking subgroups (e.g., unhealthy drinkers [UHDs] and those drinking below unhealthy drinking thresholds [non-UHDs]). Study 1 evaluated the factor structure and measurement invariance of the AUDIT across UHDs/non-UHDs (n = 1,350, 79.3 % female, 31.2 % UHDs). Study 2 validated a statistically derived 8-item AUDIT (AUDIT-8) in an independent sample of UHDs (n = 238, 49.2 % female). Confirmatory factor analysis examined factor structure and measurement invariance between UHDs/non-UHDs. Item response theory and differential item functioning evaluated sources of variance and removed problematic items, yielding the AUDIT-8. Predictive validity and test-retest reliability of the AUDIT-8 were examined. In study 2, convergent validity and factor structure of the AUDIT-8 were examined. A unidimensional AUDIT model fit best. Invariance models suggested differences across UHDs/non-UHDs. Items 1-3 (consumption) were most relevant for the non-UHDs. Items 4-8 (problems) were most relevant for UHDs; however, items 4-8 were not relevant and rarely endorsed among non-UHDs. Items 9-10 performed poorly in both groups. Test-retest reliability and predictive validity of AUDIT-8 was acceptable. In a second sample of UHDs, AUDIT-8 had a unidimensional structure and acceptable convergent validity with measures of consumption, affect and drinking motives. Results suggest that only items 1–3 of the AUDIT should be administered in screenings for UHD among university students, with items 4–8 administered among positive screens to confirm UHD. Items 9–10 should be removed. Implications for AUDIT-8 as a two-part screening tool are discussed.