Testing the validity of the AUDIT-C and AUDIT-3 to detect unhealthy alcohol use among high-risk populations in Zambia: A secondary analysis from two randomized trials.

Abstract Background This study evaluated the test characteristics of brief versions of the Alcohol Use Disorders Identification Test (AUDIT), the AUDIT-C and AUDIT-3, compared to the full AUDIT in populations with heavy drinking living in Zambia and compared differences in effect size estimates when using brief versions in clinical trials. Methods Data were obtained from two randomized trials of the Common Elements Treatment Approach (CETA) for reducing unhealthy alcohol use among adult couples and people living with HIV (PLWH) in Zambia. The full AUDIT was administered to participants at baseline and at 6- or 12-month follow-up. Sensitivity and specificity of the brief versions were calculated in comparison to the full AUDIT. Mixed effects regression models were estimated to calculate the effect sizes from the trials using the brief versions and these were compared to the originally calculated effect sizes using the full version. Results The AUDIT-C performed well at cut-off ≥ 3 for both men (sensitivity: >80%; specificity: >76%) and women (sensitivity: >84%; specificity: >88%). The AUDIT-3 performed best at cut-off ≥ 1, but with comparatively reduced validity for men (sensitivity: >77%; specificity: ≥60%) and women (sensitivity: ≥72%; specificity: >62%). Effect sizes were different by up to 52% using the AUDIT-C and up to 60% for the AUDIT-3 compared to the AUDIT. Conclusions The AUDIT-C is recommended as a brief screening tool for community-based and clinic-based screening in Zambia among populations with high prevalence of unhealthy alcohol use. For research studies, the full AUDIT is recommended to calculate treatment effect.