Ethnic and sex bias in primary care screening tests for alcohol use disorders

Background: The use of self-report screening tests for alcohol use disorders in the primary care setting has been advocated. Objective: To test for ethnic and sex bias in three self-report screening tests for alcohol use disorders in a primary care population. Design: Cross-sectional study with patients randomly selected from appointment lists. Setting: University-based family practice clinic. Patients: Probability sample of 1333 adult family practice patients stratified by sex and ethnicity. Measurements: Patients completed 1) a diagnostic interview to determine the presence of a current alcohol use disorder and 2) three screening tests: the CAGE questionnaire, the Self-Administered Alcoholism Screening Test (SAAST), and the Alcohol Use Disorders Identification Test (AUDIT). Results: The areas under the receiver-operating characteristic (ROC) curves for the CAGE questionnaire and the SAAST ranged from 0.61 to 0.88 and were particularly poor for African-American men and Mexican-American women. For the AUDIT, the area under the ROC curves was greater than 0.90 for each patient subgroup. The sensitivity of the CAGE questionnaire and the SAAST at standard cut-points was lowest for Mexican-American women (0.21 and 0.13, respectively). Positive likelihood ratios for the AUDIT were similar to or higher than those for' the other screening tests, whereas negative likelihood ratios were lowest for the AUDIT (<0.33), indicating the superiority of this test in ruling out a disorder. Conclusions: A marked inconsistency in the accuracy of common self-report screening tests for alcohol use disorders was found when these tests were used in a single clinical site with male and female family practice patients of different ethnic backgrounds. The AUDIT does not seem to be affected by ethnic and sex bias.