Performance of alcoholism screening questionnaires in elderly veterans

Abstract

To validate three alcoholism screening questionnaires in elderly male veterans. Participants were 120 male veterans aged 65 years or older. In this cross-sectional study, consecutive patients in the outpatient general medical practice at the Omaha VA Medical Center were interviewed with the alcohol module of the Revised Diagnostic Interview Schedule (DIS-III-R) and three alcoholism screening questionnaires, the Michigan Alcoholism Screening Test-Geriatric Version (MAST-G), the CAGE, and the Alcohol Use Disorders Identification Test (AUDIT). Performance characteristics (sensitivity, specificity, predictive values, likelihood ratios, and areas under the receiver operating curve [ROC]) of the screening questionnaires were determined in comparison with the DIS-III-R. Forty-three participants (36%) met DSM-III-R (Diagnostic and Statistical Manual of Mental Disorders) criteria for alcohol abuse or dependence; 23% were active drinkers and 13% were inactive. Fifty of the 120 (42%) reported abstinence from drinking during the preceding year. A MAST-G score > or = 5 had a sensitivity and specificity of 70% and 81%, respectively. A CAGE score > or = 2 had a sensitivity and specificity of 63% and 82%. Using active drinkers only, an AUDIT score of > or = 8 had a sensitivity and specificity of 33% and 91%. The positive predictive values (PPV) for the MAST-G, CAGE, and AUDIT were 67%, 66%, and 69%, respectively; the negative predictive values were 83%, 80%, and 68%, respectively. Areas under the receiver operating curves for the MAST-G, CAGE, and AUDIT were 0.84 +/- 0.04, 0.77 +/- 0.05, and 0.56 +/- 0.08, respectively. The MAST-G and the CAGE outperformed the AUDIT for detecting alcohol abuse and dependence in an elderly male veteran population. The CAGE, requiring only four easily memorized questions to achieve similar accuracy, appears to offer an advantage over the 24-item MAST-G.