Screening properties of questionnaires and laboratory tests for the detection of alcohol abuse or dependence in a...

Early identification of alcohol abuse or dependence is important in general practice because many diseases are influenced by alcohol. General practitioners, however, fail to recognise most patients with alcohol problems.

To assess the diagnostic performance of the CAGE and AUDIT questionnaires, their derivatives, and laboratory tests in screening for alcohol abuse or dependence in a primary care population (male and female patients), attending their general practitioner (GP).

A diagnostic cross-sectional study.

A random sample of patients who were over 18 years of age (n = 1992) attending 69 general practices situated in the same region in Belgium.

Alcohol questionnaires (CIDI 1.1, section I, CAGE, AUDIT, AUDIT-C, Five-Shot, and AUDIT Piccinelli) were completed, demographic information was recorded, and patients underwent conventional blood tests, including mean corpuscular volume, liver function tests, the gamma-glutamyl transferase test, and carbohydrate-deficient transferrin (CDT, estimated using %CDT). Calculations of sensitivity, specificity, positive predictive value, negative predictive value, odds ratios with their 95% CIs, and receiver operating characteristic (ROC) curves for different scores of the questionnaires and laboratory tests, using DSM-III-R as the reference standard.

The past-year prevalence of alcohol abuse or dependence in this population was 8.9% (178/1992) of which there were 132 male and 45 female patients attending a general practice. The GPs identified 33.5% of patients with alcohol abuse or dependence. Among male patients, all questionnaires had reasonable sensitivities between 68% and 93% and hence at lower cut-points than recommended. Only the sensitivity of the CAGE, even at its lowest cut-point of > or = 1 was lower (62%). In female patients the sensitivities were lower; however, odds ratios were higher for different questionnaires. The receiver operating characteristic (ROC) curves did not differ between the questionnaires. The laboratory tests had low diagnostic accuracy with areas under the ROC curves (AUCs) between 0.60 and 0.67 for female patients and 0.57 and 0.65 for male patients.

This is one of the largest known studies on alcohol abuse or dependence among family care practices. We confirm earlier results that the AUDIT questionnaire seems equally appropriate for males and females; however, screening properties among male patients are higher. Nevertheless, the Five-Shot questionnaire is shorter and easier to use in a general practice setting and has nearly the same diagnostic properties in male and female general practice patient populations. We confirm that conventional laboratory tests are of no use for detecting alcohol abuse or dependence in a primary care setting. Also, the %CDT cannot been used as a screening instrument in this general practice population.