Performance of screening instruments for identifying alcohol dependence in the general population, compared with...

Screening instruments for problem drinking have been developed in clinical populations, and little is known of their performance in the general population. Sensitivity and specificity of the CAGE and the TWEAK for ICD-10 and/or DSM-IV criteria for alcohol dependence are compared among those sampled from the southern region in the 1995 National Alcohol Survey (n = 776), and from emergency room (ER) (n = 1327) and primary care clinic (n = 767) samples in Jackson, MS. Sensitivity of the CAGE was higher in both the ER (85%) and primary care (82%) samples than in the general population (75%), although differences were not significant, and this held for males and females alike. Sensitivity of the TWEAK was lower in the general population sample (83%) than in the ER sample (89%), but higher than in the primary care sample (75%). In the general population, sensitivity of both screeners was better (although not significantly so) among those making an ER or primary care visit during the last year than among those not doing so, with the largest difference found for the CAGE (85% vs. 65%, respectively). In a merged sample of the three sites, the interaction of site by screening instrument was not a significant predictor of alcohol dependence for either the CAGE or TWEAK. Data suggest that screening instruments may not perform as well in the general population as in some clinical populations, with more variation apparent across groups for some screeners than for others. Although further analysis is necessary to explore this issue further, attention should be given to selection of the best instrument for use in a given population.