Screening for alcohol problems in the U.S. general population: a comparison of the CAGE and TWEAK by gender, ethnicity, and services utilization

Abstract

Objective: The purpose of this study was to compare the performance of two screening instruments for alcohol problems, the CAGE and the TWEAK, against ICD-10 and DSM-IV criteria for alcohol dependence by gender and ethnicity, and to evaluate whether characteristics associated with health services utilization may affect the performance of screening instruments, in a representative sample of the U.S. adult general population.

Method: Data are from the Alcohol Research Group's 1995 National Alcohol Survey; these were weighted to account for the design effect inherent in multistage cluster sampling and oversampling of blacks and Hispanics. Effective sample size was 2,443: 797 blacks, 642 Hispanics and 1,004 whites and others (primarily Asian and Native American). Sensitivity and specificity were examined for both screening instruments. Logistic regression was used to evaluate the predictive value, separately, for the CAGE and TWEAK, controlling for gender, ethnicity, region of the country and service use (primary care and emergency room).

Results: The TWEAK was more sensitive for men than for women, but no differences were found by service use, while the CAGE was more sensitive for men who had used the emergency room during the preceding year compared with those who had not. The TWEAK was more sensitive than the CAGE among white and Hispanic men, among men who had not used the emergency room and among whites who had no service use. Performance of neither the CAGE nor the TWEAK was found to vary by region of the country.

Conclusions: Data suggest that while the performance of screening instruments may vary across demographic subgroups in the general population, instruments may perform equally well for identifying problem drinkers in general populations as in clinical populations.