Analysis of cut points for screening instruments for alcohol problems in the emergency room

The purpose of this study is to evaluate the performance of a number of alcohol screening instruments (CAGE, Brief MAST, AUDIT, TWEAK), in an emergency room population, against ICD-10 criteria for harmful drinking and for alcohol dependence from the CIDI, by gender, race and injury status.

A probability sample of patients were interviewed (N = 1,330) at the University of Mississippi Medical Center. Analyses were carried out on only current drinkers (n = 771; 58% of the sample). Instruments were compared on sensitivity, specificity and area under the curve using receiver operating characteristic (ROC) analysis to determine optimum cut points.

Overall the TWEAK and AUDIT performed best (in relation to sensitivity and area under the ROC curve) at standard cut points (3 and 8, respectively) for both harmful drinking and alcohol dependence. These instruments did not perform equally well across all subgroups, however, particularly among women, and among whites where the CAGE, with a cut point of 1, outperformed both the TWEAK and AUDIT at standard cut points. The TWEAK at a cut point of 2 also did as well as the CAGE at a cut point of 1 among women.

These data suggest that standard screening instruments do not perform equally well across ethnic and gender subgroups and that consideration of alternative cut points may be appropriate in some populations.