Objective: To improve the CAGE (Cut down, Annoyed, Guilty, Eye opener) questionnaire's predictive accuracy in screening college students. Participants: The sample consisted of 219 midwestern university students who self-administered a confidential survey. Methods: Exploratory factor analysis, confirmatory factor analysis, receiver operating characteristics (ROC), and Cronbach's alpha were used to analyze factor structure, validity, and reliability. Results: The modified CAGE correctly classified students with alcohol abuse ("AA students"; area under the curve [AUC] = 0.7765) and students with alcohol dependency ("AD students"; AUC = 0.8392) more often than CAGE (AA students: AUC = 0.6977; AD students: AUC = 0.7437), and these differences are statistically significant (AA students: ?[superscript 2](1) = 14.72, p < 0.001; AD students: ?[superscript 2](1) = 7.71, p < 0.01). Using 2-point cut scores, CAGE correctly identified 59.38% of AD students as AD, whereas the modified CAGE correctly identified 87% of AD students as AD. Using 1-point cut scores, CAGE correctly identified 65% AA students, whereas the modified CAGE identified 85.29%. Conclusions: The modified CAGE has better accuracy than CAGE in predicting AA and AD among college populations.