Developing a Screening Questionnaire for Problem Drinking in College Students

Abstract

The CAGE questionnaire is one of several brief screening instruments that has been successful in detecting alcoholism in a general population. There is evidence, however, indicating that none of these instruments has sufficient validity for detecting the less severe pattern of problem drinking that is more typical of college students. This study analyzed survey data, including CAGE questions, to identify items that would have sufficient test operating characteristics more valid for screening problem drinkers. Problem and normal drinking patterns were defined by combining certain categories of quantity-frequency data with categories of alcohol-related problems. Problem and normal drinkers were identified as those students at the ends of the quantity-frequency/alcohol-problems continuum. Differences between the two groups in response frequencies to a 17-item alcohol-use survey were examined using chi-square analyses. Five items showing highly significant differences were identified, and sensitivity, specificity, and positive predictive values were calculated at possible cutoff points. These items were (1) a positive response to any 2 of the 4 CAGE items, (2) reporting rarely or never choosing nonalcoholic beverages at social events, (3) driving under the influence at least 6 to 10 times or more in the last year, and (4) having started regular use of alcohol before the college years. At the recommended cutoff score for a positive test, the items have a sensitivity of 88%, specificity of 87%, and positive predictive value of 52%. When used for screening in a population similar to this, it will fail to identify 12% of the problem drinkers and will falsely classify 13% of all normal drinkers as problem drinkers. These data represent a considerable improvement in basic screening capability for problem drinking over that demonstrated with the CAGE.