Effective measurement of problematic drinking for college students: Reducing differential item functioning across gender and race

Very few students (aged 18 -- 24) recognize alcohol-related problems, seek treatment, or receive efficacious interventions for problematic alcohol use. There is concern that current problematic alcohol use screening measures for college students may miss opportunities for detection and referral, and more evaluation of these measures is warranted. This dissertation evaluated the metric invariance (across race and gender), reliability, criterion validity, and construct validity of four problematic alcohol use screening tests -- the Alcohol Use Disorders Identification Test (AUDIT), the Rutgers Alcohol Problem Index, the Young Adult Alcohol Problems Screening Test, and the College Alcohol Problems Scale, Revised -- all designed to screen people for problematic alcohol use. All four measures demonstrated partial metric invariance and several items were recommended for removal from each measure. All measures demonstrated good to excellent reliability. Furthermore, a majority of the measures demonstrated expected associations with measures of alcohol consumption, alcohol dependence, and desirable responding; however, unexpected construct validity results were discussed. Receiver Operating Characteristics curves demonstrated acceptable to good criterion validity for all measures, using an alcohol dependence diagnosis as the criterion. However, there was substantial variation across ideal cut score values by race and gender for most of the measures. Ultimately, recommendations were made for using the AUDIT as the most appropriate tool to screen and refer college students. Using the total pool of invariant items across all measures, a fifth measure was constructed, but it did not substantially improve the detection of students with problematic drinking styles.