Alcohol screening in general practices using the AUDIT: How many response categories are necessary?

Aims: The Alcohol Use Disorders Identification Test (AUDIT) is a common screening instrument. This study analyses if response categories of the AUDIT might be dichotomized without affecting the psychometric properties of the questionnaire. Methods: Participants between 18 and 65 years were recruited from general practices in two northern German cities. In total, 10,803 screenings were conducted (refusal rate: 5.9%). For those who were screened positive, the Munich-Composite International Diagnostic Interview (M-CIDI) was used for identification of 12-month Alcohol Use Disorders and at-risk consumption (exceeding 20/30 g per day). Abstinent subjects and screening positives without diagnostic interview were excluded from the analysis, leaving a sample of 7,112 subjects. ROC-Curves were calculated separately for each item in order to identify an optimal cut-off value. Finally, a version of the AUDIT based on dichotomized items was compared to the original version and its short-form, the AUDIT version based on three questions dealing with consumption AUDIT-C. Results: As an optimal cut-off value for items on consumption, drinking at least once a week, having more than 1-2 drinks per occasion, and drinking 6 or more drinks in one sitting at least once a month were identified. For all questions on alcohol-related problems or dependence symptoms, having 'ever occurred' differed best between subjects with and without Alcohol Use Disorders or at-risk consumption. Sensitivity and specificity of the dichotomized version of the AUDIT did not differ from the original version, and both full versions performed superior compared to the AUDIT-C. Conclusion: Data indicate that the AUDIT response categories may be dichotomized without affecting its validity.