Test-retest reliability of the alcohol use disorder identification test in a general population sample.

Abstract

A number of different screening tests are frequently used in alcohol research, but our knowledge about the reliability of many of them is quite limited. Recently, this problem has received more attention. This article examines the test-retest reliability of one of these instruments-the Alcohol Use Disorder Identification Test (AUDIT)-in a general population sample. A general population sample (n = 457) was tested and, after approximately 1 month, was retested by using the AUDIT. Correlation between the two tests has been examined with the intraclass correlation coefficient and the kappa coefficient in analysis of dichotomous variables. Specificity and sensitivity at a number of different cutoff scores have also been analyzed by using the first test as a criterion. On the item level, the correlations ranged between 0.6 and 0.8. The overall reliability of total AUDIT scores was 0.84. When stratified by gender, age, and consumer status, the total score reliability approximated 0.80 for all the categories except low alcohol consumers (0.51). Agreement using the recommended cutoff score of 8+ was also examined. The reliability (kappa) observed in the whole sample was 0.691, which was interpreted as a substantial agreement. By this cutoff, 91% were correctly classified at retest compared with the first test. AUDIT 8+ showed higher reliability for males, young people, and moderate consumers and low reliability among low consumers. In terms of reliability, the most optimal cutoff for women turned out to be 6 or more. According to these results, the test-retest reliability of AUDIT is high. The next step might be to examine to what extent the findings apply within health-care settings, which is what the test originally was designed for.