Reliability of self-reported alcohol consumption in a general population survey.

Reliability estimates for the measures of alcohol consumption used in a household survey were examined. Respondents who had not consumed any alcohol within the previous year and those under age 18 were excluded from the survey, yielding a sample of 1395 (48% men). With one survey that involved a 2-week recall period and another that involved a 4-week period, three methods for estimating reliability were employed: alternate forms, test-retest and a combined method. Validity was also examined using a 30-day drinking diary as a criterion. The findings indicate high levels of reliability, averaging .91 for the consumption measures of beer, wine and distilled spirits. Validity estimates were also fairly substantial but not uniformly so across the different beverage types. In general, the results suggest that these consumption measures can be used with considerable confidence.