Are epidemiological and sociological surveys a proper instrument for detecting true problem drinkers? (The low sensitivity of an alcohol survey in Rotterdam).

The sensitivity of an alcohol survey in Rotterdam is estimated at only 16%. Taking account of sampling fluctuations, an upper confidence limit of 47% for the sensitivity can be computed. The statistical procedure is not based upon interviewing true problem drinkers to be identified from registers, but upon an anonymous verification of positively scoring individuals and a sample of negatively scoring individuals in the survey with registers of true problem drinkers. The above results may have implications for the interpretation of survey based prevalence figures.