Abstract
In Switzerland in 1986, an anonymous classroom questionnaire-survey on the consumption of solvents and illegal drugs carried out on a representative sample of 15 to 16-year-old schoolchildren showed a lifetime prevalence of about 10% for cannabis consumption, 12% for the use of solvents, and almost no use of other illegal drugs. These results, however, are contaminated by a relatively high percentage of non-responses to questions on use of illegal drugs (10%).
This paper deals first with an approach to solving the general problem of missing information in responses to interviews or questionnaires on the use of illegal drugs, and the specific problem of non-response to particular questions. It then describes a probabilistic estimation of effective prevalence rate on the basis of a number of known and credible predictors of illegal-drug use.
The methods used consist of a review of the scientific literature on methodological problems of asking questions about drug use. Discriminant and logistic regression analyses are applied to assess possible reasons for non-responses as well as to calculate the probabilistic estimation of effective prevalence rate. The discussion of the results considers ways of reducing the numbers of non-responders in similar future surveys.