A primary concern in epidemiologic studies of drug abuse is the development of measurement strategies that are reliable and valid. Not only is it important to develop effective measures of drug abuse outcomes, but it is also necessary to obtain reliable, unbiased measures of related risk factors. Methodologic studies designed to address these concerns will increase confidence in estimates of prevalence and incidence of drug abuse as well as producing indicators of association that are unlikely to be biased. Little is currently known about the reliability of instruments that measure problems related to the use of illicit drugs. In particular, there is a relative dearth of information pertaining to test-retest reliability, although there are a number of studies that have examined inter-rater reliability. However, test-retest reliability is a prerequisite to inter-rater reliability. For this reason, the Etiology Branch is undertaking a test-retest study that is designed to assess the reliability of the substance abuse supplement to the 1991 National Health Interview Survey. This questionnaire consists of 122 items that assess the use of various drugs in the lifetime and last year, then specifically addresses problems related to the use of marijuana and cocaine. Two hundred individuals who enter the recruitment process at the ARC will complete the questionnaire at baseline and three weeks later. A random sample of one-half of these participants will be asked to complete an additional questionnaire two hours after responding to the first questionnaire. The test-retest reliability of individual items will be analysed, as well as the reliability of counts of symptoms or diagnostic categories based on combinations of symptoms. Another important concern in the conduct of field studies of drug abuse is the impact of the interview environment on self-reports of drug use. A study that was recently completed in the Branch using data from the 1990 National Household Survey on Drug Abuse (NHSDA) found that adolescents (12-17 years old) were less likely to report use of licit and illicit drugs if a parent was present during the interview but no difference if another person was present. This finding has implications for the conduct of future household surveys. Other methodologic issues addressed in this project include development and application of statistical methods relevant to drug abuse research. One recently completed project developed a statistical procedure for producing multi-dimensional confidence regions for depicting uncertainty around parameter estimates in conditional logistic regression.
