Aims: To assess the utility of biological testing in a general population survey for estimating prevalence and evaluating self-report data quality.
Design: An audio computer-assisted interview was administered to subjects from June 2001 to January 2002. Immediately following the interview, subjects were requested to participate in hair, oral fluid and urine testing.
Setting: Subjects were from randomly selected households in the City of Chicago using multi-stage sampling methods. Interviews were conducted in subjects' homes.
Participants: The data represent 627 randomly selected adult participants, ages 18-40 years.
Measurements: Prevalance, kappa, conditioned kappa, sensitivity, specificity, under-reporting, 'mixed model' and logistic regression.
Findings: Higher rates of marijuana use were generated from survey reports than from drug testing. Drug testing generated higher prevalence rates than survey reports for recent use of cocaine and heroin. Under-reporting of recent drug use was apparent for all three substances. Sensitivity was particularly low for cocaine and heroin. Race was related to under-reporting, with African Americans less likely to report marijuana use despite a positive test result.
Conclusions: The utility of drug testing for surveys depends on the type of substance examined as well as on the type of test employed. Multiple tests have more utility than a single test. Drug testing is useful for identifying the levels and sources of under-reporting in a survey and provides a basis for adjusting prevalence estimates based on self-reports.