Reliability of interview responses of injecting drug users

We compared 23 injecting drug users (IDUs) who were part of a larger study at retest periods of less than greater than one month for test-retest reliability of their responses to a detailed questionnaire covering demographics, drug use, HIV risk behavior and attidues and knowledge about HIV. Data indicated that there was high test-related reliability for demographic characteristics, and for the social context of drug use for the shorter retest duration. Reliabilities for sexual histories, drug use, and equipment use declined over time. Reliabilities for items where respondents were asked to estimate probabilities were low, as were data on attitudes. We conclude that reliability for clear matters of fact such as demographic characteristics is high, for matters of sexual or drug-related behavior over a month period are reliable over the short but not the longer term (since the behvavior reported in the retest may have changed), and that reliability for estimates of the percentage of time particular behaviors occured was low. Reliability of self-report data of IDUs appears to depend on the context and time-frame of the question, rather than on the nature of the population.