Reliability of self-reported HIV risk behaviors of drug users.

In AIDS research, relatively little attention has been paid to reliability of self-report of drug users. The authors examined the test-retest reliability of the Risk Behavior Assessment (NIDA, 1991) questionnaire. This structured-interview questionnaire was administered twice to 196 drug users in 5 cities over a 48-hr period. Findings indicated that respondents consistently self-report drug use, injection practices, and sexual behaviors; discrepancies do not appear to reflect systematic decreases or increases in self-report; unreliability is associated with poorly worded questions and respondent characteristics; and discrepant reports warrant attention in analysis and interpretation of data. Measurement error has implications for estimating risks, understanding relationships between behavior and HIV transmission, and interpreting change after interventions. Items with low reliability have been revised, and further reliability studies are examining whether revisions have led to improved reliability.