Quality of treatment data. Reliability over time of self-reports given by clients in treatment for substance abuse.

This study examines the reliability over a 2-month period of self-reports of drug use, sexual behaviors, and use of treatment services provided by 2,968 clients participating in a large, multisite, prospective study of drug treatment in the United States-the Drug Abuse Treatment Outcome Study (DATOS). Analyses focus on responses to 62 pairs of logically related questions that were asked at two points in time: (1) 1 month after entry into treatment, and (2) 3 months after entry into treatment. Subjects' responses to questions asked at these two time points are assessed for logical consistency. Prior analyses of self-reports provided by DATOS clients at one point in time (entry into treatment) found surprisingly high levels of within-interview consistency in their reporting of alcohol use (Turner & Hubbard, 1995) and cocaine use (Adair, Craddock, Miller, & Turner, 1995). The crosstemporal tests of consistency reported in this article eliminate several potential sources of artifactual consistency that may have affected prior analyses, (e.g., consistency imposed by an interviewer or constructed by a respondent during the course of a single interview). Contrary to expectations, crosstemporal comparisons reveal high levels of logical consistency in clients' responses. The mean percent of substantively inconsistent responses ranges from 0.7% for questions asking about frequency of drug use to 4.4% for questions asking about sexual behaviors.