The program's main objective was to reduce the spread of HIV by counseling drug users and by improving drug-treatment programs in the area. A second objective was to examine the correlates of risk behavior. The evaluation used random regression models to obtain both individual level and group level estimates of behavioral change over time. The power of these models lies in their ability to provide information on how individual change over time is unique (a random effect) while simultaneously accounting for how individuals change uniformly over the same time period (a fixed effect). A structured interview was administered six times over an 18-month period. Of those persons assessed at baseline (n=475), 95 percent (n=451) were also reinterviewed in the last interview at 18 months. Both group and individual level changes in risk behavior were assessed using the random regression models. The study focused on three potential risk behaviors for HIV/AIDS: number of sexual partners, frequency of condom use, and injection drug use. For each risk behavior, a separate statistical model was estimated. The results of the random regression models show significant reductions in number of sexual partners and injection drug use. Also, a number of variables, such as perceived risk for AIDS and knowledge of HIV/AIDS, were statistically significant covariates of risk behavior. 1 figure, 6 tables, 2 notes, and 52 references