Aside from hemophiliacs, the highest seroprevalence rates have been documented among drug users in the New York metropolitan area: as early as 1984, rates of infection as high as 50 percent were reported in some studies (Marmor et al., 1987). However, alarming HIV seroprevalence rates also have recently been observed among other groups, including female Job Corps students, young gay men, alcoholics in treatment, and psychiatric patients.
While there is some evidence that HIV rates among gay males in some cities have plateaued (San Francisco Department of Public Health, 1994), other trends give cause for alarm. For example, the overall prevalence of HIV among Job Corps students, at 3 per 1,000 is high given their youth (Conway et al., 1993). Among female Job Corps students, the HIV rate rose from 2.1 per 1,000 in 1988 to 4.2 per 1,000 in 1992. The rate among male students decreased during this time from 3.6 to 2.2 per 1,000.
One indicator of trouble is the relapse to unsafe sex by men who had adopted safer practices (Stall et al., 1990; Stall et al., 1992). A second indicator of possible negative future trends is the high rate of unsafe behavior reported by the low-risk population (Catania et al., 1992; Tanfer et al., 1993). In fact, rates of risk behavior may provide a more accurate window to the future than that offered by AIDS case rates or HIV seroprevalence.
With respect to injection drug users, numerous HIV seroprevalence studies have been conducted in many cities around the world, and have generated prevalence estimates ranging from 0 percent to greater than 50 percent of the relevant populations. A number of these cities have seroprevalence rates of 30 percent or higher among drug injectors, and several have seen the rapid spread of HIV among drug injectors soon after the introduction of HIV into the drug-using population (Brown and Beschner, 1993; CDC, 1990, 1993b; Des Jarlais et al., 1989; Des Jarlais, Friedman, Choopanya, et al., 1992; Hahn et al., 1989).
Although most seroprevalence studies have been performed among convenience samples, and most have involved prevalent rather than incident HIV infections, they have generated useful information regarding the levels of HIV infection within these defined population groups. HIV incidence can be an important predictor for the future course of the epidemic. Although some studies have measured incidence among various highrisk groups (including gay men and injection drug users), there is a need for additional data on HIV incidence (Metzger et al., 1993; Moss et al., 1994; Rezza et al., 1990; van Ameijden et al., 1992; Vlahov et al., 1990; Vlahov, 1994). Such data provide insight into the spread of