Given the substantial uncertainty in our ability to specify numbers of injection drug users in the United States, it is likewise problematic to estimate accurately the characteristics of this elusive population. Yet existing surveys, drug abuse indicator databases, and other research offer considerable insight in beginning to identify the composition of injection drug users within our borders.
A review of available data reveals at least two predominant themes. First, the overall population of injection drug users in the United States is quite heterogeneous in composition. It is emphatically not the case that all injection drug users are male minorities located in large urban areas, as stereotypes would imply. Second, the composition and characteristics of this population are continually evolving and have changed markedly over time. These facts suggest that the useful improvements that can be made in prevalence estimation techniques are not limited merely to reducing the margins of error for a national approximation of numbers of injectors. To offer meaningful insight, future estimation models will also need to reflect important regional differences and be generated at regular intervals so as to account for variation over time.
NIDA's National AIDS Demonstration Research (NADR) programs provide the most comprehensive profile of active drug injectors not in drug abuse treatment in this country (Brown and Beschner, 1993). However, their description of the characteristics of injection drug users is based on a large sample of out-of-treatment injectors, so the information may be limited by a volunteer/selection bias. As noted earlier in this chapter, it is extremely difficult, if not impossible, to get a truly representative sample of injection drug users. The most feasible approach to generating sound inferences is to obtain information from multiple sampling schemes, which includes the efficiency of sampling entrants to drug treatment programs and entrants into the criminal justice system, as well as community-based recruitment schemes, as was done for NADR. Each scheme has limitations, but in combination they allow for sound inferences (Vlahov and Polk, 1988; Alcabes et al., 1991, 1992; Watters and Biernacki, 1989). In particular, surveys of entrants into treatment for drug abuse at multiple sites (Hahn et al., 1989; Prevots et al., 1995; Battjes et al., 1994) and prison-based surveys at multiple sites (Vlahov et al., 1991a) show similar patterns of geographic diversity of risk behaviors and HIV seroprevalence, which helps to validate estimates and inferences derived from the NADR database.
In the NADR study, a review of the sociodemographic characteristics