impact of the modeled policies. Due to uncertainty in inputs, we specified uncertainty ranges and conducted sensitivity analyses.

The following discussion is divided into four categories: HIV population characteristics, HAART use, HIV clinical services and costs, and effects of financing options. The text parallels Table A-1.

According to Fleming et al. (2002), approximately 950,000 individuals are infected with HIV. Among those infected, Fleming further estimates that approximately 670,000 are aware of infection. Of these, approximately 360,000 have AIDS (CD4 < 200, or AIDS-defining condition). Based on data from in-care populations, we assume that 150,000 of these have AIDS by clinical criteria and 210,000 by CD4 < 200 (Bozzette et al., 1998). As individuals are more likely to be aware of HIV infection later in disease (e.g., if infected for longer and/or symptomatic), we assumed a greater number of aware individuals in each more severe disease state.

Information regarding income level and insurance status of HIV-infected individuals was obtained from HCSUS, a national probability sample of people with HIV in care. According to Bozzette et al. (1998), the majority (72 percent) of those with HIV are low income (household income < $25,000). Among these low-income individuals, nearly 25 percent are uninsured and 61 percent rely on Medicaid or Medicare. In contrast, of those with incomes > $25,000, most (78 percent) rely on private insurance (Bozzette et al., 1998).

Information regarding disease stage and insurance status was also obtained from HCSUS (not shown in table). Individuals with clinical AIDS are about two-thirds as likely to have private insurance as are asymptomatic individuals, and are correspondingly more likely to have Medicaid. This reflects the impoverishing effects of severe AIDS as well as the AIDS disability requirement for HIV-associated Medicaid eligibility.

Estimates of the current prevalence of HAART use are drawn from a number of sources. Studies of HIV-infected populations (AIDS and HIV non-AIDS) in New York State and in three metropolitan areas used local data sources (HIV/AIDS surveillance, lab reporting, Medicaid and AIDS Drug Assistance Program [ADAP] billing claims) from 2001 in a framework endorsed by the Health Resources and Services Administration (HRSA) to estimate participation in HIV care (Kahn, 2002). Studies of Medicaid and ADAP populations were conducted for 1998 in four heavily HIV-affected states (Kahn et al., 2002). Additional older estimates provide