BACKGROUND: An extensive literature supports expanded HIV screening in the United States. However, the question of whom to test and how frequently remains controversial. OBJECTIVE: To inform the design of HIV screening programs by identifying combinations of screening frequency and HIV prevalence and incidence at which screening is cost-effective. DESIGN: Cost-effectiveness analysis linking simulation models of HIV screening to published reports of HIV transmission risk, with and without antiretroviral therapy. DATA SOURCES: Published randomized trials, observational cohorts, national cost and service utilization surveys, the Red Book, and previous modeling results. TARGET POPULATION: U.S. communities with low to moderate HIV prevalence (0.05% to 1.0%) and annual incidence (0.0084% to 0.12%). TIME HORIZON: Lifetime. PERSPECTIVE: Societal. INTERVENTIONS: One-time and increasingly frequent voluntary HIV screening of all adults using a same-day rapid test. OUTCOME MEASURES: HIV infections detected, secondary transmissions averted, quality-adjusted survival, lifetime medical costs, and societal cost-effectiveness, reported in discounted 2004 dollars per quality-adjusted life-year (QALY) gained. RESULTS OF BASE-CASE ANALYSIS: Under moderately favorable assumptions regarding the effect of HIV patient care on secondary transmission, routine HIV screening in a population with HIV prevalence of 1.0% and annual incidence of 0.12% had incremental cost-effectiveness ratios of 30,800 dollars/QALY (one-time screening), 32,300 dollars/QALY (screening every 5 years), and 55,500 dollars/QALY (screening every 3 years). In settings with HIV prevalence of 0.10% and annual incidence of 0.014%, one-time screening produced cost-effectiveness ratios of 60,700 dollars/QALY. RESULTS OF SENSITIVITY ANALYSIS: The cost-effectiveness of screening policies varied within a narrow range as assumptions about the effect of screening on secondary transmission varied from favorable to unfavorable. Assuming moderately favorable effects of antiretroviral therapy on transmission, cost-effectiveness ratios remained below 50,000 dollars/QALY in settings with HIV prevalence as low as 0.20% for routine HIV screening on a one-time basis and at prevalences as low as 0.45% and annual incidences as low as 0.0075% for screening every 5 years. LIMITATIONS: This analysis does not address the difficulty of determining the prevalence and incidence of undetected HIV infection in a given patient population. CONCLUSIONS: Routine, rapid HIV testing is recommended for all adults except in settings where there is evidence that the prevalence of undiagnosed HIV infection is below 0.2%.