During February-November 1991 in Abidjan, the Ivory Coast, interviews with 1257 male first-time blood donors and screening of their blood for antibodies to HIV-1 and HIV-2 were conducted at the National Blood Transfusion Center to determine whether it is feasible to exclude or defer donors at high risk of HIV infection. 143 (11.4%) men had antibodies to HIV. The multivariate analysis showed that the strongest risk factor for HIV infection was sex with prostitutes in the last 5 years without regular condom use (odds ratio [OR] = 10, confidence interval [CI] = 6.6-15) followed by prostitute contact in the last 5 years regardless of condom use (OR = 8.4, CI = 5.6-12.6), and an age of 30-39 years (OR = 6.9, CI = 18). 27% of all donors with prostitute contact in the last 5 years were infected (positive predictive value). If the center had used prostitute contact in the last 5 years as a criterion, it would have discarded 31% of all units of blood. Thus, more than 75% of all potentially excluded units would have been from donors infected with HIV. The sensitivity, specificity, and positive predictive values ranged from 15% to 98%, 38% to 91%, and 17% to 30%, respectively. These findings suggest that it is feasible in Abidjan to exclude donors based on behavioral characteristics of those most at risk of HIV infection. Donor deferral should be evaluated as a strategy to improve blood safety in resource-poor areas with high rates of HIV infection.