To study the risk for human immunodeficiency virus (HIV) infection and the patterns of use and associated toxicity of zidovudine among health care workers after an occupational exposure to HIV. An ongoing, prospective surveillance project conducted by the Centers for Disease Control and Prevention. Exposed workers voluntarily reported by 312 U.S. health care facilities from August 1983 to June 1992. Four of 1103 enrolled workers with percutaneous exposure to HIV-infected blood seroconverted (HIV seroconversion rate, 0.36%; upper limit of the 95% Cl, 0.83%); no enrolled workers with mucous membrane (n = 75) or skin (n = 67) contact seroconverted. During October 1988 to June 1992, 31% of 848 enrolled workers used zidovudine after exposure; this proportion increased from 5% during October through December 1988 to 43% during January through June 1992. Despite using zidovudine after exposure, one worker became infected with a strain of HIV that was apparently sensitive to zidovudine. Adverse symptoms, most commonly nausea, malaise or fatigue, and headache, were reported by 75% of workers using zidovudine; 31% of workers did not complete planned courses of zidovudine because of adverse events. The risk for HIV seroconversion after percutaneous exposure to HIV-infected blood is 0.36%, which is similar to previous estimates. Zidovudine is used after exposure by a sizable proportion of health care workers enrolled in the project despite frequent, minor, associated symptoms. Documented failures of postexposure zidovudine prophylaxis, including in one worker enrolled in this study, indicate that if zidovudine is protective, any protection afforded is not absolute. Postexposure zidovudine, if used, requires careful consideration of possible risks and benefits.