To evaluate human immunodeficiency virus type I ( HIV-1) infection among patients of primary care physicians, we performed anonymous, unlinked HIV-I antibody testing on leftover blood specimens submitted to HI huge commercial clinical laboratories for complete blood cell count or hematocrit determination, the most commonly ordered diagnostic tests from January through August 1990. 55.613 specimens submitted by general internists. pediatricians, and family practitioners were sampled: 1.104 (2.0%) had HIV-I antibody. Seroprevalence among the laboratories varied 50-fold, from 0.3 to 12.4%. The HIV-I prevalence at each laboratory was not always consistent with the AIDS incidence in the area served by the laboratory. Overall the seroprevalence was almost eight times higher in men (3.9%) than in women (0.5%). Specimens from seropositive persons, especially from men. were unevenly distributed among the physician practices: only three practices submitted approximately 50% of all specimens from seropositive men. These data indicate that a few physicians treat the majority of HIV-1-infectcd primary care patients. The HIV-I prevalence among specimens at a clinical laboratory is thus determined by whether few physicians submit specimens to that laboratory. These results could be of use. for instance, in analyzing proposals to mandate physician reporting of HIV-I infection. The high HIV-I prevalence among laboratory specimens underscores the potential for exposure to HIV-I-infected blood by clinical laboratory personnel and emphasizes the need for universal precautions for all blood specimens.