Early detection of HIV infection improves prognosis and reduces transmission, but 30%-40% of cases are diagnosed late. A comprehensive and systematic review of medical encounters before diagnosis has not been done. This study reviews 5 years of medical encounters before the diagnosis of HIV infection in members of a large managed care organization where access to care is reasonably good. Patient characteristics, HIV risk factors, and clinical events preceding diagnosis were examined and tested for association with late diagnosis (CD4 cell count of <200/microL at diagnosis). Of 440 HIV-infected patients, 62% had CD4 cell counts of <350/microL, 43% had CD4 cell counts of <200/microL, and 18% had CD4 cell counts of <50/microL at diagnosis. Twenty-six percent of all patients had risks documented >1 year before diagnosis. Only 22% of patients had one of eight clinical indicators suggested in the literature as reasons to test for HIV >1 year before diagnosis. In multiple logistic regression, older age, male sex, race, risk group, no prior HIV testing, physician-initiated testing, and having any of eight clinical indicators before diagnosis were each associated with late diagnosis (p <or=.05). Late diagnosis remains a challenge despite good access to care. In our setting, effective risk assessment before symptoms arise offers greater potential for raising the mean CD4 cell count at diagnosis than does increased awareness of selected HIV-associated clinical prompts.