The goals of this study were to quantify delays in diagnosing HIV infection in patients presenting for medical care prior to the identification of their HIV infection, and to determine characteristics of visits preceding the identification of HIV infection. This was a retrospective cohort study of consecutive newly diagnosed HIV infected patients in a large urban setting. Patients were predominantly minorities and identified from all inpatient and outpatient HIV tests performed between December 29, 1998 and December 27, 1999. Data were collected from all emergency department and clinic visits for each patient included in this study for a 3-year period preceding the first diagnosis of HIV infection. Two hundred seventy-eight patients tested positive for HIV during the study period. Of these, 76 (27%) met inclusion criteria. Twenty-three patients made a total of 53 health care visits preceding the diagnosis of their HIV infection. The median delay in diagnosis of HIV infection was 2 days (interquartile range [IQR], 1-22 days; range, 0-1093 days) for all subjects, although 19 (25%; 95% confidence interval [CI]: 16%-36%) patients had a delay in diagnosis of 30 days or more. Among those patients who had made prior health care visits, the median delay in diagnosis was 112 days (IQR, 33-690 days; range, 2-1093 days). No specific risk factor, historical clue, physical examination finding, or laboratory finding reliably identified patients with HIV infection. Documentation in the medical record of specific risk factors and clinical characteristics suggestive of HIV infection was poor. This study documented missed opportunities and delays in diagnosis of patients with unrecognized HIV infection. Clinicians must maintain a high index of suspicion for HIV infection in all patients.