The recent report from the Centers for Disease Control and Prevention [1] identified 113 patients who were seropositive for the human immunodeficiency virus (HIV) but who were considered unlikely to have acquired the infection in a medical setting. This conclusion was based on molecular biology test results and by the classification of most of these patients into groups with increased risk for HIV infection.

The accuracy of molecular biology as a test has limitations, however, and was only done on isolates from 30 patients. I believe that the exclusion characteristics regarding risk category are too rigid and poorly defined [1, 2]. The categories "multiple sexual partners," "heterosexual transmission," and "history of sexually transmitted diseases" are potentially very large groups and should not be considered the only exclusion categories. Even with these overly strict categories, five patients had "no risk factors" identified.

The authors offer no comment on whether an attempt was made to elicit details of illnesses related to HIV seroconversion. Given the often long and asymptomatic nature of HIV infection, this history can be a vital clue to the likely date of infection. In a recently identified cluster of patients with medically acquired HIV infection in Australia, this factor was the most important in identifying minor surgery as the cause of HIV infection [3].

A detailed report on one of these health care workers [2] notes that he was mentally impaired, practiced in an area with a high prevalence of HIV infection, had poor infection-control techniques, and used defective disinfection and cleaning processes. It would therefore be amazing if HIV was not sporadically transmitted from patient to patient. However, the conclusion and title of the report strongly implied that the virus was not transmitted in this practice, even though less than 20% of the patients were tested for HIV [2].

At least six cases of health care worker-to-patient transmission of HIV have occurred in about 24 000 patients [1]. If you add to this number patients who had no identifiable risk factors and patients excluded because of overly strict risk factor application, HIV may have been transmitted in as many as 20 patients. We are probably not seeing an infection rate of 1 per 1000 patients. However, the only way to identify the real risk is through further investigations. Unfortunately, I believe that misleading titles of associated reports [2] and the recommendation that retrospective patient notification (and presumably investigation) need not be done routinely [1] will significantly decrease the likelihood that sufficient data are collected and the real risk quantified.