The ANRS-EN12-VESPA 2003 survey interviewed a national, representative sample of the HIV-positive population attending hospital outpatient services. Conducting the survey in a hospital made it easier to identify and contact HIV-positive patients, but involved various biases: the role of the physicians in choosing which patients would be interviewed, lack of time for certain categories of patients, oral and written language skills required to answer questions. The survey design enabled us to study these biases in detail. Overall, significant biases were observed, but the resulting distortions in the sample were quite small. Among the factors responsible for non-participation in the survey, those related to the patient weighed more than those linked to survey organization. Alongside the fact that persons infected through homosexual transmission were more likely to respond, which probably reflects a greater ability and willingness to participate, we also noted the importance of work and family constraints, whose effects are probably sexually differentiated, as well as the language barrier, which puts foreigners at a disadvantage, especially for the written questionnaire.