This survey presents several limitations. As it is the case for all cross-sectional studies, it is not possible to know if the risk factors associated with HIV pre-existed the time of infection. The pooling of population sampled with different methodologies threatens the validity of the statistical tests in bivariable analysis. However, this heterogeneity was somehow addressed in the regression analysis using GEE. Syphilis testing did not include a quantitative RPR test; hence syphilis test results included both active syphilis and serologic scars.