Accuracy of HIV-related risk behaviors reported by female sex workers, Iran: a method to quantify measurement bias in marginalized populations.

We quantified discrepancies in reported behaviors of female sex workers (FSW) by comparing 63 face-to-face interviews (FTFI) to in-depth interviews (IDI), with corroboration of the directions and magnitudes of reporting by a panel of psychologists who work with FSW. Sensitivities, specificities, positive and negative predictive values (PPV and NPV) were assessed for FTFI responses using IDI as a "gold standard". Sensitivities were lowest in reporting symptoms of sexually transmitted infections (63.9 %), finding sex partners in venues (52.4 %) and not receiving HIV test results (66.7 %). Specificities (all >83 %) and PPVs (all >74.0 %) were higher than NPV. FSW significantly under-reported number of clients, sexual contacts and non-condom use sex acts with clients and number of days engaging in sex work in the preceding week. This study provides a quantified gauge of reporting biases in FSW behaviors. Such estimates and methods help better understand true HIV risk in marginalized populations and calibrate survey estimates accordingly.