Background: Self-reported condom use has been used for three decades in behavioral research. However, reporting bias and its association with sexually transmitted infections (STIs) including HIV have seldom assessed in large-scale studies. The objectives of this study were to use a biologic measure to validate self-reported condom use and to investigate if reporting bias was associated with syphilis among older female sex workers (FSWs) in China.
Methods: The study was conducted among 1,245 older FSWs who were 35 years old and older in three cities (about 400 per each). Respondent-driving sampling was used to recruited study subjects. Prevalent and active syphilis was tested in blood samples. Prostate-specific antigen (PSA) was tested in vaginal secretion samples. The presence of PSA indicates unprotected sex. If an older FSW reported having had no sex in the past 24 hours and PSA was present, she was classified as having a discordant report on no sex. Similarly, if an older FSW reported having had only protected sex in the past 24 hours, but PSA was present, she was classified as having a discordant report on protected sex.
Results: The prevalence of syphilis was 23% (95%CI: 21-26%) and the rate of active syphilis was 10% (95%CI: 9-12%) among older FSWs. Among 493 older FSWs who reported having had no sex in the past 24 hours, 135 were tested positive for PSA, resulting the proportion of discordant report in no sex was 27%. Among 445 older FSWs who reported having used condoms for every sexual intercourse, 166 were tested positive for PSA, i.e., the proportion of discordant report in protected sex was 37%. After controlling for confounding variables, older FSWs who had prevalent syphilis were more likely to have discordant report in no sex [prevalence ratio (PR) = 1.77; 95%CI, 1.29-2.42)] and in protected sex (PR =1.56; 95%CI, 1.14-2.14) than those who did not have prevalent syphilis. Similarly, older FSWs who had active syphilis were more likely to have discordant report in no sex (PR =1.94; 95%CI, 1.37-2.74) and in protected sex (PR =1.65; 95%CI, 1.12-2.43) than those who did not have active syphilis. Similar results were presented in separated analyses across the three study sites.
Conclusions: Self-reported condom use has strong reporting bias and may not be used as an outcome to measure effectiveness of STI interventions as the bias is also associated with occurrence of STIs. PSA test provides a valid bio-approach for assessing actual condom use and can be used in large scale studies.