To Be or Not to Be Bayesian Correction for Misclassification of Self-reported Sexual Behaviors Among Men Who Have Sex with Men

Background: Inferring sexual behavior of a stigmatized minority through self-reported sexual identity is subject to misclassification and can lead to biased results. We quantify the degree of this misclassification and perform a Bayesian correction of the risk of HIV infection in relation to self-reported sexual behavior. Methods: Sensitivity and specificity of self-reported men who have sex with men in ascertaining sexual behavior was derived from validation data, as was the informative prior on the association of same-sex behavior with self-reported HIV infection. Using these priors, we performed two separate Bayesian analyses of National Epidemiologic Survey on Alcohol and Related Conditions data for the odds of self-reported HIV positivity, adjusting for differential misclassification of self-reported same-sex behavior indicated by either partner gender or sexual identity. Results: We found differential exposure misclassification with specificity exceeding sensitivity, and higher misclassification rates based on sexual identity compared with partner gender. Sexual identity and partner gender displayed different associations with HIV infection in the raw data but these became virtually identical when adjusted for estimates of misclassification of sexual behavior by these two indicators. The estimate of prevalence of same-sex behavior associated with an elevated risk of HIV infection decreased after adjustment for misclassification. Conclusions: Studies of risk due to same-sex behavior are likely biased when they rely on self-identification for ascertainment of risk factors, especially when self-reported identity is used. The implications of our findings on risk modeling cannot be assumed to be trivial due to substantial shifts in distributions of risk and prevalence of exposure.