Sex Misclassification in the BRFSS and its Implications for Assessing Transgender and Gender Nonconforming Reproductive Health: A Quantitative Bias Analysis

National surveys based on probability sampling methods, such as the Behavior Risk Factor and Surveillance Survey (BRFSS), are crucial tools for unbiased estimates of disparities in health and healthcare access among gender minorities. The BRFSS began offering an optional sexual orientation and gender identity module in 2014, capturing transgender and gender nonconforming (TGNC) identity among respondents. Although the BRFSS provides much needed data on the TGNC community, self-identified TGNC respondents are still vulnerable to misclassification of sex assigned at birth. This study applied quantitative bias analysis to explore the magnitude and direction of the systematic bias present in measures of reproductive health that result from this misclassification. Our findings suggest that TGNC individuals with misclassified sex assigned at birth are demographically distinct from those who are asked sex-specific questions, suggesting that there is significant selection bias present in the BRFSS measures of reproductive health. We use multivariate Poisson regression with robust standard errors to estimate the association between gender identity and four sex-specific outcomes: lifetime PSA testing, lifetime Pap testing, hysterectomy, and pregnancy. We then compare two analytic methods to explore the degree of bias in these estimates, including single and multiple imputation models and simple probabilistic bias adjustments. Our results demonstrate that estimated associations among gender nonconforming respondents are the most vulnerable to small degrees of bias, while estimates among transgender women and men are more robust to small and moderate degrees of bias. This study provides evidence that BRFSS data report non-representative samples of TGNC individuals who are asked questions about their reproductive health, and that these outcomes are subsequently vulnerable to bias. Therefore, researchers who use BRFSS data to examine reproductive health outcomes should not reported results as unbiased population-based estimates. This study further demonstrates that implementation of validated sex assigned at birth and gender identity question in national surveys is critical.