Diana Tordoff1, Michele Andrasik2,3, Anjum Hajat1. 1. From the Department of Epidemiology, University of Washington, Seattle, WA. 2. Department of Global Health, University of Washington, Seattle, WA. 3. HIV Vaccine Trials Network, Fred Hutchinson Cancer Research Center, Seattle, WA.
Abstract
BACKGROUND: National surveys based on probability sampling methods, such as the Behavioral Risk Factor and Surveillance System (BRFSS), are crucial tools for unbiased estimates of health disparities. In 2014, the BRFSS began offering a module to capture transgender and gender nonconforming identity. Although the BRFSS provides much needed data on the this population, these respondents are vulnerable to misclassification of sex assigned at birth. METHODS: We applied quantitative bias analysis to explore the magnitude and direction of the systematic bias present as a result of this misclassification. We use multivariate Poisson regression with robust standard errors to estimate the association between gender and four sex-specific outcomes: prostate-specific antigen testing, Pap testing, hysterectomy, and pregnancy. We applied single and multiple imputation methods, and probabilistic adjustments to explore bias present in these estimates. RESULTS: Combined BRFSS data from 2014, 2015, and 2016 included 1078 transgender women, 701 transgender men, and 450 gender nonconforming individuals. Sex assigned at birth was misclassified among 29.6% of transgender women and 30.2% of transgender men. Transgender and gender nonconforming individuals excluded due to sex-based skip patterns are demographically distinct from those who were asked reproductive health questions, suggesting that there is noteworthy selection bias present in the data. Estimates for gender nonconforming respondents are vulnerable to small degrees of bias, while estimates for cancer screenings among transgender women and men are more robust to moderate degrees of bias. CONCLUSION: Our results demonstrate that the BRFSS methodology introduces substantial uncertainty into reproductive health measures, which could bias population-based estimates. These findings emphasize the importance of implementing validated sex and gender questions in health surveillance surveys. See video abstract at, http://links.lww.com/EDE/B562.
BACKGROUND: National surveys based on probability sampling methods, such as the Behavioral Risk Factor and Surveillance System (BRFSS), are crucial tools for unbiased estimates of health disparities. In 2014, the BRFSS began offering a module to capture transgender and gender nonconforming identity. Although the BRFSS provides much needed data on the this population, these respondents are vulnerable to misclassification of sex assigned at birth. METHODS: We applied quantitative bias analysis to explore the magnitude and direction of the systematic bias present as a result of this misclassification. We use multivariate Poisson regression with robust standard errors to estimate the association between gender and four sex-specific outcomes: prostate-specific antigen testing, Pap testing, hysterectomy, and pregnancy. We applied single and multiple imputation methods, and probabilistic adjustments to explore bias present in these estimates. RESULTS: Combined BRFSS data from 2014, 2015, and 2016 included 1078 transgender women, 701 transgender men, and 450 gender nonconforming individuals. Sex assigned at birth was misclassified among 29.6% of transgender women and 30.2% of transgender men. Transgender and gender nonconforming individuals excluded due to sex-based skip patterns are demographically distinct from those who were asked reproductive health questions, suggesting that there is noteworthy selection bias present in the data. Estimates for gender nonconforming respondents are vulnerable to small degrees of bias, while estimates for cancer screenings among transgender women and men are more robust to moderate degrees of bias. CONCLUSION: Our results demonstrate that the BRFSS methodology introduces substantial uncertainty into reproductive health measures, which could bias population-based estimates. These findings emphasize the importance of implementing validated sex and gender questions in health surveillance surveys. See video abstract at, http://links.lww.com/EDE/B562.
Authors: Diana M Tordoff; Brian Minalga; Bennie Beck Gross; Aleks Martin; Billy Caracciolo; Lindley A Barbee; Jennifer E Balkus; Christine M Khosropour Journal: Sex Transm Dis Date: 2022-02-01 Impact factor: 2.830
Authors: Oluwadamilola T Oladeru; Sung Jun Ma; Joseph A Miccio; Katy Wang; Kristopher Attwood; Anurag K Singh; Daphne A Haas-Kogan; Paula M Neira Journal: Am J Clin Oncol Date: 2022-03-01 Impact factor: 2.787
Authors: Diana M Tordoff; Sahar Zangeneh; Christine M Khosropour; Sara N Glick; Raymond Scott McClelland; Dobromir Dimitrov; Sari Reisner; Ann Duerr Journal: J Acquir Immune Defic Syndr Date: 2022-04-15 Impact factor: 3.771
Authors: Stephen J Mooney; Lin Song; Adam Drewnowski; James Buskiewicz; Sean D Mooney; Brian E Saelens; David E Arterburn Journal: Obesity (Silver Spring) Date: 2021-10-04 Impact factor: 5.002