Sources of Missing Sexual Orientation and Gender Identity Data in the Behavioral Risk Factor Surveillance System

Abstract

This paper describes the sources, magnitude, and correlates of missing data in the Behavioral Risk Factor Surveillance System Sexual Orientation and Gender Identity module. Missing data from the Behavioral Risk Factor Surveillance System Sexual Orientation and Gender Identity module fielded from 2014 to 2019 were ascribed to 4 sources: the optional nature of the Sexual Orientation and Gender Identity module, out-of-state cell phone respondents, interview termination, and item nonresponse. The prevalence of missingness from these 4 sources was examined in relation to sociodemographic factors and survey process factors. Data were analyzed in 2018-2020. From 2014 to 2019, of 2,698,738 Behavioral Risk Factor Surveillance System respondents, 1,330,025 (44.8%, weighted) were in states that did not administer the Sexual Orientation and Gender Identity module. Among 723,301 cell phone interviews in states administering the module, 12.5% (weighted) were out of state. Among 1,316,174 otherwise potential respondents, 9.4% (weighted) terminated the interview before Sexual Orientation and Gender Identity module administration. Among 1,205,177 administered the module, item-level missingness was 3.4% for sexual orientation and 1.3% for gender identity. Correlates of missingness varied considerably at each stage. Missing the Behavioral Risk Factor Surveillance System Sexual Orientation and Gender Identity data is much more prevalent than item nonresponse alone would suggest. Analytic techniques that consider only item nonresponse, such as complete case analysis, risk producing biased findings. Including the Sexual Orientation and Gender Identity module in the required core demographics section is the only feasible method to reduce the amount and complexity of missing data.