Does the approach used to measure sexual identity affect estimates of health disparities differently by race? A randomized experiment from the National Survey of Family Growth

Abstract

The inclusion of “something else” as a response option for survey questions about sexual identity has been shown to significantly moderate estimated differences between sexual identity subgroups in the distributions of various health outcomes in national health surveys, specifically for substance use/misuse and reproductive health. This moderation effect may be larger depending on the respondent's race, possibly due to the intersectionality of multiple minority identities and compounding minority stressors. To test this hypothesis, we conducted secondary analyses of data from five consecutive years of the National Survey of Family Growth (2015–2019) [n = 18,903 (8510 males, 10,393 females)] in the United States, where respondents were randomly assigned to receive either a four-category version of the sexual identity question that included “something else” as a response option along with lesbian/gay, bisexual, and heterosexual, or a three-category version that omitted “something else.” We focused on model-based estimates of distributions of health outcomes across subgroups defined by both race and sexual identity and tested the race moderation hypothesis using three-way interactions involving sexual identity, question type received (four-category vs. three-category), and race (Black, White, and other). Based on this randomized experiment, we found evidence of significant differences between the four-category and three-category question versions of sexual identity, in terms of the associations between sexual identity and several health outcomes, along with moderation of these differences by the respondent's race. For some outcomes, such as wanting a child or another child, the magnitudes of the estimated sexual identity differences among racial minority individuals, specifically between bisexual and heterosexual women, were sensitive to the measurement of sexual identity. These results further suggest a need for improved sexual minority measurement, especially among racial minorities. These differences were also mostly found among women, suggesting further intersectionality of groups exposed to specific stressors.