We aggregated WA-BRFSS data from 2003 to 2010. WA-BRFSS is an annually conducted telephone survey for noninstitutionalized adults aged 18 years and older.¹² Beginning in 2003, WA-BRFSS included a measure of sexual orientation. Data from 2003 to 2010 were aggregated to create a sufficient sample to test the study research questions. All variables included in the study have been consistently asked each year.

The categories of race and ethnicity include non-Hispanic White, African American, Asian American, Native Hawaiian or other Pacific Islander, American Indian or Alaskan Native, Hispanic, Multiracial, and other race or ethnicity. We merged Asian American with Native Hawaiian or other Pacific Islander, and excluded Multiracial and other race or ethnicity from further analyses because of insufficient sample size. First, we calculated weighted prevalence estimates of background characteristics and the 4 sexual orientation responses—including (1) “heterosexual” or “lesbian, gay, or bisexual,” (2) “other,” (3) “not sure or don’t know,” and (4) “refusing to answer”— by race and ethnicity and compared the estimates based on 95% confidence intervals. Second, we used adjusted multinomial logistic regression to examine the odds comparing respondents of color to non-Hispanic White respondents for self-identifying as “other,” responding “not sure or don’t know,” or refusing to answer relative to self-identifying as “heterosexual” or “lesbian, gay, or bisexual (LGB).” In the adjusted model, we controlled for confounding variables of age (in years), income (≤ 200% federal poverty level vs > 200% DHHS federal poverty level), and education (≤ high school vs ≥ some college) as nonresponse was associated with older age, lower income, and lower educational achievement (data not shown). Year of interview was also added to the model to account for the clustering of the data. Sample weights provided by WA-BRFSS were applied to all analyses.

Table 1 demonstrates weighted estimates of background characteristics and responses to sexual orientation question by race and ethnicity. The result of the χ² test indicates that there is a significant relationship between sexual orientation responses and race and ethnicity (χ²[12] = 1713.38; P < .001).When considering 95% confidence intervals, compared with non-Hispanic White Americans, African Americans, Asian Americans, and Hispanics were more likely to respond “not sure or don’t know”; Asian Americans and Hispanics were also more likely to refuse to answer. The rates of self-identifying as “other” were not different by race and ethnicity. The results of adjusted multinomial logistic regression analyses (Table 2) indicate that the adjusted odds of responding “not sure or don’t know” relative to self-identifying as “heterosexual or LGB” for African Americans, Asian Americans, and Hispanics were about 2.6, 12.5, and 6.4 times greater, respectively, than those for non-Hispanic Whites. The adjusted odds of refusing to answer relative to self-identifying as “heterosexual or LGB” for Asian Americans and Hispanics were about 4.4 and 2.0 times greater, respectively, than those for non-Hispanic Whites. The adjusted odds of nonresponse for American Indians and Alaskan Natives were not different from those for non-Hispanic Whites.

Our findings, based on WA-BRFSS, indicate that nonresponse to a commonly used sexual orientation question is associated with race and ethnicity. Furthermore, we discovered that there is diversity in nonresponses within racial and ethnic minorities. Asian Americans and Hispanics were more likely to respond “not sure or don’t know” as well as refuse to answer compared with non-Hispanic Whites.

Those who answered “not sure or don’t know” or refused to answer might be unable to interpret the stated categories of sexual orientation because the terms heterosexual, lesbian, gay, and bisexual may not be part of everyday language for some respondents.^4,13 In particular, the nonresponse rates for Asian Americans and Hispanics, which consist of relatively higher proportions of foreign-born immigrants,¹⁴ may be higher because of the unfamiliarity with the sexual orientation terms. For example, a previous study notes that foreign-born respondents in the United States are less likely to identify as lesbian, gay, or bisexual compared with US-born respondents.¹⁵ According to national surveys, many Asian Americans tend to identify their sexual orientation as “queer,”¹⁶ which is often excluded from analyses, and some of the sexual orientation terms used in US surveys are not translatable (e.g., no comparable word for “straight” in Spanish).¹⁷ Future research will need to investigate to what extent nonresponse among Asian Americans and Hispanics is related to fluency in US English or having different terms in their native languages. It will also be important to test whether including questions on sexual behavior and sexual attraction in addition to sexual orientation enhances our understanding of sexual identification among diverse racial and ethnic groups and if such an approach reduces cultural bias. A study indicates that foreign-born men who report having sex with men are less likely to identify their sexual orientation as gay than their US-born counterparts.¹⁸ Cognitive interviewing techniques to evaluate sources of response error in measurement and alternative measures of sexual identity may help develop culturally responsive measures related to sexuality.

We also observed some interesting similarities in nonresponses among racial and ethnic minority groups. Although African Americans showed slightly higher rates of responding “not sure or don’t know” than did non-Hispanic Whites, African Americans, and American Indians and Alaskan Natives did not show as salient nonresponses as did Asian Americans and Hispanics did when compared with non-Hispanic Whites. It has been suggested that the discourse and construction of sexual orientation in the United States has had influence across diverse racial and ethnic minority groups as well as non-Hispanic Whites.¹⁹ Thus, it may be problematic to simply assume that people of color are more likely to not identify their sexual orientation within the categories stated in a survey. Further investigation should be conducted to better understand the diversity in response patterns within racial and ethnicminority groups.

The findings may not be generalizable to the US population because the data used in this study are only representative of Washington State. Still, the findings emphasize the necessity of improving measurements of sexual orientation to reduce racial and ethnic bias. Existing evidence suggests cumulative risk of health disparities by sexual orientation and race and ethnicity.²⁰ It is imperative to have population-based data that can estimate prevalence of key health indicators for racially and ethnically diverse LGB populations. Thus, these nonresponses should not be simply ignored but rather further investigated and understood so that better measures can be constructed. Recently, DHHS announced that measures of sexual orientation will be added to the National Health Interview Survey by 2013.²¹ As we move forward in health disparities research, it is important for population-based surveys to consider the cultural sensitivity of sexual orientation measures. Only through addressing the increasing diversity in our society will be prepared to address and reduce health disparities.