Prevalence of Self-Reported Diabetes by Sexual Orientation: Results from the 2014 Behavioral Risk Factor Surveillance System

Data source

Data for this study come from the 2014 BRFSS, a cross-sectional, nationally representative survey of the civilian, noninstitutionalized population aged 18 years and older. The BRFSS is conducted annually by the Centers for Disease Control and Prevention (CDC) in conjunction with state health departments in all 50 states plus the District of Columbia. The BRFSS has been administered since 1984, and to better reflect the changing demographic structure of the United States, cell phone only households were added to the sampling frame in 2011.³⁹ Approximately 450,000 adults are randomly selected for the survey every year and are asked a standard core set of questions, including information about sociodemographic characteristics, health conditions, and healthcare access. In addition, states have the option to add BRFSS-supported modules on specific topics, or states can develop and include their own questions in their statewide BRFSS.

Study sample

Currently, the BRFSS core questionnaire does not ask participants to identify their sexual orientation, but several states have independently added sexual orientation questions to their own BRFSS surveys in previous years.^17,18,40 State-added questions are not submitted to the CDC, so analyzing BRFSS data on sexual minority populations across state borders previously required permission from each individual state. Starting in 2014, the BRFSS offered states an optional and unified sexual orientation module, and the following 19 states added sexual orientation questions to their statewide BRFSS surveys: Delaware, Hawaii, Idaho, Indiana, Iowa, Kansas, Kentucky, Louisiana, Maryland, Minnesota, Montana, Nevada, New York, Ohio, Pennsylvania, Vermont, Virginia, Wisconsin, and Wyoming.

In the BRFSS sexual orientation module, respondents were asked, “Do you consider yourself to be: straight, lesbian or gay, bisexual.” Our final sample included 3776 adults who identified as lesbian, gay, or bisexual and 142,852 adults who identified as straight or heterosexual. Our analysis excluded respondents who indicated their sexual orientation as something other than the response options (n = 441), respondents who did not know the answer (n = 1161), or respondents who refused to answer (n = 2556). The nonresponse rate for the BRFSS sexual orientation question was similar to the nonresponse rate for other sensitive questions (i.e., race/ethnicity).⁴¹ The median state-level response rate for the BRFSS was 47% in 2014 (ranging from 25.1% in California to 60.1% in South Dakota).⁴²

Diabetes measures and covariates

We classified respondents as having diagnosed diabetes if they answered “yes” to the question, “Has a doctor, nurse, or other health professional ever told you that you have diabetes?” Women indicating gestational diabetes during pregnancy and adults indicating they were told they had prediabetes were not considered to have diabetes. We excluded respondents who indicated they did not know (n = 160) or refused to answer (n = 60). No participants in our study sample had missing information regarding diabetes diagnosis. Other demographic, socioeconomic, health behavior, and healthcare categorical variables used in this study were sexual orientation (lesbian, gay, bisexual, or heterosexual); age in years (18–29, 30–39, 40–49, 50–64, 65 or older, missing); race and ethnicity (White, Black, Hispanic, other/multiple races, and missing); relationship status (married or living with an unmarried partner, formerly married [divorced, separated, or widowed]), never married, refused/do not know; reporting any exercise in the past 30 days (yes, no, and missing); being a current smoker, former smoker, or lifetime nonsmoker; having consumed ≥7 drinks (women) or ≥14 drinks (men) per week (yes, no, and missing); educational attainment (less than high school, high school graduate, some college, college graduate, and missing); employment status (employed, unemployed, not in paid work force, do not know/not sure, and missing); household income in U.S. dollars (<$10,000, $10,000–24,999, $25,000–34,999, $35,000–49,999, $50,000–74,999, $75,000 or greater, and missing); body mass index (BMI) measured in kg/m² (<25, ≥25<30, ≥30, and missing); self-reported health status (excellent/very good/good, poor/fair, and missing); having health insurance (insured, uninsured, and missing); having a usual source of care (has usual care, no usual care, and refused/missing); having a routine checkup in the prior 12 months (yes, no, and missing); and having an unmet care need because of cost (yes, no, and missing) in the prior 12 months.

Analytic methods

We used descriptive statistics to characterize the study sample; weighting was used for all analyses. Then, we examined the prevalence of having been diagnosed with diabetes by sex and sexual orientation, first for the entire sample and then for select characteristics. We also estimated a series of logistic regression models to compare the prevalence of diabetes between lesbian/gay, bisexual, and heterosexual adults while incrementally adjusting for sociodemographic characteristics, health behaviors, and access to care measures. Results from the logistic regression models are presented as adjusted odds ratios (AORs) with 95% confidence intervals (CIs). With the exception of sexual orientation where respondents with missing data were excluded a priori, respondents' missing data for one or more predictor variables were retained in our regression analyses; missing data retained in analyses were coded as their own category. We conducted all analyses in R 3.3.1⁴³ using the survey package version 3.31-2,⁴⁴ to weight the data to account for the complex survey design of the BRFSS according to the technical instructions provided by CDC.⁴⁵ The Vanderbilt Institutional Review Board designated that this study did not involve human subjects research and granted exempt approval for this study.

Demographic, socioeconomic, and health characteristics

Table 1 presents the demographic, socioeconomic, and health characteristics of adults in the study sample by sex and sexual orientation. Gay and bisexual men were younger and more racially and ethnically diverse than heterosexual men. A comparatively lower percentage of gay and bisexual men reported being married or living with a partner. Gay men more often reported a BMI of <25 kg/m² compared with heterosexual men. Gay men most often reported and bisexual men least often reported having completed a college education. Compared with both gay and heterosexual men, bisexual men had lower incomes, more often lacked health insurance, and more often had an unmet healthcare need because of cost. Bisexual men also less frequently reported visiting a doctor for a routine checkup in the past 12 months and reported overall poor or fair health more often.

Similar to sexual minority men, lesbian and bisexual women were also younger and more racially and ethnically diverse than their heterosexual counterparts, with these trends being particularly pronounced for bisexual women. Lesbian and bisexual women reported being married or partnered less often than heterosexual women, with this trend strongest for bisexual women. Both lesbian and bisexual women reported higher BMIs than heterosexual women. Lesbian women more commonly reported that they had completed some college or hold college degrees than either heterosexual women or bisexual women. Bisexual women often indicated that they did not complete high school—nearly 1 in 5 bisexual women did not have a high school diploma, compared with ∼1 in 10 heterosexual women and 1 in 20 lesbian women.

Lesbian women more often reported incomes at the high and low extremes of the income scale, with more lesbian women reporting household incomes of <$10,000 and $75,000 or more per year than heterosexual women, respectively. Bisexual women had the lowest incomes of all sexual minority groups, with nearly 50% reporting household incomes <$35,000; >10% of bisexual women reported that they earned <$10,000 per year. Compared with heterosexual women, a lower percentage of lesbian and bisexual women had health insurance. Lesbian and bisexual women also more often reported having an unmet healthcare need because of cost, and less frequently reported visiting a doctor for a routine checkup in the past 12 months. Bisexual women more often reported fair or poor health than either lesbian or heterosexual women.

Prevalence of diabetes by sex and sexual orientation

Tables 2 and 3 present the prevalence of diabetes by sexual orientation and for select characteristics for men and women, respectively. Overall, 14.2% (95% CI 9.7–18.8) of bisexual men, 11.4% (95% CI 8.6–14.2) of gay men, and 10.8% (95% CI 10.4–11.2) of heterosexual men had an unadjusted self-reported diabetes diagnosis, as did 8.5% (95% CI 5.2–11.8) of lesbian women, 5.7% (95% CI 4.0–7.3) of bisexual women, and 10.2% (95% CI 9.8–10.5) of heterosexual women. Age-stratified analyses revealed that gay men and lesbian women had higher diabetes prevalence estimates than their heterosexual counterparts in every age category; bisexual men and women had the highest overall diabetes prevalence estimate for groups in the two oldest age categories, although none of these findings were significant. Among heterosexual men and women, Black adults were found to have a statistically higher prevalence of diabetes than did White adults. In contrast, among sexual minorities, no significant differences in diabetes prevalence among racial/ethnic minorities were found.

Odds of diabetes prevalence

Next, we performed a series of binomial logistic regression models to generate AORs to compare the prevalence of diabetes by sexual orientation. After adjusting for age, race/ethnicity, and BMI (Table 4, Model 3), both gay men (OR = 1.72; 95% CI 1.29–2.31) and bisexual men (OR = 1.69, 95% CI 1.12–2.56) had increased odds of a lifetime diabetes diagnosis. In contrast, lesbian women had slightly increased odds (OR = 1.11, 95% CI 0.72–1.72), whereas bisexual women had slightly decreased odds (OR = 0.93, 95% CI 0.66–1.33) of diabetes compared with heterosexual women, although these findings were not significant.

After adjusting for socioeconomic characteristics, health behaviors, and healthcare access variables (Table 4, Model 7), the heightened diabetes prevalence remained for gay men (OR = 1.50, 95% CI 1.09–2.07) and bisexual men (OR = 1.55, 95% CI 1.00–2.07). After performing the same analysis among women, lesbian women still had increased odds (OR = 1.22, 95% CI 0.76–1.95), whereas bisexual women had lower odds (OR = 0.88, 95% CI 0.62–1.26) of having diabetes than heterosexual women, although again these findings were not significant.

Limitations

There were several limitations to using the 2014 BRFSS for this study. First, all responses to the BRFSS were self-reported, which can lead to recall and response bias when describing height, weight, access to care, sociodemographic characteristics, and health conditions, including diagnoses for diabetes. In addition, reporting sexual orientation may suffer from selection bias. Our sample of LGB adults only includes noninstitutionalized adults randomly selected among landline and cell phone users who were comfortable disclosing their sexual orientation to BRFSS surveyors. Missing from our analysis were homeless adults and adults residing in institutionalized settings, including nursing homes, institutionalized medical facilities, incarceration facilities, and homeless shelters. Thus, our study may be missing LGB individuals from the lower end of the socioeconomic spectrum who were not captured by the BRFSS and who are over-represented in these populations.⁴⁷

Our results may not be generalizable to the entire U.S. sexual minority population, as our study only included data from 19 states. However, these states were geographically diverse. Nevertheless, our study would have benefited if the optional sexual orientation and gender identity module had been added to the core BRFSS questionnaire.

Our study would have also benefited from additional data missing in the BRFSS. For example, the BRFSS does not measure other dimensions of sexual orientation, including sexual behavior or sexual attraction. Thus, our study does not consider individuals who are sexually active with or attracted to people of the same gender but who do not identify as LGB. BRFSS also does not measure variables that could contribute to the higher diabetes prevalence observed, in particular, for sexual minority men, such as the use of HIV antiretroviral medications, nor does the BRFSS include objective measures of diabetes (e.g., study-assessed biometrics such as blood glucose levels or glycated hemoglobin tests for HbA1c levels) and diabetes risk factors (e.g., study-assessed body weight and waist measurements). Finally, the BRFSS is a cross-sectional survey and cannot definitively establish the causal directions of the observed associations between sexual orientation and diabetes, as cross-sectional studies are prone to omitted variable bias. Missing and unmeasured variables—such as structural stigma and dysregulation of the HPA axis due to minority stress and depression⁴⁸—may provide alternative explanations for the relationship between sexual orientation and health outcomes.⁴⁹

Notwithstanding these limitations, we use data from a large population-based survey to analyze the prevalence of diabetes and diabetes risk factors by sexual orientation—making this one of the first studies to focus in depth on diabetes among sexual minorities. We found that both gay and bisexual men were more likely to be living with diabetes than heterosexual men. Our results serve as a reminder that not only are LGB adults more likely to have mental health conditions and substance use disorders, but increasingly, LGB adults have been shown to have higher prevalence of physical health chronic conditions, such as diabetes. Healthcare providers should consider the higher burden of mental health conditions, substance use, and stress levels among sexual minorities and be mindful of how these factors, including any medications used to address them, may contribute to the physical health disparities increasingly described among sexual minority populations, including diabetes,^23,38 cardiovascular disease,^38,50 and cerebrovascular disease.⁵¹ Healthcare providers should also continue their efforts to build welcoming and safe clinical environments for their sexual minority patients. Building trust between providers and sexual minority patients will help facilitate discussions on recommended screenings and best practices for reducing risk factors for diabetes. Meanwhile, future studies should investigate whether the development of enhanced screening guidelines for physical health conditions, such as diabetes, among sexual minority populations, is warranted.