Demonstrating the Importance and Feasibility of Including Sexual Orientation in Public Health Surveys: Health Disparities in the Pacific Northwest

Abstract

Objectives. We identified health disparities for a statewide population of lesbian, gay, and bisexual (LGB) men and women compared with their heterosexual counterparts.

Methods. We used data from the 2003–2006 Washington State Behavioral Risk Factor Surveillance System to examine associations between sexual orientation and chronic health conditions, health risk behaviors, access to care, and preventive services.

Results. Lesbian and bisexual women were more likely than were heterosexual women to have poor physical and mental health, asthma, and diabetes (bisexuals only), to be overweight, to smoke, and to drink excess alcohol. They were also less likely to have access to care and to use preventive services. Gay and bisexual men were more likely than were heterosexual men to have poor mental health, poor health-limited activities, and to smoke. Bisexuals of both genders had the greatest number and magnitude of disparities compared with heterosexuals.

Conclusions. Important health disparities exist for LGB adults. Sexual orientation can be effectively included as a standard demographic variable in public health surveillance systems to provide data that support planning interventions and progress toward improving LGB health.

A limited number of studies have described chronic disease health risks among lesbian, gay, and bisexual (LGB) adults. Few of these studies have been population based, and those were often conducted in limited geographic areas or did not include a heterosexual comparison group. Most have relied on convenience samples or other targeted study designs and studied only specific health issues such as smoking or HIV risk factors.

The patchwork of available studies indicates that LGB adults have important health disparities. Compared with heterosexual women, lesbian and bisexual women have been shown to have poorer overall health and mental health^1–3; higher rates of smoking,^4–11 alcohol consumption,^8,11–13 asthma,^3,13 and obesity²; and less access to health care,¹² including routine preventive screenings such as Papanicolaou (Pap) tests or mammograms,^14–16 although they were more likely to have had HIV tests.¹⁵ Gay and bisexual men have reported higher rates of smoking^4–6,9 and alcohol use⁸ and poorer general health and mental health^1,3 compared with heterosexual men. An LGB companion document to the Healthy People 2010 initiative identified 29 specific objectives that prioritized sexual minorities, but data by sexual orientation were not available in public health surveillance systems to track most of those objectives.¹⁷ In a recent review of sexual and gender minority health issues, Mayer et al.¹⁸ called for more inclusion of sexual minority identifiers in national data sets as a necessary next step in elimination of health disparities.

In 2003, Washington began to include a question about sexual orientation in its Behavioral Risk Factor Surveillance System (BRFSS). The BRFSS is a telephone-based survey of adults that is sponsored by the Centers for Disease Control and Prevention (CDC) and implemented throughout the United States.¹⁹

For previous analyses, we combined data from 2 states (Washington and Oregon) collecting LGB information in BRFSS since 2003 into a single data set to gain sufficient numbers to describe LGB smoking behaviors.⁶ After 4 years of data collection, we now have enough LGB respondents from Washington alone to examine a variety of indicators. The purpose of this study was to describe a variety of health indicators for a statewide population of LGB men and women compared with their heterosexual counterparts. By demonstrating the feasibility and relevance of collecting information on sexual orientation in the BRFSS, we provide justification for public health surveillance systems to progress beyond “don't ask, don't tell” policies.

METHODS

We used combined Washington State BRFSS data from 2003 to 2006.²⁰ The BRFSS is an ongoing, population-based, random-digit-dialed telephone survey. Eligible respondents were English- or Spanish-speaking noninstitutionalized adults who lived in a household with a landline telephone. Once an eligible household was reached, an adult was randomly selected from among the household's adults for participation. Sampling weights were calculated on the basis of each respondent's probability of selection, and data were also poststratified to the age and gender distribution of the Washington population. Annual response rates (calculated by the Council of American Survey Research Organizations [CASRO] method²¹) ranged between 43% and 47% for the 4 years of our study.

Measures

With the exception of sexual orientation and poverty level, we based all measures on CDC core questions that are fielded in every state's survey. Exact wording for the questions is available on the CDC Web site.¹⁹ A few questions were included only during either even- or odd-numbered years according to a planned rotation schedule, but most questions were asked every year. Responses in which the respondent answered “don't know” or refused to answer the question were dropped from analysis. Unless otherwise specified, we used established CDC categories for “at risk” status.¹⁹

Sexual orientation.

Respondents were asked, “Now I'm going to ask you a question about sexual orientation. Do you consider yourself to be (A) heterosexual or straight, (B) homosexual, gay, or lesbian, (C) bisexual, or (D) something else? Remember, your answers are confidential.” Respondents who answered “something else” or “don't know” or who refused to answer were not included in further analyses.

Demographic characteristics.

Respondents provided their exact age and the highest level of education completed. We combined separate race and ethnicity responses to classify respondents either as Hispanic (any race) or by their preferred race (for non-Hispanics). Classification as above, at, or below 200% of the federal poverty level was calculated on the basis of reported household income, number of adults and children in the household, and published federal poverty level guidelines for each specific year.²² We chose 200% of federal poverty level because in Washington this is the standard for eligibility for low-income health service plans.

Health measures.

We grouped respondents into “fair” or “poor” general health. For mental health, we classified respondents according to whether or not they reported 10 or more days of poor mental health during the past month. “Limited activities” were defined as activities that were limited in any way because of physical, mental, or emotional problems.

Respondents reported whether they had ever been told by a health professional that they had the following: diabetes (not including prediabetes or gestational diabetes alone), asthma (and still had asthma at time of survey), hypertension or high blood pressure (not included if borderline or during pregnancy alone), or high cholesterol.

Overweight (including obesity), as measured by body mass index (BMI; weight in kilograms divided by height in meters squared), was determined from self-reported height and weight. Overweight was defined as a BMI of 25 or higher.

“Heavy drinking” was defined as 2 or more drinks per day for men and 1 or more per day for women. “Current smoking” was defined as lifetime consumption of 100 or more cigarettes and current smoking “every day” or “some days.”

“Insufficient physical activity” was defined as not performing any of the following: moderate activity (brisk walking, biking, gardening) 30 minutes per day, 5 or more days per week; vigorous activity (running, aerobics) 20 minutes per day, 3 or more days per week; or, for employed people, spending most of the time on the job walking or doing heavy labor or physically demanding work. “Insufficient fruit and vegetable consumption” was defined as not eating fruits and vegetables 5 or more times per day.

Reported health care coverage included insurance, membership in a health maintenance organization (HMO), or a government plan such as Medicare. Respondents also indicated whether they had one or more people they thought of as their “personal doctor” (this could also have been an alternative health care provider).

Reception of a flu shot within the past year was determined for all respondents. Lifetime history of colonoscopy or sigmoidoscopy screening was determined for respondents aged 50 years or older. Pap test screening within the past 3 years was determined for women aged 18 years and older. Mammogram screening within the past 2 years was determined for women aged 40 years and older. Prostate-specific antigen test screening within the past 2 years was determined for men aged 40 years and older.

Data Analysis

Analyses were completed with Stata version 9.0 (StataCorp LP, College Station, Texas), with P < .05 considered statistically significant. We used the Pearson χ² test of independence to determine whether age, race, and education distributions varied by sexual orientation. We used multiple logistic regression models to test for associations between sexual orientation and health indicators, stratified by gender and adjusted for age and education. Because distributions were not substantially different and the number of minority respondents was small, we did not adjust models for race/ethnicity.

RESULTS

A total of 30 845 male and 48 655 female respondents were available from the 2003–2006 Washington BRFSS. Of all respondents in our data set, 498 men identified themselves as gay and 235 as bisexual; 589 women identified themselves as lesbian and 561 as bisexual.

Among both men and women, very few (0.2%) responded that their sexual orientation was “something else.” Visual inspection of additional information provided by respondents to describe “something else” indicated that some were transgendered and that others provided answers such as “human” or “asexual.” About 1% (0.8% of men and 1.2% of women) responded “don't know,” and 1.2% of men and 1.6% of women refused to give an answer about their sexual orientation. These respondents were not included in further analysis.

Demographics

Table 1 describes the demographic composition of the respondents, stratified by sexual orientation within gender. Among men, 1.9% reported being gay and 0.9% reported being bisexual. Among women, 1.4% reported being lesbian and 1.6% reported being bisexual.

TABLE 1.

Characteristics of Study Population, by Gender and Sexual Orientation: Washington Behavioral Risk Factor Surveillance System, 2003–2006

	Men				Women
	Heterosexual (n = 30 112), %	Gay (n = 498), %	Bisexual (n = 235), %	P	Heterosexual (n = 47 505), %	Lesbian (n = 589), %	Bisexual (n = 561), %	P
Age, y				< .001				< .001
18–29	22.1	16.5	38.8		20.2	24.5	48.0
30–39	19.9	30.0	23.4		18.7	22.4	26.6
40–49	20.8	31.9	18.1		20.6	31.8	13.3
50–59	17.8	13.7	11.1		17.7	15.3	8.8
60–69	10.4	5.6	5.1		10.5	4.2	1.7
≥70	8.9	2.3	3.5		12.4	1.9	1.7
Race/Ethnicitya				.57				.03
White	83.7	88.0	82.6		85.6	85.5	82.5
Black	2.2	1.8	2.5		1.8	1.6	3.9
Asian/Pacific Islander	4.0	3.8	1.8		3.6	3.1	2.4
Native American	1.4	1.3	2.8		1.3	1.6	3.1
Other race	0.4	0.3	1.3		0.3	0.3	0.5
Multiracial	0.2	0.1	0.0		0.3	0.7	0.8
Hispanic	8.1	4.7	9.0		7.1	7.2	6.8
Education				< .001				< .001
Less than high school	8.6	4.1	17.0		7.8	5.7	12.5
High school graduate/GED	24.5	16.9	19.3		24.2	11.7	24.9
Some college	29.2	25.9	36.4		34.1	34.1	30.5
College graduate	37.8	53.2	27.4		33.9	48.6	32.1
Poverty levelb				.05				< .001
>200% of the federal poverty level	75.4	79.9	68.0		73.1	66.7	54.7
≤200% of the federal poverty level	24.6	20.1	32.0		26.9	33.3	45.3
Total	97.2	1.9	0.9		97.0	1.4	1.6

Note. GED = general equivalency diploma. Sample sizes are unweighted; percentages are weighted to adjust for sampling and poststratified to the age and gender distribution of the state.

^aBased on “preferred race” for multiracial people; White, Black, other, and multiracial categories exclude Hispanic ethnicity.

^bAs defined by federal guidelines for each survey year.

Lesbian and gay respondents were significantly younger than were their heterosexual counterparts (χ² test; P < .001), and bisexuals were the youngest in each gender group. Among men, the mean age of respondents was 44.4 years for heterosexuals, 41.3 for gays, and 37.1 for bisexuals. Among women, the mean age of respondents was 46.3 for heterosexuals, 40.0 for lesbians, and 32.9 for bisexuals.

The racial/ethnic composition of survey respondents reflected the distribution of Washington, which is more than 80% non-Hispanic White. Race/ethnicity was not associated with sexual orientation among men. Bisexual women were somewhat less likely than were other women to be non-Hispanic White, and the overall association of race/ethnicity with sexual orientation was significant for women (χ² test; P = .03).

Educational attainment was significantly different by sexual orientation (χ² test; P < .001 for both genders). For both men and women, bisexuals appeared to be the least educated (although they were also the youngest group), whereas gays and lesbians were most likely to be college graduates.

The association between poverty level and sexual orientation was only marginal among men but significant among women. Of all groups, bisexual women were most likely (45.3%) to be living at or below 200% of the federally defined poverty level and gay men were least likely (20.1%).

Chronic Disease and Negative Health Outcomes

We examined measures of chronic health conditions and overall health (Table 2). Among men, the prevalence of poor physical health was similar for heterosexual, gay, and bisexual men, but after adjustment for demographic differences, the odds for poor physical health were significantly greater for gay men than they were for heterosexual men. The adjusted odds of having poor mental health and limited activities because of health problems were significantly greater for both gays and bisexuals than they were for heterosexual men. Male odds of having asthma, diabetes, hypertension, or high cholesterol did not differ significantly by sexual orientation.

TABLE 2.

Prevalence of Chronic Disease and Negative Health Outcomes and Their Associations With Sexual Orientation, by Gender: Washington Behavioral Risk Factor Surveillance System, 2003–2006

	Men				Women
Indicator and Secxual Orientation	No.	Prevalence, % (95% CI)	AORa (95% CI)	P	No.	Prevalence,% (95% CI)	AORa (95% CI)	P
Fair/poor physical health
Heterosexual (Ref)	30 013	12.4 (12.0, 12.9)	1.00		47 375	13.4 (13.0, 13.8)	1.00
Gay or lesbian	497	12.4 (9.4, 16.1)	1.5 (1.1, 2.0)	.02	587	13.3 (10.1, 17.3)	1.5 (1.1, 2.1)	.01
Bisexual	235	12.8 (7.4, 21.2)	1.1 (0.6, 2.2)	.68	559	19.8 (16.0, 24.4)	2.3 (1.7, 3.1)	< .001
Poor mental health (≥ 10 d in past mo)
Heterosexual (Ref)	29 658	10.2 (9.7, 10.6)	1.00		46 687	14.4 (14.0, 14.9)	1.00
Gay or lesbian	492	19.7 (15.5, 24.6)	2.3 (1.7, 3.2)	< .001	583	22.1 (17.9, 26.9)	1.7 (1.3, 2.2)	< .001
Bisexual	232	21.0 (14.4, 29.7)	2.1 (1.3, 3.3)	.002	553	38.3 (33.2, 43.8)	3.1 (2.4, 3.9)	< .001
Limited activities due to health problems
Heterosexual (Ref)	29 953	21.0 (20.4, 21.6)	1.00		47 268	23.8 (23.3, 24.3)	1.00
Gay or lesbian	498	24.0 (19.7, 28.8)	1.4 (1.1, 1.9)	.004	585	32.2 (27.2, 37.7)	2.0 (1.6, 2.6)	< .001
Bisexual	235	29.9 (22.3, 38.8)	2.1 (1.3, 3.1)	.001	559	35.1 (30.2, 40.4)	2.6 (2.1, 3.3)	< .001
Diabetes
Heterosexual (Ref)	29 961	6.8 (6.5, 7.2)	1.00		47 458	6.3 (6.0, 6.5)	1.00
Gay or lesbian	496	5.4 (3.5, 8.3)	1.1 (0.7, 1.8)	.63	589	5.1 (3.3, 7.7)	1.3 (0.8, 2.0)	.27
Bisexual	233	7.9 (4.2, 14.2)	1.8 (0.9, 3.6)	.10	561	5.8 (3.8, 8.8)	1.8 (1.1, 2.8)	.02
Asthma
Heterosexual (Ref)	29 915	6.7 (6.3, 7.1)	1.00		47 131	11.2 (10.9, 11.6)	1.00
Gay or lesbian	496	8.2 (5.5, 11.9)	1.3 (0.8, 1.9)	.29	585	17.7 (13.7, 22.5)	1.7 (1.3, 2.3)	.001
Bisexual	234	9.0 (5.1, 15.2)	1.3 (0.7, 2.4)	.36	554	21.0 (17.0, 25.7)	2.0 (1.5, 2.6)	< .001
Hypertensionb
Heterosexual (Ref)	14 930	25.1 (24.3, 26.0)	1.00		22 979	22.7 (22.1, 23.4)	1.00
Gay or lesbian	261	18.2 (13.2, 24.7)	0.8 (0.5, 1.2)	.29	285	14.7 (9.8, 21.4)	1.0 (0.6, 1.7)	.996
Bisexual	120	21.7 (13.8, 32.4)	1.1 (0.6, 2.0)	.72	265	17.0 (12.2, 23.1)	1.6 (1.1, 2.5)	.02
High cholesterolb
Heterosexual (Ref)	11 775	36.9 (35.8, 38.0)	1.00		18 782	33.5 (32.6, 34.3)	1.00
Gay or lesbian	220	32.4 (25.2, 40.6)	1.0 (0.7, 1.5)	.91	239	29.2 (22.6, 36.8)	1.1 (0.8, 1.6)	.54
Bisexual	83	35.2 (22.6, 50.3)	1.1 (0.6, 2.1)	.69	183	24.8 (17.9, 33.2)	1.0 (0.7, 1.6)	.94

Note. AOR = adjusted odds ratio; CI = confidence interval.

^aFrom a logistic model that includes sexual orientation, age, and education (less than high school, high school/general equivalency diploma, some college, college graduate).

^bQuestions were asked only in 2003 and 2005.

Among women, the prevalence of poor physical health was nearly equal for heterosexual and lesbian women. However, after adjustment for demographic differences, lesbians and bisexuals had significantly greater odds than did heterosexual women of having poor general health, poor mental health, and limited activities because of health problems. Both lesbian and bisexual women were more likely than were heterosexual women to have asthma. Bisexual women had about double the odds for diabetes and also increased odds for hypertension compared with heterosexual women.

Chronic Disease Risk Behaviors

We examined measures of individual behaviors that are associated with development of chronic diseases (Table 3). Gay men were significantly less likely than heterosexual men to be overweight but had twice the odds for smoking. Gay men and heterosexual men had similar odds for heavy drinking, insufficient physical activity, and insufficient fruit and vegetable consumption. Compared with heterosexual men, bisexual men had twice the odds for heavy drinking, smoking, and insufficient exercise; odds for overweight and for insufficient fruit and vegetable consumption were similar.

TABLE 3.

Prevalence of Chronic Disease Risk Behaviors and Their Associations With Sexual Orientation, by Gender: Washington Behavioral Risk Factor Surveillance System, 2003–2006

	Men				Women
Indicator and Sexual Orientation	No.	Prevalence,% (95% CI)	AORa (95% CI)	P	No.	Prevalence,% (95% CI)	AORa (95% CI)	P
Overweight (BMI ≥ 25 kg/m2)
Heterosexual (Ref)	29 708	23.5 (22.9, 24.2)	1.00		44 297	22.3 (21.8, 22.8)	1.00
Gay or lesbian	494	14.2 (10.9, 18.2)	0.6 (0.4, 0.8)	< .001	572	28.8 (24.3, 33.7)	1.6 (1.3, 2.0)	< .001
Bisexual	229	20.6 (14.3, 28.7)	0.9 (0.6, 1.4)	.5	529	29.7 (24.9, 35.0)	1.6 (1.2, 2.0)	.001
Heavy drinkingb
Heterosexual (Ref)	29 543	5.0 (4.7, 5.4)	1.00		46 937	5.1 (4.8, 5.3)	1.00
Gay or lesbian	490	5.8 (3.5, 9.5)	1.2 (0.7, 2.1)	.44	586	9.2 (6.3, 13.1)	1.8 (1.2, 2.7)	.004
Bisexual	233	10.8 (5.6, 19.6)	2.0 (1.0, 4.1)	.05	550	13.5 (9.9, 18.2)	2.9 (2.0, 4.1)	< .001
Current smoking
Heterosexual (Ref)	29 978	19.2 (18.5, 19.8)	1.00		47 250	16.6 (16.1, 17.0)	1.00
Gay or lesbian	495	29.5 (24.4, 35.2)	2.2 (1.6, 2.9)	< .001	583	29.8 (25.0, 35.1)	2.4 (1.8, 3.2)	< .001
Bisexual	234	38.7 (29.5, 48.8)	2.3 (1.6, 3.3)	< .001	555	38.1 (32.9, 43.6)	2.5 (2.0, 3.2)	< .001
Insufficient physical activity (work and leisure combined)c
Heterosexual (Ref)	14 175	32.7 (31.7, 33.8)	1.00		21 691	39.6 (38.8, 40.5)	1.00
Gay or lesbian	255	32.5 (25.5, 40.4)	1.0 (0.7, 1.5)	.80	274	38.1 (30.8, 46.0)	1.1 (0.8, 1.5)	.53
Bisexual	114	43.9 (31.8, 56.8)	1.9 (1.1, 3.4)	.02	259	34.8 (27.9, 42.5)	1.0 (0.7, 1.4)	.93
Insufficient fruits and vegetablescd
Heterosexual (Ref)	14 957	81.4 (80.6, 82.2)	1.00		23 026	70.4 (69.6, 71.2)	1.00
Gay or lesbian	261	74.9 (67.5, 81.1)	0.7 (0.5, 1.0)	.07	285	70.6 (62.7, 77.5)	1.1 (0.7, 1.5)	.73
Bisexual	120	77.0 (66.1, 85.2)	0.7 (0.4, 1.2)	.16	265	76.7 (69.8, 82.3)	1.2 (0.8, 1.7)	.34

Note. AOR = adjusted odds ratio; CI = confidence interval; BMI = body mass index. BMI was calculated as weight in kilograms divided by height in meters squared.

^aFrom a logistic model that includes sexual orientation, age, and education (less than high school, high school/general equivalency diploma, some college, college graduate).

^bDefined as 2 or more drinks per day for men and 1 or more per day for women.

^cQuestions were asked only in 2003 and 2005.

^dDefined as less than 5 times per day.

Both lesbian and bisexual women had higher odds for overweight, heavy drinking, and smoking than did heterosexual women. After adjustment for demographic differences, lesbian, bisexual, and heterosexual women had similar odds for insufficient physical activity and for insufficient fruit and vegetable consumption.

Protective Health Care Services

We examined the prevalence of protective health care services, including access to care, preventive services, and screenings (Table 4). For gay men, the odds of having protective health care factors were either similar to those of heterosexual men (for health insurance, prostate-specific antigen test, and colonoscopy) or significantly better (for having a personal doctor and receiving flu shots). By contrast, for bisexual men, the odds of having protective health care factors were either similar to those of heterosexual men (for having a personal doctor, flu shot, and colonoscopy) or significantly worse (for having health insurance and prostate-specific antigen test).

TABLE 4.

Prevalence of Protective Health Care Factors and Their Associations With Sexual Orientation, by Gender: Washington Behavioral Risk Factor Surveillance System, 2003–2006

	Men				Women
Indicator and Sexual Orientation	No.	Prevalence, % (95% CI)	AORa (95% CI)	P	No.	Prevalence, % (95% CI)	AORa (95% CI)	P
Current health insurance
Heterosexual (Ref)	30 021	83.9 (83.2, 84.5)	1.00		47 431	88.0 (87.5, 88.4)	1.00
Gay or lesbian	496	87.5 (83.3, 90.7)	1.1 (0.8, 1.7)	.49	586	81.3 (76.7, 85.3)	0.6 (0.4, 0.8)	.002
Bisexual	234	67.6 (57.1, 76.6)	0.5 (0.3, 0.8)	.004	561	74.5 (69.3, 79.2)	0.7 (0.5, 0.9)	.006
Have a personal doctor
Heterosexual (Ref)	29 969	72.5 (71.8, 73.3)	1.00		47 389	84.7 (84.3, 85.2)	1.00
Gay or lesbian	496	82.5 (77.4, 86.7)	1.8 (1.3, 2.5)	.001	587	79.1 (74.1, 83.4)	0.8 (0.5, 1.0)	.07
Bisexual	234	62.1 (52.2, 71.1)	0.9 (0.6, 1.4)	.58	558	75.5 (70.5, 79.9)	0.9 (0.7, 1.2)	.50
Received a flu shot in past y
Heterosexual (Ref)	30 023	30.1 (29.4, 30.8)	1.00		47 391	33.6 (33.0, 34.1)	1.00
Gay or lesbian	495	44.1 (38.4, 50.0)	2.2 (1.7, 2.9)	< .001	587	26.5 (22.0, 31.5)	1.0 (0.7, 1.2)	.72
Bisexual	235	19.0 (13.9, 25.5)	0.8 (0.5, 1.1)	.14	559	18.0 (14.5, 22.2)	0.8 (0.6, 1.1)	.18
Pap test in past 3 y (women aged ≥ 18 y)b
Heterosexual (Ref)					17 329	85.3 (84.5, 86.0)	1.00
Gay or lesbian					252	76.3 (66.7, 83.8)	0.5 (0.3, 0.7)	.001
Bisexual					254	84.2 (77.1, 89.4)	0.9 (0.6, 1.4)	.60
Mammogram in past 2 y (women aged ≥ 40 y)b
Heterosexual (Ref)					17 996	74.9 (74.0, 75.8)	1.00
Gay or lesbian					207	63.1 (54.1, 71.3)	0.6 (0.4, 0.9)	.01
Bisexual					105	52.6 (39.7, 65.1)	0.4 (0.2, 0.7)	< .001
PSA test in past 2 y (men aged ≥ 40 y)b
Heterosexual (Ref)	10 503	46.8 (45.6, 48.1)	1.00
Gay or lesbian	150	34.6 (25.8, 44.6)	0.8 (0.5, 1.3)	.41
Bisexual	66	20.2 (11.1, 33.9)	0.3 (0.2, 0.6)	.002
Ever had colonoscopy/sigmoidoscopy (aged ≥ 50 y)b
Heterosexual (Ref)	8384	61.8 (60.4, 63.2)	1.00		13 520	60.3 (59.2, 61.5)	1.00
Gay or lesbian	87	56.4 (43.0, 68.9)	0.9 (0.5, 1.6)	.77	105	56.6 (44.1, 68.4)	1.0 (0.6, 1.7)	.91
Bisexual	44	66.2 (48.4, 80.3)	1.2 (0.6, 2.7)	.61	56	66.5 (50.3, 79.6)	1.5 (0.8, 3.0)	.24

Note. CI = confidence interval; OR = odds ratio; PSA = prostate-specific antigen.

^aFrom a logistic model that includes sexual orientation, age, and education (less than high school, high school/general equivalency diploma, some college, college graduate).

^bQuestion asked in 2004 and 2006 only.

Lesbian and bisexual women had significantly worse odds than did heterosexual women for having health care insurance and receiving mammograms. Lesbians also had worse odds than did heterosexual women for having a routine Pap test. Odds were similar for having a personal doctor, receiving a flu shot, and colonoscopy.

DISCUSSION

We found that sexual minorities were at risk for multiple health outcomes and risk behaviors and had fewer protective health care services compared with their heterosexual counterparts. Although many of these disparities had been previously reported in research, Washington's surveillance system modification illustrates the real-world feasibility of collecting comprehensive, population-based LGB health data, with a heterosexual comparison group, and also the capacity to measure progress toward improving LGB community health.

Both male and female LGB respondents in our study were younger and had different education patterns than did heterosexual respondents. Adjustment for age and education was an important part of identifying disparities by sexual orientation; crude prevalence did not always yield an accurate picture of differences because so many health factors were also associated with age and education or socioeconomic status.²³ We do not have information about why differences in age and education exist between LGB and heterosexual respondents. Potentially, older people may feel less comfortable reporting LGB sexual orientation.

Lesbian and bisexual women had more health disparities relative to heterosexual women than did gay and bisexual men relative to heterosexual men. Of the 7 health outcomes we explored, only 1 outcome did not show elevated risk for either lesbian or bisexual women. Lesbian and bisexual women in our study were at greater risk for overweight, heavy drinking, and smoking, which are all leading contributors to early mortality.²⁴

We were able to disaggregate bisexual and lesbian or gay individuals when describing health risks. Among both men and women, bisexuals had the greatest number of increased risks compared with their heterosexual and lesbian or gay counterparts. For example, bisexual women, but not lesbians, had significantly increased odds for diabetes and hypertension. We did not find other studies reporting similar results; however, most previous studies have not been able to include separate estimates for lesbian and bisexual women. We do not have satisfactory hypotheses about the cause of this greater disparity among bisexuals. We note that bisexual orientation may be fixed or it may be a self-perception that allows for a shift between heterosexuality and homosexuality.¹⁷ Potentially, stress associated with changing self-perception could result in other negative health outcomes. Further study is warranted.

We noted that smoking was a key risk factor for LGB groups. This was 1 indicator where both sexual minority men and women had higher odds for risk than heterosexuals, and the magnitude of the odds ratio was large compared with other health risks. Our findings confirm that tobacco control should be a high-priority issue for people working to improve the health of LGB communities.

Although some variables did not indicate disparities between sexual minorities and heterosexuals, this does not mean that those specific issues should not still be areas of concern. For example, at least one third of respondents of any orientation are not getting enough exercise.

Our findings add to the evidence that LGB groups have important health disparities; however, the mechanism for this disparity is not well understood. Some have suggested that increased health risk behaviors (smoking, alcohol) among LGB groups are a response to the pressures of homophobia and discrimination.⁴ In a recent article, Cochran and Mays suggested that excess prevalence of some chronic health conditions among LGB people can be explained by psychological distress and HIV-positive status.³ Lack of preventive care may be the result of either lack of access or unwillingness to use insensitive care. Improved information about LGB health, by standard inclusion of sexual orientation in public health surveillance systems, is a foundation for addressing these health disparities and understanding their mechanisms and whether the patterns vary regionally.

Our experience also provides evidence that information about sexual orientation and health can be efficiently gathered as part of established population-based health surveys. The cost of adding the sexual identity question to the state BRFSS each year (approximately $1000) is a fraction of the cost of a separate survey. Although public health agencies may be concerned that asking about sexual orientation is too controversial, the very small percentage of respondents who answered “don't know” or who refused to respond suggests otherwise. The majority (97%) of respondents gave clear information about their sexual orientation. In contrast, for a question about household income, only 88% of respondents provided usable information. In addition, the survey interviewers and managers received very few complaints about this question. A small number of other states have also added sexual orientation to BRFSS surveys without experiencing adverse events.²⁵ We examined measures of health that were available from multiple years of the BRFSS, that were asked of the general population, and that were of significant public health interest. There are many other indicators routinely collected in public health surveillance systems, and standard inclusion of sexual orientation as a demographic variable would allow for a combination of data from states or years. This would provide the ability to generate robust estimates for LGB persons, including reliable estimates of health variables for subgroups of LGB persons.

Limitations

Our question on sexual orientation was based on self-identification, which may be a less sensitive measure than sexual behaviors or attraction.²⁶ We chose this single measure to use the least amount of limited time in the BRFSS questionnaire. Some people might have been unwilling to reveal their LGB orientation as part of the survey, resulting in misclassification of sexual orientation; however, this should result in a bias toward the null. We do not have information about whether LGB individuals would be more or less likely than heterosexuals to participate in, or to be sampled for, the survey (for example, whether LGB individuals were more or less likely than heterosexuals to live in a household with a landline telephone).

Determination of LGB status based on self-identification is satisfactory to the extent that public health interventions are usually tailored to those who openly self-identify as LGB and are therefore more likely to read LGB-specific periodicals, attend LGB events or venues, and respond to marketing campaigns addressing LGB individuals. However, further studies to confirm differences in chronic disease risks based on both self-identification and attraction or behaviors would be useful.

The health of transgendered individuals is sometimes considered together with that of LGB individuals. The responses to our question on sexual orientation did not offer a transgender option, although a few people described themselves as transgendered in the response option of “something else.” We did not have enough transgender respondents in this survey to describe these individuals. Being “transgendered”—that is, feeling misclassified as either male or female by birth gender—is more consistent with the concept of gender identity (how individuals perceive their own gender) than with sexual orientation (whether one is attracted to same-gender individuals or not). Since the prevalence of transgendered individuals is expected to be extremely small, describing and monitoring the health status of transgendered populations would require either an extremely large sample or a targeted sampling design.

We note that the total number of individuals dropped from our analysis because they answered “something else” or “don't know” or refused to answer the question on sexual orientation exceeded the number who reported being LGB (2230 vs 1883). The “don't know” and “refused” groups were older than both the heterosexual and LGB groups. We do not know whether these were individuals who did not understand the sexual orientation question, were uncomfortable answering (because of being a sexual minority or for some other reason), legitimately felt that they were “something else” (such as transgendered), or were honestly uncertain about their sexuality. Potentially, the addition of some proportion of these individuals to the LGB group could substantially change the nature of our observed disparities. Further investigation of demographic and other characteristics of the nonresponding group may be useful.

In our logistic models, we used only education to control for socioeconomic status (rather than poverty level) because there was a relatively large proportion of missing data for household income (approximately 12%), which would have reduced our available sample size. However, we explored models using the binary poverty level measure, and generally the magnitude of disparities was similar, although the number of significant findings was somewhat decreased.

The BRFSS has a number of inherent limitations: the exclusion of individuals who live in homes without telephones or who have only cellular telephone service, live in institutions (such as college dorms and assisted living facilities), or do not speak English or Spanish. In addition, the response rates were not optimal. Generalization of findings from our study to the true LGB population should be considered carefully, bearing in mind the overall limitations of the BRFSS.

Conclusions

We found further evidence of numerous health-related disparities—including disparities in health outcomes, risk behaviors, and protective health care services—for LGB men and women compared with heterosexual men and women. Data collection did not incur substantial economic or political costs. Inclusion of sexual orientation as a standard demographic variable in public health surveillance systems is an effective and efficient means of collecting data that are essential for describing the prevalence of health measures, identifying disparities, developing interventions, and measuring progress toward improving community health.