Discrimination, Sexual Orientation Discrimination, and Severity of Tobacco Use Disorder in the United States: Results From the National Epidemiologic Survey on Alcohol and Related Conditions-III

Abstract

Introduction

Tobacco use is more prevalent among sexual minority populations relative to heterosexual populations. Discrimination is a known risk factor for tobacco use. However, the relationship between exposure to different forms of discrimination, such as racial or ethnic discrimination and sexual orientation discrimination, and tobacco use disorder (TUD) severity has not been examined.

Aims and Methods

Using data from the 2012–2013 National Epidemiologic Survey of Alcohol and Related Conditions-III (n = 36 309 US adults), we conducted multivariable logistic regression analyses to examine the associations among racial or ethnic discrimination, sexual orientation discrimination, and TUD severity for lesbian or gay-, bisexual-, and heterosexual-identified adults. Consistent with the Diagnostic and Statistical Manual of Mental Disorders Fifth Edition (DSM-5), past-year moderate-to-severe TUD was defined as the presence of ≥4 DSM-5 TUD symptoms.

Results

Higher levels of lifetime racial or ethnic discrimination were associated with significantly greater odds of past-year moderate-to-severe TUD among sexual minorities (adjusted odds ratio [AOR] = 1.03, 95% confidence interval [CI] = 1.01–1.05) and heterosexuals (AOR = 1.04, 95% CI = 1.03–1.05). Stressful life events, mood disorder, and anxiety disorder had significant associations with moderate-to-severe TUD among sexual minorities (AOR range: 1.86–5.22, p < .005) and heterosexuals (AOR range: 1.71–3.53, p < .005). Among sexual minorities, higher levels of racial or ethnic and/or sexual orientation discrimination were associated with greater odds of any TUD (AOR = 1.02, 95% CI = 1.01–1.03).

Conclusions

Sexual minorities and heterosexuals who experience higher levels of racial or ethnic discrimination are at heightened risk of having moderate-to-severe TUD. Exposure to higher levels of discrimination also increases the risk of having any TUD among sexual minority adults. Health providers and tobacco cessation professionals should be cognizant of the minority stressors experienced by their clients and their potential impact on TUD severity.

Implications

This study is the first to show how experiences of racial or ethnic and sexual orientation discrimination are associated with DSM-5 TUD severity among sexual minority and heterosexual populations. Individuals exposed to multiple minority stressors may have increased vulnerability for developing TUD and related adverse health consequences. Our study underscores the importance of considering racial or ethnic discrimination and the multiple minority statuses that individuals may hold. Eliminating all forms of discrimination and developing interventions that are sensitive to the role that discrimination plays in TUD severity may attenuate the tobacco use disparities between sexual minority and heterosexual adults.

Introduction

Approximately one in five adults in the United States currently use tobacco products, including various combustible (eg, cigarettes, cigars, etc.), noncombustible (eg, smokeless tobacco, chewing tobacco, etc.), and electronic (eg, e-cigarettes) products.¹ Cigarette smoking remains the leading preventable cause of disease, disability, and death and can be attributed to an estimated 480 000 deaths in the United States each year.² The prevalence of past-year and lifetime tobacco use disorder (TUD), as defined by the Diagnostic and Statistical Manual of Mental Disorders Fifth Edition (DSM-5),³ were estimated to be 20% and 27.9%, respectively, in the United States during 2012–2013.⁴ A growing body of literature suggests that tobacco use is more prevalent among sexual minority populations (eg, gay, lesbian, and bisexual)^2,5–12 and that these populations have greater odds of TUD^13,14 relative to heterosexual people. Consequently, sexual minorities may have a heightened risk of adverse health consequences related to their elevated prevalence of tobacco use.¹⁵

In addition to the targeted marketing and promotion efforts conducted by the tobacco industry^15–18 and the social norms of greater acceptability around cigarette use in venues that target sexual minority individuals,^11,15 other suggested explanations for the tobacco use disparities include the excess stress sexual minorities experience resulting from prejudice, stigma, and discrimination. Meyer’s Minority Stress Model posits that in addition to everyday life stress, exposure to excess stress in the social environment related to minority status may contribute to health risk behaviors which, in turn, result in adverse health outcomes.^19,20 Stigma and discrimination can occur at the individual or structural level, creating conditions that produce inequitable treatment and limit minority individuals’ opportunities and resources.²¹ Discrimination can be pervasive and occur across multiple domains of life, including public spaces, health care, employment, and educational settings.²² Discrimination can also be more subtle, in the form of microaggressions—words or actions, often unintentional, which convey an implicit negative bias toward marginalized groups.²³

Discrimination is known to be a risk factor for tobacco product use²⁴ and nicotine dependence.²⁵ Studies have shown a disproportionate prevalence of victimization and discrimination experienced by sexual minorities compared with heterosexual people.^12,26 There is much evidence to support the positive associations between sexual orientation discrimination and substance use disorder (SUD),²⁷ TUD,^8,28 or smoking behavior²³ among sexual minority populations. These studies have found that sexual minorities who experienced any or higher levels of discrimination had greater probability of having a SUD, TUD, or engaging in smoking behavior relative to those who did not experience any or lower levels of discrimination. Moreover, experiencing sexual orientation discrimination in combination with other forms of discrimination increased the odds of SUD, demonstrating the potential cumulative effect of stressors related to the intersection of multiple minority identities.²⁷

The Minority Stress Model is also relevant for racial or ethnic minority people, who experience a disproportionate prevalence of discrimination compared with non-Hispanic White people.^24–26 Similarly, evidence suggests that racial or ethnic discrimination is associated with smoking among certain racial or ethnic minorities²⁹ and may be a mechanism which mediates the relationship between race or ethnicity and smoking.²⁶ To our knowledge, studies have focused on a binary TUD outcome rather than examine the association of discrimination with severity of TUD among sexual minority populations. TUD severity has important implications for tobacco cessation treatment; treatment length is often longer and relapse is more common among people with more severe TUD and withdrawal symptoms.^30,31 Additional research is needed to understand the influence of discrimination on TUD severity while considering the context of other intersecting identities, as sexual minorities who belong to other marginalized communities may experience multiple minority stressors, which may influence the severity of their TUD.

Stress reduction and regulation of negative affect are key drivers for tobacco use.^15,32 Among the general population, having two or more stressful life events occurring in the past-year increases the odds of new and ongoing TUD compared with having none or one stressful life event.³³ The higher prevalence of SUD experienced by sexual minorities is mediated by the stressful life events and sexual orientation discrimination they experience.³⁴ Studies also suggest that mental health is a significant correlate for TUD in the general population⁴ and for cigarette use among sexual minorities.^15,35 Notably, there is also heightened prevalence of mood and anxiety disorders among sexual minorities compared with heterosexual people.³⁶ Moreover, the experience of discrimination increases the odds of past-year mental health disorders among sexual minorities.³⁶ Thus, in addition to minority stressors, general life stressors, and mental health are important factors in tobacco use and TUD.

Although the Minority Stress Model has been instrumental in conceptualizing the disparities that exist among minority populations, it also emphasizes the importance of identifying resiliencies and protective factors to cope with the effects of stigma.²⁰ Social support can positively influence health behaviors and affect health outcomes.³⁷ Among young sexual minority women, connection to a sexual minority community and having friends with the same sexual identity have been found to be protective factors in relation to their smoking behavior.³⁸ Greater social support is associated with lower odds of TUD.¹³

Informed by the Minority Stress Model, our study addresses important gaps in the literature on TUD among sexual minority populations by (1) examining the association between racial or ethnic discrimination and TUD severity among sexual minority and heterosexual populations, and (2) assessing the relationship between racial or ethnic and sexual orientation discrimination and TUD severity among sexual minority populations. Based on previous research, we hypothesize the following: (1) racial or ethnic discrimination will be positively associated with TUD severity, and (2) among sexual minorities, higher levels of discrimination based on racial, ethnic, or sexual minority status will be positively associated with TUD severity.

Methods

Study Population

Our data come from the National Epidemiologic Survey of Alcohol and Related Conditions-III (NESARC-III), which was conducted from April 2012 through June 2013 and collected data from US noninstitutionalized adults aged 18 and older. The overall sample size was 36 309, and the overall response rate was 60.1%.³⁹ Respondents were selected from a multistage, stratified cluster sample. A unique feature of the NESARC-III was the inclusion of the Alcohol Use Disorder and Associated Disabilities Interview Schedule-5 (AUDADIS-5), a reliable and valid diagnostic interview that aligns with the diagnostic symptom criteria in the DSM-5.⁴⁰ All procedures, including informed consent, received full human subjects review, and IRB approval. Detailed information about the methodology and sampling of the NESARC-III are described elsewhere.³⁹

To ensure an adequate sample size for sexual orientation and race or ethnicity categories, we excluded respondents who were “not sure” (n = 199) about their sexual identity as well as those who identified as Asian, Native Hawaiian, other Pacific Islander (n = 1765), and American Indian or Alaska Native (n = 499) because cell sizes were too small (<30) to produce reliable estimates when we restricted the sample to sexual minorities. The final sample size for our study was 35 706 respondents.

Measures

Tobacco Use Disorder

Our dependent variable of interest was the severity of past-year DSM-5 TUD, constructed from items that map onto the 11 symptom criteria of DSM-5 TUD (Supplementary Table 1). Any TUD was defined as the presence of at least two of the symptoms associated with DSM-5 TUD.^3,4Moderate-to-severe TUD was defined as the presence of four or more of the eleven symptoms associated with DSM-5 TUD.

Sexual Orientation

Consistent with prior work,^8,34,41,42 sexual orientation was comprised of three dimensions: sexual identity, sexual attraction, and sexual behavior. Sexual identity was assessed by asking respondents to identify which category best described them: heterosexual (straight), gay or lesbian, bisexual, and not sure. Our sample was restricted to individuals who identified as heterosexual (straight), gay or lesbian, or bisexual. Sexual attraction was assessed by asking respondents to identify which category best described their sexual attraction to other people: only attracted to females, mostly attracted to females, equally attracted to males and females, mostly attracted to males, and only attracted to males. Sexual behavior was assessed by asking respondents with whom they have had sex in their entire life: only females, only males, both males and females, and never had sex. Respondents were coded into one of four categories based on their responses to the identity, attraction, and behavior questions: lesbian or gay-identified; bisexual-identified; heterosexual-identified with discordant attraction or behavior (ie, respondents who identified as heterosexual and reported same-sex attraction or behavior); and heterosexual-identified with concordant attraction and behavior (ie, respondents who identified as heterosexual and reported opposite-sex attraction and behavior). Consistent with previous research,³⁴ the term sexual minorities refers to lesbian or gay-identified, bisexual-identified, and heterosexual-identified with discordant attraction or behavior.

Minority Stress-Related Variables

Variables related to the Minority Stress Model included experiences of discrimination, social support, stressful life events, past-year DSM-5 anxiety disorder, and past-year DSM-5 mood disorder. Adapted from the Experiences of Discrimination (EOD) scale,^43,44 respondents were asked six questions about the frequency of discrimination they experienced based on their race or ethnicity and sexual minority status, respectively: (1) ability to obtain health care or health insurance coverage; (2) health care treatment; (3) in public settings; (4) in other situations such as employment or workplace, educational settings, encounters with law enforcement; (5) verbal harassment; and (6) physical assault or threats of harm. Responses ranged from never (0) to very often (4). The racial or ethnic discrimination items were asked of all respondents. Hispanic or Latino respondents received questions about discrimination based on their Hispanic or Latino identity specifically, whereas non-Hispanic or Latino respondents (including non-Hispanic white respondents) received questions about discrimination based on their race or ethnicity generally. In this study, the racial or ethnic discrimination scales had excellent reliability (α = 0.92 for Hispanic or Latino respondents and α = 0.90 for non-Hispanic or Latino respondents) as did the sexual orientation discrimination scale (α = 0.94). The lifetime racial or ethnic discrimination measure had a possible range from 0 to 48, reflecting the frequency of racial or ethnic discrimination experiences within the past-year and/or prior-to-past-year. The lifetime racial or ethnic and sexual orientation discrimination measure was created by summing the responses to the racial or ethnic discrimination and sexual orientation discrimination items within the past-year and/or prior-to-past-year, with a possible range from 0 to 96. This measure applied to sexual minorities only because heterosexual-identified individuals with concordant attraction and behavior did not receive the items about discrimination based on sexual minority status.

Social support was assessed using a validated measure, the Interpersonal Support Evaluation List (ISEL),⁴⁵ consisting of six items measuring positive statements about a respondent’s social network (eg, “If I were stranded 10 miles from home, someone I know would come and get me”) and six negative statements (eg, “If a family crisis arose, it would be difficult to find someone who could give me good advice about how to handle it”). The negative statements were reverse coded to match substantive meaning. For example, answering “definitely false” to a negative statement would have the same substantive meaning as answering “definitely true” to a positive statement. A higher value of the combined items indicates a stronger self-perceived social network availability. The social support measure had good internal consistency in this study (α = 0.83).

We assessed the respondents’ life stressors in the past-year using the Stressful Life Events scale,⁴³ which measured the presence of one or more of sixteen common life stressors. Examples of stressful life events include death of a family member; homelessness; and separation, divorce, or termination of a steady relationship. The scale had a possible range of 0–16. Consistent with prior work,⁴⁶ this variable was recoded into three categories: none, 1–2 stressful life events, and 3+ stressful life events.

We also included DSM-5 mood and anxiety disorders. The presence of any past-year DSM-5 anxiety disorder was defined as having met criteria for any one of five possible DSM-5 anxiety disorders: specific phobia, social phobia, panic disorder, agoraphobia, or generalized anxiety disorder. The presence of any past-year DSM-5 mood disorder was defined as having met criteria for any one of three possible DSM-5 mood disorders: bipolar, dysthymia, or major depression.

Sociodemographic Variables

Our main variables assessing sociodemographic characteristics were sex (male or female; nonbinary sex was not a response option); age (18–24, 25–44, 45–64, or 65+); race or ethnicity (White, African American, or Hispanic); educational attainment (high school degree or less, attending some college but not receiving a degree, or receiving a college degree or higher); urbanicity (urban or rural); geographic region (Northeast, Midwest, South, and West); relationship status (married or cohabitating, widowed, divorced, separated, or never married); and religiosity or spirituality (very important, somewhat important, not very important, or not important at all).

Missing data rates were found to be less than 5.0% on most individual variables. The one exception was the sexual orientation discrimination variable (12.3% missing), with the majority of missing data on this variable arising due to nonresponse from heterosexual-identified individuals with discordant behavior or attraction. These respondents may have perceived that the sexual orientation discrimination questions were not relevant to them. To determine if the missing values made a difference in our analyses, we conducted multiple multivariate hot deck imputation on respondents who were eligible for the sexual orientation discrimination questions.

Statistical Analysis

We first examined bivariate associations of sociodemographic characteristics and sexual orientation using the -survey tabulate- (svy: tab) commands in Stata to account for the complex sample design features of the NESARC-III, including codes identifying sampling clusters and strata, and survey weights. We estimated the mean amount of lifetime racial or ethnic discrimination and lifetime racial or ethnic and sexual orientation discrimination across the sexual orientation subgroups as well. We conducted a Rao–Scott test of significance for each bivariate association involving categorical measures.⁴⁷ For the comparisons of mean lifetime racial or ethnic discrimination and mean lifetime racial or ethnic and sexual orientation discrimination, we used a design-adjusted Wald-F test.⁴⁸

We then conducted multivariable logistic regression analyses, again accounting for the NESARC-III sample design. We used the subpop() option within Stata’s svy: logit command to distinguish between sexual minorities and heterosexuals with concordant sexual attraction and behavior while computing appropriate estimates and standard errors (SEs). Two binary dependent variables were used in these analyses: any TUD and moderate-to-severe TUD. We first estimated the relationships of lifetime racial or ethnic discrimination with any TUD and with moderate-to-severe TUD among the whole study sample. We subsequently estimated the relationship of lifetime racial or ethnic and sexual orientation discrimination with any TUD and with moderate-to-severe TUD among sexual minorities only. We adjusted for the following covariates in these multivariable analyses: sex, age, race or ethnicity, educational attainment, urbanicity, geographic region, social support, relationship status, religiosity or spirituality, number of life stressors, any past-year DSM-5 anxiety disorders, and any past-year DSM-5 mood disorders. Given the number of models fitted, we evaluated significance at the 0.005 level.⁴⁹ We also performed a design-adjusted goodness-of-fit test for each of our models.⁵⁰

Results

Overall, an estimated 1.6% of the NESARC-III target population identified as lesbian or gay, 1.6% identified as bisexual, 6.7% identified as heterosexual with discordant (ie, same-sex) attraction or behavior, and 90.1% identified as heterosexual with concordant (ie, opposite-sex) attraction and behavior. Table 1 displays the estimated distributions of the measures of interest within sexual orientation subgroups. This table presents sample sizes and estimated column percentages for each variable, with the exception of the social support scale, the lifetime racial or ethnic discrimination scale, and the lifetime racial or ethnic and sexual orientation discrimination scale, where the means and SEs for each scale are provided. Respondents’ reported scores for lifetime racial or ethnic discrimination scale among the overall study sample ranged from 0 to 48. Bisexual-identified individuals had the highest estimated mean score for lifetime racial or ethnic discrimination (3.55, SE = 0.28). Respondents’ reported scores for the lifetime racial or ethnic and sexual orientation discrimination scale among sexual minorities ranged from 0 to 82. Lesbian or gay-identified individuals had the highest estimated mean score for lifetime racial or ethnic and sexual orientation discrimination (8.54, SE = 0.50). Sexual minorities reported higher prevalence of any TUD and moderate-to-severe TUD relative to heterosexual individuals. Sexual orientation subgroups stratified by race or ethnicity further showed that African American bisexual individuals experienced the highest prevalence of any TUD (37.3%) and moderate-to-severe TUD (24.9%) (Figure 1). Results of supplemental analyses on sexual orientation subgroups stratified by sex showed that bisexual women and bisexual men had the highest prevalence of any TUD and of moderate-to-severe TUD relative to their counterparts in other sexual orientation subgroups (Supplementary Figure 1).

Table 1.

Estimated Distributions of Sociodemographic Features and Key Study Measures for Subgroups of the Target NESARC-III Population Defined by Sexual Identity and Attraction or Behavior (N = 35 706)

Variable	Sexual minorities
	Lesbian or gay-identified	Bisexual- identified	Heterosexual-identified, discordant attraction or behavior	Heterosexual-identified, concordant attraction and behavior
Categories	n (%)	n (%)	n (%)	n (%)
Total sample	586 (1.6%)	566 (1.6%)	2396 (6.7%)	32 158 (90.1%)
Sex*
Male	321 (58.2%)	144 (28.8%)	872 (39.4%)	14 295 (49.0%)
Female	265 (41.8%)	422 (71.2%)	1524 (60.6%)	17 863 (51.1%)
Age*
18–24	110 (18.3%)	171 (36.3%)	316 (13.5%)	3822 (12.6%)
25–44	238 (39.0%)	266 (42.6%)	967 (36.2%)	12 098 (34.1%)
45–64	200 (34.9%)	106 (17.9%)	759 (32.8%)	10 967 (35.4%)
≥65	38 (7.8%)	23 (3.2%)	354 (17.5%)	5271 (17.9%)
Race or ethnicity
White	333 (70.4%)	289 (68.9%)	1308 (72.9%)	16 994 (71.5%)
African American	116 (13.6%)	150 (16.6%)	495 (12.6%)	6852 (12.6%)
Hispanic	115 (16.0%)	101 (14.6%)	413 (14.5%)	6286 (16.0%)
Educational attainment*
High school degree or less	160 (28.1%)	230 (40.0%)	922 (36.8%)	13 693 (39.0%)
Some college	145 (21.3%)	175 (30.8%)	582 (23.9%)	6966 (21.4%)
College degree or higher	281 (50.6%)	161 (29.2%)	892 (39.3%)	11 499 (39.7%)
Urbanicity*
Urban	537 (88.6%)	507 (85.5%)	2081 (82.9%)	26 541 (78.1%)
Rural	49 (11.4%)	59 (14.5%)	315 (17.1%)	5617 (21.9%)
Geographic region*
Northeast	110 (25.4%)	96 (21.2%)	355 (17.8%)	4531 (18.2%)
Midwest	91 (15.3%)	127 (22.5%)	475 (20.1%)	6739 (21.6%)
South	207 (31.8%)	201 (32.4%)	849 (33.4%)	13 041 (37.5%)
West	178 (27.6%)	142 (23.9%)	717 (28.7%)	7847 (22.8%)
Social support scale (0–36) [mean, SE]	30.1 (0.3)	28.0 (0.4)	28.8 (0.1)	30.3 (0.1)
Relationship status*
Married or cohabitating	126 (30.1%)	139 (31.2%)	1028 (54.3%)	15 323 (59.2%)
Widowed	7 (1.0%)	14 (2.5%)	141 (5.4%)	2376 (5.9%)
Divorced	40 (6.2%)	64 (11.0%)	370 (12.4%)	4694 (10.9%)
Separated	15 (3.6%)	29 (3.7%)	120 (3.4%)	1392 (2.9%)
Never married	398 (59.1%)	320 (51.6%)	737 (24.5%)	8373 (21.1%)
Religiosity or spirituality*
Very important	240 (35.9%)	244 (37.9%)	1281 (50.5%)	18 744 (54.5%)
Somewhat important	178 (32.6%)	179 (33.5%)	656 (28.9%)	8814 (29.4%)
Not very important	81 (16.1%)	65 (12.1%)	226 (9.6%)	2459 (8.7%)
Not important at all	86 (15.4%)	76 (16.5%)	232 (11.0%)	2097 (7.3%)
Stressful life events*
None	129 (23.1%)	81 (14.0%)	625 (26.6%)	10 368 (33.3%)
1–2	213 (36.5%)	181 (33.1%)	970 (42.1%)	13 941 (44.5%)
3+	236 (40.4%)	298 (52.9%)	779 (31.3%)	7677 (22.2%)
Any past-year DSM-5 anxiety disorder*
No	474 (77.8%)	415 (71.1%)	1966 (81.6%)	28 252 (87.8%)
Yes	112 (22.2%)	151 (28.9%)	430 (18.4%)	3906 (12.2%)
Any past-year DSM-5 mood disorder*
No	458 (78.3%)	394 (68.8%)	1904 (80.2%)	28 170 (88.1%)
Yes	128 (21.7%)	172 (31.2%)	492 (19.8%)	3988 (11.9%)
Discrimination scales
Lifetime racial or ethnic discrimination scale (0–48) [mean, SE]*	2.70 (0.2)	3.55 (0.3)	3.00 (0.1)	2.33 (0.1)
Lifetime sexual orientation discrimination scale (0–48) [mean, SE]	5.83 (0.3)	2.59 (0.3)	0.54 (0.1)	—
Lifetime racial or ethnic and sexual orientation discrimination scale (0–96) [mean, SE]*	8.54 (0.5)	6.15 (0.5)	3.63 (0.2)	—
Any TUD*	147 (24.6%)	187 (33.4%)	510 (21.1%)	5307 (16.8%)
Moderate-to-severe TUD*	91 (15.7%)	111 (19.4%)	306 (12.1%)	3075 (9.8%)

DSM-5 = Diagnostic and Statistical Manual of Mental Disorders Fifth Edition; NESARC-III = National Epidemiologic Survey of Alcohol and Related Conditions-III; SE = standard error; TUD = tobacco use disorder.

Rao–Scott test of significance: *p < .005.

Figure 1.

Estimated prevalence of any TUD and moderate-to-severe TUD among racial or ethnic groups by sexual orientation. TUD = tobacco use disorder.

Table 2 presents the estimated odds ratios in the multivariable logistic regression models analyzing the associations between lifetime racial or ethnic discrimination and any TUD and moderate-to-severe TUD, respectively, among sexual minorities and heterosexual individuals. Results show that greater lifetime racial or ethnic discrimination was positively associated with any TUD among sexual minorities (adjusted odds ratio [AOR] = 1.03, 95% confidence interval [CI] = 1.02–1.05) and heterosexuals (AOR = 1.03, 95% CI = 1.03–1.04), after adjusting for the other covariates (each of which had a significant association with any TUD in the bivariate analyses). Similarly, greater lifetime racial or ethnic discrimination was positively associated with moderate-to-severe TUD among sexual minorities (AOR = 1.03, 95% CI = 1.01–1.05) and heterosexuals (AOR = 1.04, 95% CI = 1.03–1.05). Furthermore, stressful life events, past-year DSM-5 mood disorder, and past-year DSM-5 anxiety disorder were significantly and positively associated with moderate-to-severe TUD among both sexual minority and heterosexual populations. Similar findings emerged in the multiple imputation analysis (Supplementary Table 2).

Table 2.

Estimated Multivariable Logistic Regression Models Describing Associations Between Discrimination, Any Tobacco Use Disorder and Moderate-to-Severe Tobacco Use Disorder Among Sexual Minority and Heterosexual-Identified Adults With Concordant^a Behavior and Attraction

	Any TUD (2+)		Moderate-to-severe TUD (4+)
Variables	Sexual minorities	Heterosexual-identified, concordant behavior and attraction	Sexual minorities	Heterosexual-identified, concordant behavior and attraction
Categories	N = 3210	N = 29 238	N = 3210	N = 29 238
	AOR (95% CI)	AOR (95% CI)	AOR (95% CI)	AOR (95% CI)
Lifetime racial or ethnic discrimination scale	1.03 (1.02, 1.05)*	1.03 (1.03, 1.04)*	1.03 (1.01, 1.05)*	1.04 (1.03, 1.05)*
Sex
Male	Ref	Ref	Ref	Ref
Female	0.93 (0.73, 1.18)	0.67 (0.62, 0.73)*	0.99 (0.75, 1.31)	0.65 (0.58, 0.73)*
Age
18–24	Ref	Ref	Ref	Ref
25–44	1.40 (0.96, 2.03)	1.65 (1.42, 1.91)*	1.42 (0.90, 2.24)	1.65 (1.36, 2.01)*
45–64	1.32 (0.90, 1.93)	1.30 (1.14, 1.49)*	1.33 (0.84, 2.11)	1.39 (1.13, 1.71)*
≥65	0.65 (0.37, 1.17)	0.52 (0.44, 0.63)*	0.67 (0.34, 1.31)	0.59 (0.46, 0.76)*
Race or ethnicity
White	Ref	Ref	Ref	Ref
African American	0.79 (0.58, 1.07)	0.53 (0.46, 0.61)*	0.78 (0.52, 1.15)	0.50 (0.43, 0.58)*
Hispanic	0.59 (0.41, 0.84)*	0.33 (0.28, 0.38)*	0.33 (0.23, 0.48)*	0.32 (0.27, 0.40)*
Educational attainment
High school degree or less	Ref	Ref	Ref	Ref
Some college	0.60 (0.46, 0.79)*	0.68 (0.62, 0.75)*	0.70 (0.51, 0.97)	0.72 (0.64, 0.82)*
College degree or higher	0.28 (0.21, 0.37)*	0.35 (0.31, 0.39)*	0.33 (0.23, 0.48)*	0.40 (0.35, 0.46)*
Urbanicity
Urban	Ref	Ref	Ref	Ref
Rural	0.87 (0.63, 1.20)	1.29 (1.15, 1.45)*	0.82 (0.53, 1.28)	1.29 (1.14, 1.45)*
Geographic region
Northeast	Ref	Ref	Ref	Ref
Midwest	1.61 (1.11, 2.32)	1.19 (1.03, 1.38)	1.22 (0.78, 1.93)	1.19 (0.97, 1.45)
South	1.19 (0.85, 1.66)	1.19 (1.02, 1.40)	1.05 (0.70, 1.57)	1.12 (0.93, 1.34)
West	1.11 (0.78, 1.58)	0.90 (0.77, 1.05)	1.02 (0.68, 1.54)	0.99 (0.82, 1.21)
Social support scale	1.00 (0.98, 1.02)	0.99 (0.98, 1.00)	0.99 (0.97, 1.01)	0.98 (0.97, 0.99)*
Relationship status
Married or cohabitating	Ref	Ref	Ref	Ref
Widowed	0.92 (0.41, 2.08)	1.21 (0.99, 1.47)	0.39 (0.12, 1.27)	1.21 (0.94, 1.56)
Divorced	1.25 (0.88, 1.78)	1.74 (1.56, 1.92)*	1.37 (0.88, 2.15)	1.61 (1.43, 1.80)*
Separated	2.14 (1.29, 3.53)*	1.46 (1.23, 1.73)*	1.90 (1.06, 3.39)	1.36 (1.09, 1.70)
Never married	1.34 (1.03, 1.74)	1.19 (1.06, 1.33)*	1.34 (0.95, 1.89)	1.28 (1.10, 1.49)*
Religiosity or spirituality
Very important	Ref	Ref	Ref	Ref
Somewhat important	1.21 (0.94, 1.55)	1.31 (1.18, 1.47)*	1.22 (0.90, 1.66)	1.18 (1.05, 1.32)
Not very important	1.58 (1.07, 2.35)	1.43 (1.23, 1.67)*	1.26 (0.76, 2.09)	1.24 (1.06, 1.47)
Not important at all	1.32 (0.85, 2.04)	1.35 (1.12, 1.62)*	0.87 (0.53, 1.42)	1.29 (1.04, 1.59)
Stressful life events
None	Ref	Ref	Ref	Ref
1–2	1.66 (1.14, 2.43)	1.55 (1.38, 1.74)*	2.07 (1.28, 3.33)*	1.75 (1.48, 2.07)*
3+	3.20 (2.15, 4.76)*	3.05 (2.67, 3.48)*	5.22 (3.42, 7.98)*	3.53 (2.92, 4.26)*
Any past-year DSM-5 anxiety disorder
No	Ref	Ref	Ref	Ref
Yes	1.44 (1.06, 1.97)	1.50 (1.34, 1.69)*	1.86 (1.35, 2.55)*	1.71 (1.46, 2.00)*
Any past-year DSM-5 mood disorder
No	Ref	Ref	Ref	Ref
Yes	1.85 (1.43, 2.38)*	1.58 (1.40, 1.78)*	2.19 (1.62, 2.94)*	1.95 (1.67, 2.26)*
Design-adjusted goodness-of-fit test	p = .8669	p = .6149	p = .6467	p = .7766

AOR = adjusted odds ratio; CI = confidence interval; DSM-5 = Diagnostic and Statistical Manual of Mental Disorders Fifth Edition; TUD = tobacco use disorder.

Tests of significance: *p < .005.

^aRespondents who identified as heterosexual and reported opposite-sex attraction and behavior.

Table 3 presents the estimated odds ratios from multivariable logistic regression models analyzing the associations between lifetime racial or ethnic and sexual orientation discrimination and any TUD and moderate-to-severe TUD, respectively, among sexual minorities only. After adjusting for the covariates, greater lifetime racial or ethnic and sexual orientation discrimination was significantly and positively associated with any TUD among sexual minorities (AOR = 1.02, 95% CI = 1.01–1.03), but not significantly associated with moderate-to-severe TUD among sexual minorities. In the multiple imputation analysis, greater lifetime racial or ethnic and sexual orientation discrimination was not significantly associated with any TUD among sexual minorities due to a larger SE (Supplementary Table 3). Lastly, we tested interactions between sexual orientation and each of the discrimination scales for each TUD outcome, and no significant results were found.

Table 3.

Estimated Multivariable Logistic Regression Models Describing Associations Between Sexual Orientation and Discrimination, Any Tobacco Use Disorder, and Moderate-to-Severe Tobacco Use Disorder Among Sexual Minority Adults^a

	Model 1		Model 2
	Any TUD		Moderate-to-severe TUD
Variables	N = 2836		N = 2836
Categories	AOR	95% CI	AOR	95% CI
Lifetime racial or ethnic and sexual orientation discrimination scale	1.02	(1.01, 1.03)*	1.01	(1.00, 1.03)
Sex
Male	Ref	…	Ref	…
Female	0.86	(0.66, 1.11)	0.92	(0.70, 1.21)
Age
18–24	Ref	…	Ref	…
25–44	1.31	(0.90, 1.90)	1.40	(0.89, 2.18)
45–64	1.15	(0.78, 1.70)	1.19	(0.76, 1.88)
≥65	0.58	(0.32, 1.05)	0.61	(0.30, 1.23)
Race or ethnicity
White	Ref	…	Ref	…
African American	0.78	(0.56, 1.09)	0.75	(0.51, 1.10)
Hispanic	0.58	(0.40, 0.83)*	0.46	(0.31, 0.68)*
Educational attainment
High school degree or less	Ref	…	Ref	…
Some college	0.59	(0.45, 0.77)*	0.73	(0.53, 1.00)
College degree or higher	0.27	(0.21, 0.37)*	0.32	(0.22, 0.48)*
Urbanicity
Urban	Ref	…	Ref	…
Rural	0.81	(0.58, 1.12)	0.79	(0.52, 1.22)
Geographic region
Northeast	Ref	…	Ref	…
Midwest	1.88	(1.27, 2.77)*	1.44	(0.92, 2.27)
South	1.24	(0.87, 1.76)	1.10	(0.72, 1.67)
West	1.17	(0.82, 1.65)	1.12	(0.75, 1.67)
Social support scale	1.00	(0.98, 1.02)	0.98	(0.97, 1.00)
Relationship status
Married or cohabitating	Ref	…	Ref	…
Widowed	1.01	(0.43, 2.37)	0.43	(0.13, 1.35)
Divorced	1.16	(0.80, 1.70)	1.18	(0.74, 1.89)
Separated	2.04	(1.16, 3.59)	1.67	(0.92, 3.02)
Never married	1.18	(0.90, 1.54)	1.24	(0.86, 1.77)
Religiosity or spirituality
Very important	Ref	…	Ref	…
Somewhat important	1.12	(0.87, 1.45)	1.19	(0.87, 1.62)
Not very important	1.35	(0.93, 1.98)	1.04	(0.62, 1.73)
Not important at all	1.14	(0.74, 1.74)	0.75	(0.47, 1.22)
Stressful life events
None	Ref	…	Ref	…
1–2	1.86	(1.26, 2.75)*	2.34	(1.43, 3.82)*
3+	3.50	(2.35, 5.19)*	5.15	(3.29, 8.08)*
Any past-year DSM-5 anxiety disorder
No	Ref	…	Ref	…
Yes	1.42	(1.02, 1.97)	1.80	(1.30, 2.49)*
Any past-year DSM-5 mood disorder
No	Ref	…	Ref	…
Yes	1.82	(1.40, 2.36)*	2.21	(1.62, 3.03)*
Design-adjusted goodness-of-fit test	p = .7623		p = .0668

AOR = adjusted odds ratio; CI = confidence interval; DSM-5 = Diagnostic and Statistical Manual of Mental Disorders Fifth Edition; TUD = tobacco use disorder.

Tests of significance: *p < .005.

^aSexual minority adults defined as lesbian or gay-identified, bisexual-identified, or heterosexual-identified with discordant attraction or behavior.

Discussion

This study used a nationally representative sample of US adults to examine associations of racial or ethnic discrimination and sexual orientation discrimination with TUD severity. Overall, we found a disproportionate prevalence of having any TUD and moderate-to-severe TUD among sexual minorities relative to heterosexual adults. Bisexual individuals had the highest prevalence of any TUD as well as moderate-to-severe TUD. Moreover, the prevalence varied by race or ethnicity, with African American bisexual-identified individuals bearing the greatest burden for having any TUD as well as moderate-to-severe TUD. Our findings underscore the nuance and complexity of TUD disparities and the importance of considering potential racial or ethnic differences and the multiple minority statuses that individuals may hold.

Consistent with prior research demonstrating the positive associations between discrimination and health risk behaviors,^{8,23–25,27,28} our study found that regardless of sexual orientation status, experiencing greater lifetime racial or ethnic discrimination increased the odds of having any TUD and of having moderate-to-severe TUD. Among sexual minorities, greater lifetime racial or ethnic and sexual orientation discrimination resulted in increased odds of having any TUD. Furthermore, the observation that African American bisexual individuals had the highest prevalence of any TUD and moderate-to-severe TUD suggests that people with multiple minority statuses may experience greater exposure to minority stressors, which directly influence their health risk behaviors and TUD severity.

Tobacco cessation professionals should be cognizant of the minority stressors experienced by their clients and their impact on them. To avoid structural or interpersonal discrimination, some individuals may choose not to disclose their sexual minority status. Moreover, sexual identity labels may change over time,⁵¹ vary based on situational context, or vary based on cultural, racial, ethnic, socioeconomic, and age groups.⁵² It is important for cessation professionals to avoid making assumptions about their clients’ sexual orientation. Tobacco cessation professionals should create an inclusive treatment environment where clients can feel safe to discuss their minority stressors. Receiving culturally appropriate interventions remains a challenge for sexual minority clients as cultural competency training for tobacco cessation professionals is inconsistent.⁵³ Increased training is needed on the health promotion needs of sexual minority populations, especially related to tobacco use.⁵³

Our study also found that individuals who experienced stressful life events had greater odds of moderate-to-severe TUD, compared with those who had no life stressors. Moreover, mood and anxiety disorders were significant positive correlates of moderate-to-severe TUD in both sexual minority and heterosexual populations. Tobacco cessation professionals should consider which clients may have more severe TUD and be more susceptible to relapse due to their prior stressful life events and to consider this in the context of unique minority stressors that may also be present. Vulnerability to stress and negative affect may interfere with cessation efforts and remission.³² Clients engaged in tobacco cessation treatment may need to learn alternative ways to cope with stress and negative affect, rather than using tobacco. This underscores the importance of psychiatric comorbidities, stress management, and emotional regulation during treatment and remission.⁵⁴

The strengths of this study include the use of a large, nationally representative, probability-based sample of sexual minority adults in the United States. This enabled us to include many theoretically significant covariates in our models examining TUD severity among sexual minorities. The NESARC-III also assesses multiple dimensions of sexual orientation by asking respondents about their sexual identity, attraction, and behavior, allowing us to include heterosexual-identified sexual minorities in our analyses. Additionally, the database includes separate measures of sexual orientation and racial or ethnic discrimination, along with validated measures of TUD, mood disorder, and anxiety disorder based on DSM-5 criteria.

There were several limitations to this study. First, the survey items related to tobacco use and mental health relied on self-report, which is subject to social desirability and could influence the prevalence of these outcomes. Second, the questions regarding the experiences of racial or ethnic and sexual orientation discrimination did not capture the severity of the discrimination experienced. Third, the racial or ethnic and sexual orientation discrimination scale measured the respondents’ experience of one or both types of discrimination and thus, the scores were not reflective of multiplicative discrimination. Fourth, these questions focused primarily on individual-level rather than structural-level discrimination. Federal and state protections against sexual orientation discrimination are not universal and structural discrimination is pervasive for sexual minority populations. Thus, these findings may under-represent the breadth of discrimination experienced by sexual minority populations and therefore, its association with TUD severity.

Fifth, a more recent version of the NESARC study is not yet available. Language describing sexual identity has since evolved; the operationalization of sexual minorities in this study does not represent all identities that exist (eg, asexual, demi-sexual, pansexual, etc.). Moreover, multiple changes in policy have occurred at the structural level (eg, enactment of nondiscrimination laws protecting sexual minorities), which can affect their experiences of discrimination at the interpersonal level. The United States is currently experiencing historical unrest with greater discourse on racial or ethnic discrimination. This study may not represent the frequency or severity of discrimination currently experienced by certain racial or ethnic minority groups given recent events. Sixth, the NESARC-III did not include some important measures, such as gender identity. Lastly, because of this study’s cross-sectional design, we were unable to infer causality.

Future research should include prospective longitudinal studies to elucidate the mechanisms between racial, ethnic, or multiplicative discrimination and TUD severity among sexual minority populations. There is growing awareness about sexual diversity in the United States and a call for greater understanding of the health needs of sexual minority populations.⁵² Because the landscape continues to change as sexual minorities become more visible, it is important to assess the trends related to TUD among sexual minorities. Additional research is needed to understand the association between discrimination and TUD among under-represented racial or ethnic subgroups (ie, Asian, Native Hawaiian, other Pacific Islander, American Indian or Alaska Native). Other types of discrimination (eg, based on educational attainment) should also be explored in future work. Lastly, small subgroup sample sizes may have influenced our results; additional work with larger cell sizes to examine potential subgroup differences are warranted.

In conclusion, this study elucidated the relationship between experiencing discrimination and the odds of TUD among sexual minority and heterosexual adults. While tobacco use may provide perceived temporary relief from discrimination and stress among users, it also increases the morbidity and mortality of smokers and those exposed to secondhand smoke.² Our study showed a strong association between racial or ethnic discrimination and TUD severity among sexual minority and heterosexual people. It also demonstrated that exposure to higher levels of discrimination increases the odds of TUD among sexual minority adults. Individuals exposed to multiple minority stressors may have increased vulnerability for developing TUD and related adverse health consequences. Eliminating discrimination and developing interventions that are sensitive to the role minority-based discrimination plays in TUD severity may attenuate the tobacco use disparities between sexual minority and heterosexual people.

Supplementary Material

A Contributorship Form detailing each author’s specific involvement with this content, as well as any supplementary data, are available online at https://academic.oup.com/ntr.

Funding

The development of this article was supported in part by research grants R01DA036541, R01CA203809, and R01CA212517 from the National Cancer Institute and National Institute on Drug Abuse at the National Institutes of Health.

Declaration of Interests

None declared.