Introduction

Smoking is a leading cause of death in the United States, accounting for nearly one in five deaths every year (CDC, 2021). Alongside its link to mortality, smoking is a risk factor for the development of major chronic conditions including heart disease, stroke and lung cancer (CDC, 2021), as well as renal failure, breast and prostate cancer (Carter et al., 2015). Considering that as of 2020, nearly 13% of U.S. adults (30.8 million) are persons who are currently smoking, with more than 16 million Americans suffering from smoking-related diseases (Cornelius et al., 2022), identifying population subgroups with a disproportionately high risk of smoking is important for public health knowledge and policy.

Existing research shows that risk of smoking is differentially distributed across population subgroups, including race/ethnicity (Mowery et al., 2015) and sexual orientation (Blosnich et al., 2013; Tuthill et al., 2020; Wheldon et al., 2018). Cigarette smoking rates are highest among American Indian/Alaskan Native adults and lowest among Latinx and Asian adults (Drope et al. 2018), and the health consequences of smoking are also variable by race/ethnicity. For example, black adults who smoke face a particularly higher risk of major chronic conditions such as cancer compared to other racial/ethnic groups (Alexander et al., 2016; Lortet-Tieulent et al., 2017; Shalala et al., 1998). Risk of cigarette smoking also varies by sexual orientation, with a higher risk among sexual minority persons compared to those who identify as heterosexual (Amroussia et al., 2020; Assari and Barzagan 2019). While informative, existing literature is limited in that separate strands tend to document smoking disparities based on single identities, with relatively little study on how the intersections of population characteristics shape associations with smoking. This is essential to consider, since the small number of studies that do exist demonstrate substantial within-group difference—including, for example, that black and Asian sexual minority adults report lower risk of cigarette smoking compared to their white counterparts (Ortiz et al., 2015).

In this paper, we contribute to existing scholarship by employing an intersectional perspective, both theoretically and methodologically, to examining adult cigarette smoking. The aims of an intersectional perspective lie in showing how interlocking systems of privilege and oppression are reflected in the experiences of persons at the juncture of multiple social categories or identities (Bowleg, 2012; Cho et al., 2013) and promoting empowerment of people from multiple historically oppressed and marginalized identities (Collins & Bilge, 2020). Since such categories or identities are construed as fluid, flexible and mutually constitutive (Bowleg, 2008; Hankivsky, 2012), an intersectional perspective takes less of an additive and more of a multiplicative approach. This approach rejects the premise that identities can be ranked or that social inequality “adds on” with each additional identity, all based on the implicit assumption that experiences exist as separate, independent and summative (Bowleg, 2008; Hankivsky, 2012). In support of such a perspective, we aim to show how intersections of racial/ethnic and sexual identity are associated with disparities in smoking. Despite being limited to individual-level data, we acknowledge the tendency among quantitative research with less focus on social power or structural factors in favor of individual-level interventions (Guan et al., 2021) and discuss how multiple systems of oppression such as racism and heteronormativity may shape individual disparities in smoking. Methodologically, we construct a measure that intersects identities based on the idea of social categories as always permeated by other categories (Cho et al., 2013) and thus having an inherently joint and simultaneous relationship. Rather than examining the respective main effects of racial/ethnic and sexual identity and then testing for the interactivity between the two identities, we use the method of categorical cross-classification. Relatedly, we examine whether our findings are supportive of an additive or a multiplicative approach to social inequality and discuss their implications for both intersectionality as a theory and a method.

To that end, we utilize data from the 2010 Social Justice Sexuality Project (SJSP), a national survey of lesbian, gay, bisexual, and transgender (LGBT) people, and ask the following research questions. First, how are intersections of racial/ethnic and sexual identity associated with smoking among U.S. adults? We examine smoking risk both across race/ethnicity within a specific sexual orientation (e.g., between white and black gay/lesbian adults) and across sexual orientation within a particular racial/ethnic identity (e.g., between Asian/Pacific Islander heterosexual and bisexual adults). Second, do the associations between intersections of racial/ethnic and sexual identity and smoking remain after adjusting for demographic, socioeconomic, and social support covariates? In particular, we focus on gender identity as an important covariate since literature has established not only higher smoking prevalence among men than among women (Carter et al., 2015; Smith et al., 2016; Syamlal et al., 2014) but also greater risk of smoking among the transgender population (Cornelius et al., 2022; McElroy et al., 2011).

With an intersectional perspective that facilitates identification of populations facing higher risks of smoking, more effective public health intervention strategies tailored to their unique needs and experiences will become viable.

Background

Smoking Disparities: Race/ethnicity and Sexual Orientation

An established body of work documents unequal participation in cigarette smoking across sociodemographic groups, including based on race/ethnicity and sexual orientation. Single-identity studies are consistent in documenting elevated cigarette smoking among selected racial/ethnic groups (e.g., American Indian/Alaskan Native) and among sexual minority adults, especially bisexual adults. Recent work has shown that heterogeneity within social groups requires a thoughtful consideration of how intersectional advantage and disadvantage differentially benefits and harms communities (Cech, 2022).

Focusing on cigarette smoking, studies show that disparities in smoking patterns are not clear-cut. In the United States, cigarette smoking is highest among American Indian/Alaskan Native adults (21%) and lowest among Asian adults (7.2%) (Mowery et al., 2015). Traditionally, health scholars have used white adults as the reference group in research, highlighting important differences in health risk across racial/ethnic minority status. While American Indian/Alaskan Native adults have higher rates of smoking on average compared to white adults, black, Latinx, and Asian adults report lower rates of smoking relative to white adults (Cornelius et al., 2022). Comparisons across racial/ethnic minority groups also document differential rates of cigarette smoking—for example, smoking among black adults (14.9%) is almost twice the rate of Asian adults (7.2%) (Cornelius et al., 2022). This work emphasizes the heterogeneity of health risk in smoking across race/ethnicity and highlights the complexity of health behaviors across minority groups, rather than an equal distribution of health risk.

Turning to sexual orientation, studies show that sexual minority adults report higher rates of cigarette smoking than heterosexual adults (King et al., 2021; Lee et al., 2009; Wheldon et al., 2018), and that they start smoking at earlier ages and report smoking more frequently relative to heterosexual adults (Corliss et al., 2013; Watson et al., 2018). Similar to findings on health behaviors of transgender adults, scholars have pointed to unhealthy coping to stressors related to a stigmatized sexual identity as explanation for the elevated rates of smoking among sexual minority populations (Blosnich et al., 2013; Meyer, 2003). Researchers have also emphasized the historic aggressive tobacco marketing that has targeted the LGBTQ community (Fallin et al., 2015).

Scholarship has also considered differences among specific sexual identity groups, with studies showing that bisexual adults have higher odds of cigarette smoking compared to lesbian and gay adults (Hoffman et al., 2018; Tuthill et al., 2020). Scholars point to the “double marginalization” or “bi-negativity” of bisexual adults within the heterosexual and sexual minority community as a possible explanation for increased levels of unhealthy coping (Dodge et al., 2012; Eliason, 2000). Data also indicates that there may be higher rates of cigarette smoking among sexual minority gay men relative to sexual minority lesbian women (Tuthill, 2021; Tuthill et al., 2020). Overall, this work suggests that the diversity in the experiences of sexual identity groups may contribute to important differences in health profiles, and more generally intersectional theorizing of identity and inequality can benefit from empirical assessments of the distribution of health risk across multiple minority statuses (Scheim & Bauer, 2019; Schmitz et al., 2020).

Studies that document differential exposure to health risk across minority identity groups emphasize important differences among similarly marginalized groups. For example, research shows a wider gap in cigarette smoking rates between sexual minority women relative to heterosexual women compared to smoking rates between sexual minority and heterosexual men (Fallin et al., 2015), thus highlighting a wider disparity among women across sexual identity. Additionally, LGBTQ studies document differences across racial/ethnic minority groups with some sexual minority adults of color reporting higher rates of cigarette smoking compared to others (King et al., 2021; Ortiz et al., 2015).

Despite increasing attention to the relationship between cigarette smoking and sexual orientation, there remain gaps in the literature regarding smoking patterns. Currently, there is limited health information of individuals who identify with “non-traditional” sexual identities such as pansexual, sexually fluid, or queer (Burgwal et al., 2019). This is important given that studies document an increase in individuals who do not identify as heterosexual or as a gay, lesbian, or bisexual person (Miranda et al., 2018). Researchers have also found that a large proportion of sexual minority adults who also belong to racial/ethnic minority groups choose to self-identify in more racial and culturally specific ways such as ‘‘same gender loving’,’ ‘‘in the life,’’ or ‘‘pasiva’’ (Rodriguez, 2003).

Yet, much less is known about the experiences and health profiles of sexual minority adults who identify with these less traditional sexual identities. Recent work suggests there may be important differences in health risk, indicating a need for more work that is more inclusive of a variety of sexual identities. For example, a study by Azagba et al. (2020) found that individuals who report that they “do not know” or “are unsure” of their sexual identity also report lower rates of smoking, relative to heterosexual adults and other sexual minority groups. This study also found that those who reported “other” as a sexual identity reported higher rates of smoking relative to both heterosexual and sexual minority adults.

Given the diversity of the sexual minority community and the rise in the number of individuals identifying with different variations of sexual identity categories, it is important to expand our framing of sexual identity within health disparities work. In including sexual identities that are often omitted from population health studies, we also assess smoking risk among empirically under-examined populations.

Intersectionality: Additive and Multiplicative Approaches to Health Risk

Health scholarship is increasingly responding to the call by the Institute of Medicine to draw on an intersectional framework when assessing health disparities (IOM 2011; Liu et al., 2017; Tuthill et al., 2020), finding that interlocking systems of social inequalities contribute to significant health differences across social position, highlighting the complexities of health inequities. This work includes attention to health status and health behaviors across the intersection of social identities that include race/ethnicity and sexual orientation (Amroussia et al., 2019; Grollman, 2014).

At the same time, mixed findings from health studies examining disparities among multiply marginalized groups complicate our understanding of how health risk is distributed across race/ethnicity and sexual orientation groups. Scholars have called for more consideration of how the intersection of privilege and disadvantage shapes health risk, critiquing the framing of minority groups as only disadvantaged (Ailshire & House, 2011). Some of this work emphasizes how the diversity of marginalization produces important differences in health patterns across population groups. Missing, however, is more discussion of how these patterns align with our conceptualization of health inequities within an intersectional framework. This is important since an intersectional framework is both a theoretical and a methodological approach that can effectively examine different axes of social inequality. Intersectionality can help us better assess how systems of power (such as racism and heterosexism) maintain and reproduce social hierarchies (Crenshaw, 1991; Davis, 2008). Sociologists have been drawn to intersectional analyses because of its emphasis on how social structures (such as policies, knowledge production, and institutional practices) maintain unequal power dynamics (Collins, 1990; Crenshaw, 1989). Health scholarship that draws on intersectional frameworks to assess patterns of inequities seeks to understand how consistent and widening health disparities are reproduced for socially disadvantaged individuals (Bowleg & Bauer, 2016; Cho et al., 2013).

As a theoretical orientation, intersectionality is a reflexive, critical, and epistemological framework that highlights the mechanisms that oppress, disadvantage, and privilege social groups (Bowleg, 2012; Luna & Pirtle, 2022). From its origin, intersectionality centered on the marginalization of black women, highlighting the way their lived experiences are harmed by the simultaneous impact of sexism and racism (Crenshaw, 1989, 1991). While much of the work continues to focus on the experiences of black women (Luna & Pirtle, 2022), scholars have expanded the analytical lens of intersectionality to examine how a variety of minority groups are similarly (and differently) marginalized.

As a methodology, intersectionality can be a useful tool to assess and understand how social processes generate inequalities. A diverse group of scholars using qualitative and quantitative analyses utilize intersectional approaches to highlight how systems of power are embodied, consumed, and reproduced (Berg, 2010; Bowleg & Bauer, 2016). With such an expansive use of intersectional analyses, several scholars have called for more discussions that evaluate how methodologies can build upon theoretical conversations (Ailshire & House, 2011; Cho et al., 2013). For example, within the health disparities literature, scholars have been critical of intersectional work that focuses exclusively on analyses that fail to engage with theoretical tenets of intersectionality (Richman & Zucker, 2019). These scholars argue that currently, there is more literature produced that document differences in health patterns with limited discussion about how these patterns reflect the mechanisms of oppressive systems in which they are embedded. Missing, then, is a more nuanced conceptualization of the unequal health burden across population groups.

Intersectional quantitative health research utilizes a variety of statistical analyses to capture interactive effects across identity categories. A large body of work relies heavily on interactions between identity categories to examine health disparities among different types of communities (Hancock, 2007; McCall, 2005). In generating interactions across categories of race/ethnicity and sexual orientation, these studies attempt to capture the interdependence of disadvantage across social position. A growing body of work, however, has shifted to using stratified models, multilevel modeling, and the use of intersectional identity categories to map disparate health outcomes (Agénor et al., 2019; Bauer et al., 2021). Regardless of which statistical analysis is used, however, scholars have noted the challenges in acknowledging the unique experiences across combinations of minority status, while also connecting similarities in how marginalization negatively impacts wellbeing (Ailshire & House, 2011).

In-depth analyses of cigarette smoking disparities across the intersection of multiple minority status remain limited. This dearth of empirical evidence has contributed to a lack of clarity as to how multiple types of social disadvantage sustain health disparities. For example, some research suggests there is significant heterogeneity in cigarette smoking patterns once multiple identities such as race/ethnicity and sexual orientation are considered (Emory et al., 2016; Tuthill et al., 2020). This work suggests that health risk may not distribute uniformly across social identities. Instead, it suggests that the combination of social advantage and disadvantage produces unique intersectional experiences that contribute to important differences in health risk patterns across population groups. Other studies, however, suggest that multiple minority status (regardless of minority identity) aligns with an increased likelihood in engagement with health risk behaviors due to an accumulation of social disadvantage (Grollman, 2014). This work emphasizes the aggregate of minority stressors and discrimination stemming from multiple types of inequalities as compounding health risk for minority groups. In these studies, the type of minority status is less important than the number of minority statuses in understanding health disadvantage among multiply marginalized individuals.

Overall, findings from this study make three important contributions to intersectional research. First, our results can enhance theoretical understandings of one of the core tenets of intersectionality—understanding the relationship between power and identity. Population health research that points to differences without understanding how categories of identity are linked with power dynamics neglects one of the foundational aims of intersectionality—reducing inequality (Richman & Zucker, 2019). Second, we provide thorough discussions on what the distribution of health risk across intersections of minority categories represents regarding access to resources and exposure to harm. Scholars argue that linking identity categories to the social factors that contribute to unequal health risk can better inform interventions to reduce health inequities. Finally, in using an intersectional approach in our methodology to engage with theoretical components, we maximize the empirical potential of intersectionality as an analytic tool.

Data and Methods

Data

We draw on data from the 2010 Social Justice Sexuality Project (SJSP),Footnote 1 a national survey of lesbian, gay, bisexual, and transgender (LGBT) people (see https://socialjusticesexuality.com/ for more information). The survey was administered to over 5,000 respondents from all U.S. states, and also included those from Washington, D.C. and Puerto Rico. The SJS Project was designed to examine the socio-political experiences of the LGBT population with five core themes: racial and sexual identity, spirituality and religion, mental and physical health, family formations and dynamics, and civic and community engagement. One of the strengths of this dataset is that it oversampled sexual minority persons of color, including those who identify as black, Latinx, Asian/Pacific Islander, and Multiracial. The sample, however, is non-probability based, generated through various recruiting strategies ranging from venue-based sampling at strategic events, snowball sampling, and respondent-driven sampling to sampling on the Internet. Our analytic sample is restricted to 4,114 respondents with valid information on race/ethnicity, sexual orientation, gender identity, and current cigarette smoking.

Measures

Our dependent variable is current cigarette smoking. Respondents were asked, “Do you now smoke cigarettes?” to which they could reply ‘not at all,’ ‘some days,’ or ‘every day.’ Responses were dichotomized to indicate current cigarette smoking status with 0 = not at all, 1 = some days or every day. It is important to note that our measurement of smoking is limited in its ability to distinguish between different types of smoked or non-smoked tobacco products such as e-cigarettes and menthol cigarettes, despite existing literature on racial/ethnic or sexual orientation differences in their use (Cohn et al., 2019; Ganz & Delnevo, 2021; Goodwin et al., 2023). Additionally, we decided to combine ‘some days’ smoking with ‘every day’ smoking in consideration of sample size, but it should be noted that persons who smoke daily differ from their non-daily counterparts in various sociodemographic characteristics, including race/ethnicity (Wang et al., 2018).

Our independent variable of interest is a combined measure of race/ethnicity and sexual orientation. First, racial/ethnic identity measures whether a respondent identifies as white, black, Latinx, Asian or Pacific Islander, or Multiracial/Other. In the original questionnaire, respondents were able to choose multiple options from white, black, Latinx, Asian or Pacific Islander, Native American, Multiracial, or Other. We classified those who chose more than one option as Multiracial, so that each racial/ethnic category includes those who identify exclusively with one group. Respondents who identified as Native American,Footnote 2 Multiracial, or Other were combined as ‘Multiracial/Other.’

Second, sexual identity measures whether a respondent identifies as heterosexual, gay/lesbian, bisexual, or an alternative sexual identity (e.g., Two Spirit, In the Life). In the original questionnaire, respondents were asked, “Which one label comes closest to how you describe your sexual identity?”, to which they could choose ‘straight/heterosexual,’ ‘gay,’ ‘lesbian,’ ‘bisexual,’ ‘two spirit,’ ‘queer,’ ‘in the life,’ ‘same gender loving,’ ‘macho/a,’ ‘activo/a,’ ‘pasivo/a,’ or ‘other.’ We simplified our measure by classifying those who chose options other than straight/heterosexual, gay/lesbian, and bisexual as having an alternative sexual identity.

Based on these two measures, our analytic sample includes 64 white heterosexual adults, 615 white gay/lesbian adults, 91 white bisexual adults, 152 white alternatively-identifying adults, 110 black heterosexual adults, 850 black gay/lesbian adults, 137 black bisexual adults, 309 black alternatively-identifying adults, 57 Latinx heterosexual adults, 413 Latinx gay/lesbian adults, 65 Latinx bisexual adults, 84 Latinx alternatively-identifying adults, 39 Asian/Pacific Islander heterosexual adults, 143 Asian/Pacific Islander gay/lesbian adults, 21 Asian/Pacific Islander bisexual adults, 55 Asian/Pacific Islander alternatively identifying adults, 71 Multiracial/Other heterosexual adults, 500 Multiracial/Other gay/lesbian adults, 123 Multiracial/Other bisexual adults, and 215 Multiracial/Other alternatively-identifying adults (see Table 1).

Table 1 Sample characteristics, by race/ethnicity and sexual orientation (n = 4114)
Full size table

We control for a range of demographic and socioeconomic characteristics as well as social support measures in our multivariate analyses. For demographic controls, we include age at interview (range from 18 to 81), gender identityFootnote 3 (cismale, cisfemale, male-to-female transgender, female-to-male transgender, alternative gender), nativity (1 = foreign-born, 0 = born in the U.S.), and marital status (1 = married or partnered, 0 = not partnered or other). To adjust for socioeconomic characteristics, we account for education level (1 = high school or less, 2 = some college or associate’s degree, 3 = college or more), employment status (1 = employed for wages or self-employed, 0 = student, in the military, retired, unemployed, or on public assistance), household income (1 = total annual household income below $30,000, 0 = all higher), and health care coverage (1 = has health insurance, 0 = does not have health insurance). We also adjust for measures of social support. First, we include frequent attendance at religious services (1 = at least 2–3 times per month, 0 = about once a month or less). Second, level of outness is constructed as a summed score across six items (alpha = 0.89) that ask respondents to identify how many members they are ‘out’ to among family members, friends, coworkers, neighbors, and religious and online communities, respectively, where 0 = none and 4 = all (range from 0 to 24). Responses from heterosexual adults were coded as 0 because the items were not applicable. Third, level of connectedness to LGBT communities is measured as a summed score across three items (alpha = 0.77) that ask respondents to report on how much they feel connected to their local LGBT communities; feel that the problems faced by the LGBT communities are also their problems; and feel a bond with other LGBT people, where 1 = strongly disagree and 6 = strongly agree (range from 3 to 18). Fourth, level of participation in LGBT communities is measured as a summed score across six items (alpha = 0.81), where 1 = never and 6 = more than once a week (range from 6 to 36). Respondents were asked: “Thinking about LGBT groups, organizations, and activities in general, during the past 12 months, how often have you (1) participated in political events, (2) participated in social or cultural events, (3) read newspapers or magazines, (4) used the Internet for chatrooms, social networking sites, blogs, etc., (5) received goods and/or services such as medical care and counseling, and (6) donated money to an organization.”

Analysis

All analyses were conducted using Stata 15.0. Using logistic regression predicting current cigarette smoking, we present a series of models beginning with the baseline model that only includes race/ethnicity and sexual orientation (Model 1), adjusting for age. In Model 2 we include other demographic characteristics and add socioeconomic characteristics in Model 3. In the fully adjusted model, we account for all demographic and socioeconomic characteristics as well as measures of social support (Model 4). All missing data have been multiply imputed using chained equations in Stata.Footnote 4

Results

Sample Characteristics

Table 1 presents sample characteristics by race/ethnicity and sexual orientation. In general, sexual minority adults tend to exhibit higher rates of current cigarette smoking, with the exception of white adults among whom heterosexual adults have the highest rate of current cigarette smoking (38%). Looking first at demographic characteristics, gay and lesbian adults in each racial/ethnic group tend to be the oldest, and cisgender adults constitute the majority across race/ethnicity and sexual orientation groups. Interestingly, the rates of those who identify as an alternative gender are greatest among sexual minority adults who identify with an alternative sexual identity in each racial/ethnic group. As to nativity, the foreign-born proportion is noticeable among Latinx and Asian/Pacific Islander adults, regardless of sexual orientation.

Turning to socioeconomic characteristics, those with at least a college degree tend to constitute the largest category across race/ethnicity and sexual orientation groups, except among Latinx and Multiracial/Other adults. In terms of employment status, gay and lesbian adults in each racial/ethnic group have the highest rates of employment, except for white gay and lesbian adults whose rate is slightly lower than that among white bisexual adults. As to low-income status (indicated by total annual household income below $30,000), heterosexual adults display the greatest rates of those with low income among white, Asian/Pacific Islander, and Multiracial/Other adults, while the rate is lowest among heterosexual adults for Latinx adults. Health insurance rates tend to be generally high across groups.

As to measures of social support, for frequent attendance at religious services, black adults, regardless of sexual orientation, display the highest rates. Average levels of outness tend to be highest among gay and lesbian adults regardless of race/ethnicity, followed by alternatively-identifying adults, and bisexual adults have the lowest levels except among Asian/Pacific Islander adults. In terms of both levels of connectedness to and participation in LGBT communities, bisexual adults tend to display the lowest scores in each racial/ethnic group among sexual minority adults.

Logistic Regression Models Predicting Current Cigarette Smoking

Table 2 displays odds ratios from logistic regression models predicting current cigarette smoking. In the baseline model adjusting only for age, we find significant difference for black and Asian/Pacific Islander adults. Specifically, black and Asian/Pacific Islander heterosexual adults have about 70% lower odds of current cigarette smoking compared to white heterosexual adults (p < 0.01 and p < 0.05, respectively). Black and Asian/Pacific Islander gay and lesbian adults, likewise, report about half the odds of current cigarette smoking than white heterosexual adults (all p < 0.05). When we add other demographic characteristics in Model 2, however, difference for black and Asian/Pacific Islander gay and lesbian adults turns nonsignificant, while black and Asian/Pacific Islander heterosexual adults continue to have lower odds of current cigarette smoking than white heterosexual adults. Regarding gender identity, female-to-male transgender adults report more than two times the odds of current cigarette smoking compared to those who identify as cismale (p < 0.01), and cisfemale adults also report about 18% higher odds of current cigarette smoking than cismale adults (p < 0.05).

Table 2 Odds ratios from logistic regression models predicting current cigarette smoking (n = 4114)
Full size table

When we add socioeconomic characteristics in Model 3, we find that black and Asian/Pacific Islander heterosexual adults continue to have lower odds of current cigarette smoking compared to white heterosexual adults, and female-to-male transgender adults still report about two times the odds of current cigarette smoking than those who identify as cismale. Cisfemale adults also continue to have greater odds of current cigarette smoking compared to cismale adults. In our fully adjusted model with inclusion of social support measures, we find that black and Asian/Pacific Islander heterosexual adults still have about 70% lower odds of current cigarette smoking than white heterosexual adults (p < 0.01 and p < 0.05, respectively). We also find that gay and lesbian adults from white, black, and Asian/Pacific Islander groups gain statistical significance with about half the odds of current cigarette smoking compared to white heterosexual adults.

Female-to-male transgender adults, though slightly reduced, continue to report greater odds of current cigarette smoking compared to those who identify as cismale (p < 0.05), but cisfemale adults do not significantly differ from cismale adults once we adjust for social support. Socioeconomic characteristics reveal that higher level of education is associated with significantly reduced odds of current cigarette smoking (all p < 0.001). Specifically, having some college level of education is associated with 36% lower odds of current cigarette smoking, and having at least a college degree is associated with even more reduced odds (64%) of current cigarette smoking. Additionally, having a health insurance is also associated with about 21% lower odds of current cigarette smoking. As to measures of social support, we find that whereas frequent attendance at religious services is negatively associated to current cigarette smoking, level of outness in contrast is positively associated (p < 0.001).

To facilitate comparison across groups, we visualized our findings by generating predicted probabilities of current cigarette smoking for the fully adjusted model. When we first look at differences by sexual orientation within each racial/ethnic group (see Fig. 1), we see that black and Asian/Pacific Islander adults are the only groups with significant difference by sexual orientation. Among black adults, bisexual (Pr = 0.30, 95% CI [0.22, 0.38]) and alternatively identifying (Pr = 0.28, 95% CI [0.23, 0.33]) adults are more likely to be currently smoking cigarettes than heterosexual adults (Pr = 0.18, 95% CI [0.10, 0.27]). Among Asian/Pacific Islander adults, bisexual adults (Pr = 0.53, 95% CI [0.32, 0.74]) stand out with a greater likelihood of current cigarette smoking compared to either heterosexual adults (Pr = 0.18, 95% CI [0.05, 0.31]) or gay and lesbian adults (Pr = 0.24, 95% CI [0.17, 0.31]).

Fig.1
figure 1

Adjusted Predicted Probabilities and 95% CIs of Current Cigarette Smoking by Race/ethnicity. Note (1) Letters indicate significant pairwise comparison (all at p < .05) across sexual orientation groups; a = heterosexual, b = gay/lesbian. (2) Heterosexual, gay and lesbian, bisexual, and alternatively-identifying respondents are abbreviated as ‘Het,’ ‘G/L,’ ‘Bi,’ and ‘Alt,’ respectively, due to space limit

Full size image

When we then look across racial/ethnic groups by sexual orientation (see Fig. 2), we first find that black and Asian/Pacific Islander heterosexual adults (Pr = 0.18, 95% CI [0.10, 0.27] and Pr = 0.18, 95% CI [0.05, 0.31], respectively) are less likely to be currently smoking cigarettes compared to white heterosexual adults (Pr = 0.40, 95% CI [0.28, 0.52]), and Latinx heterosexual adults (Pr = 0.35, 95% CI [0.22, 0.47]) are more likely to be currently smoking cigarettes than black heterosexual adults. Among gay and lesbian adults, black respondents (Pr = 0.23, 95% CI [0.20, 0.26]) are less likely to be currently smoking cigarettes compared to Latinx (Pr = 0.34, 95% CI [0.30, 0.39]) and Multiracial/Other groups (Pr = 0.33, 95% CI [0.29, 0.37]), whereas they do not differ significantly from white and Asian/Pacific Islander adults. Latinx gay and lesbian adults, on the other hand, have a higher likelihood of current cigarette smoking compared to all other racial/ethnic groups, except for those who identify as Multiracial/Other. Gay and lesbian adults who identify as Multiracial/Other (Pr = 0.33, 95% CI [0.29, 0.37]) are more likely to be currently smoking cigarettes than their white, black and Asian/Pacific Islander counterparts. Lastly, among bisexual adults, Asian/Pacific Islander adults (Pr = 0.53, 95% CI [0.32, 0.74]) are distinctive with a greater likelihood of current cigarette smoking compared to black and Latinx adults (Pr = 0.30, 95% CI [0.22, 0.38] and Pr = 0.27, 95% CI [0.17, 0.37], respectively).

Fig.2
figure 2

Adjusted Predicted Probabilities and 95% CIs of Current Cigarette Smoking by Sexual Orientation. Note (1) Letters indicate significant pairwise comparison (all at p < .05) across racial/ethnic groups; a = White, b = Black, c = Latinx, d = Asian/Pacific Islander. (2) Asian/Pacific Islander and Multiracial/Other respondents are abbreviated as ‘Asian/PI,’ and ‘Multi/Oth,’ respectively, due to space limit

Full size image

Discussion

Results highlight important variation in risk of cigarette smoking across the intersection of race/ethnicity and sexual orientation, emphasizing the heterogeneity of health patterns across social position. Additionally, our study provides valuable information about smoking patterns for empirically under-examined sexual identity groups. Findings also reinforce results from previous literature that document health protective factors against risk of smoking. Overall, there are three key findings to highlight.

First, our analyses support a more multiplicative framing of how health risk is distributed across identity groups. Our results show that while some respondents with two minority identities maintain a higher predicted probability of cigarette smoking (such as Asian/Pacific Islander bisexual adults), others do not (such as black gay and lesbian adults). Often, work on sexual minority persons of color provide evidence of a “double jeopardy” effect, where the aggregate of two disadvantaged statuses contributes to higher health risk (Stein et al., 2010). Our results, however, suggest more of a multiplicative interpretation of “double jeopardy” where the combination of social disadvantage compounds each other to produce more variable risk of cigarette smoking across the intersection of social identities. This contrasts an additive interpretation of “double jeopardy,” which would find more equal risk of smoking among shared minority identities (Berdahl & Moore, 2006; Jackson et al. 1982).

To start, we found racial/ethnic differences in odds of cigarette smoking among heterosexual adults. Models document significantly lower odds of smoking among Asian/Pacific Islander and black heterosexual adults relative to white heterosexual adults. These findings align with previous work documenting differential rates of cigarette smoking across race/ethnicity (Cornelius et al., 2022). Our study, however, does not find evidence of significant differences in odds of cigarette smoking for heterosexual adults in other racial/ethnic minority groups (Latinx and Multiracial/Other adults). We also find significant differences in odds of cigarette smoking across sexual orientation. White, black, and Asian/Pacific Islander gay and lesbian adults have significantly lower odds of cigarette smoking relative to white heterosexual adults. Similar to findings among heterosexual adults, our study does not find evidence of significant differences in odds of cigarette smoking for gay and lesbian adults in other racial/ethnic minority groups (Latinx and Multiracial/Other adults). Thus, while people of color in the sample may be similarly disadvantaged through their racial/ethnic minority status, variation in odds of cigarette smoking across race/ethnicity suggests more group-specific differences in exposure to health risk.

Our within-group analysis also suggests that sexual minority adults of color have higher predicted probabilities of cigarette smoking relative to their heterosexual counterparts. This is especially true for black and Asian/Pacific Islander respondents, where we document significant pairwise comparisons in likelihood of cigarette smoking. These findings align with previous work which documents that “doubly marginalized” people (such as sexual minority persons of color) have accumulated health risk from two disadvantaged identities (Berdahl & Moore, 2006). This work posits that while racial/ethnic minority persons are marginalized by racist systems, sexual minority adults encounter additional disadvantage from heterosexism that contributes to accumulation of health risk.

Yet, the variation in likelihood of cigarette smoking across social position suggests that the combination of social identities matters. Rather than finding that all sexual minority adults of color share uniform levels of risk of cigarette smoking as “doubly marginalized” people, our analysis highlights important variability in predicted likelihood of cigarette smoking across the intersection of identity groups. For example, we found that while some racial/ethnic groups (such as Asian/Pacific Islander adults) have more disparate likelihood of cigarette smoking across sexual identity, others (such as black and white adults) have more evenly distributed risk.

We also found evidence that there may be more diverse experiences of social disadvantage. For example, research suggesting a “bisexual disadvantage” due to bi-negativity within both heterosexual and sexual minority communities (Friedman et al., 2014) would expect bisexual racial/ethnic minority adults to always have the highest risk of smoking. This work implies that bisexual adults of color are, in a way, a “triple marginalized” group, since they are marginalized within an already stigmatized group. Our analysis, however, suggests that this pattern may not be as evident across all racial/ethnic groups. We found that while bisexual Asian/Pacific Islander adults have the highest likelihood of cigarette smoking relative to other Asian/Pacific Islander sexual identity groups, bisexual adults in other racial/ethnic groups have similar or lower predicted probabilities of cigarette smoking relative to other sexual identity groups. For example, Latinx bisexual adults had a lower reported rate of cigarette smoking compared to gay and lesbian or alternatively-identifying Latinx adults. This suggests a less clear-cut distribution of cigarette smoking risk in considering heterogeneity in combinations of minority status.

Second, our findings suggest that additional intersectional analyses that consider how cigarette smoking risk is distributed across social position are warranted to better understand health disadvantage among marginalized communities. For example, our results document significant gender differences in odds of cigarette smoking. By including alternatively-identifying adults in our analytic sample (such as gender fluid, non-gender binary, and gender nonconforming), we assess cigarette smoking patterns among adults who identify outside the gender binary. Results show that there are no significant differences in odds of cigarette smoking between alternatively-identifying adults and cisgender men. Additionally, in considering meaningful differences in gender transition and disaggregating transgender adults by current gender identity (male-to-female and female-to-male), we found evidence of significant health risk of cigarette smoking among transgender men. It is noteworthy that while transgender men have significantly higher odds of cigarette smoking, transgender women do not. Thus, while transgender men and women share a similarly marginalized gender identity, variation in risk of cigarette smoking indicates that there may be important differences that contribute to smoking disparities within the transgender community. Our finding aligns with a growing body of scholarship that documents differences in experience between transgender men and transgender women (Catalano, 2015; Connell, 2010; Schilt, 2006). These differential experiences in marginalization indicate a need to more carefully consider how they relate to gendered health disparities.

Studies that assess cigarette smoking disparities across the intersections of race/ethnicity, sexual orientation, and gender identity can provide more insight into how health risk is distributed among multiply marginalized groups. While it is beyond our ability to fully flush out these intersections in the current paper (due to small sample sizes, especially among gender minority populations), this represents a clear area where additional scholarship is needed.

Our findings also point to the importance of being more inclusive of sexual minority identity groups. The increase in individuals who do not identify as heterosexual or as a gay, lesbian, or bisexual person (Miranda et al., 2018) calls for more consideration of health disparities across additional types of sexual identity categories. In including adults who identify with sexual identity categories beyond gay, lesbian, and bisexual in our analytic sample, we provide a more inclusive health profile of the sexual minority community. Our analysis does not show significant differences in odds of cigarette smoking among alternatively identifying adults compared to heterosexual white adults. Results show, however, that racial/ethnic minority adults who identify with an alternative sexual identity have a higher predicted probability of cigarette smoking relative to their heterosexual counterparts.

Additionally, our analysis identifies racial/ethnic differences in risk of cigarette smoking for alternatively identifying adults compared to other sexual identity groups. For example, black alternative adults have a higher likelihood of cigarette smoking relative to black heterosexual adults but a similar likelihood of smoking relative to black bisexual adults. Asian/Pacific Islander alternative adults, on the other hand, have a higher likelihood of cigarette smoking relative to Asian/Pacific Islander heterosexual adults and a much lower likelihood of cigarette smoking relative to Asian/Pacific Islander bisexual adults. Findings also document that white alternative adults have a lower predicted probability of cigarette smoking relative to their heterosexual counterparts, but a similar likelihood of cigarette smoking relative to other white sexual minority adults. Our findings indicate that while there may be similarities in how adults who identify with “non-traditional” sexual identities are exposed to health risk, racial/ethnic differences suggest meaningful differences in risk across the intersection of identity categories.

Finally, our findings point to important social determinants of health that are significant predictors of cigarette smoking. Variation in these determinants of health seems to be meaningful contributors to variation in health risk across identity groups. Models show that educational status, engagement with a religious community, level of outness, and health insurance are significantly associated with cigarette smoking. These findings highlight the link between access to health-related resources and health behaviors, rather than presuming that identity categories are solely predictive of health risk. Previous studies document how power systems that maintain social hierarchies maintain and reproduce economic disadvantage among minoritized communities through discriminatory practices that reduce access to employment, housing, and education (Hayward et al., 2000; Williams & Sternthal, 2010).

Yet, our demographic analysis also presents important differences in socioeconomic patterns across the intersection of race/ethnicity and sexual orientation. While the analytic sample is (overall) economically vulnerable, variability in socioeconomic profiles highlights complexity in the distribution of economic disadvantage. Across racial/ethnic groups, employment status and household income differed by sexual orientation. For example, reported household income among bisexual adults did not always reflect a “bisexual economic disadvantage” found in previous studies which point to a financial disadvantage among bisexual adults relative to other sexual identity groups (Badgett & Schneebaum, 2012). For several racial/ethnic minority groups (such as Asian/Pacific Islander, Latinx and Multiracial/Other adults), bisexual adults were similarly economically disadvantaged relative to other sexual identity groups. Thus, in considering economic disadvantage among racial/ethnic minority groups, economic vulnerability among sexual minority adults was less pronounced. We highlight that similarities and differences across the intersection of racial/ethnic and sexual identity groups need to be considered to better understand the relationship between economic disadvantage and health behaviors, including smoking.

Models also showed the significance of educational status as a protective buffer against smoking. We found that even “some college” significantly reduced odds of cigarette smoking, emphasizing the health value of education that has been found among heterosexual populations (Amroussia et al., 2020). This finding is particularly noteworthy for our analytic sample, which is relatively highly educated. Generally, sexual minority adults were higher educated than their heterosexual counterparts regardless of race/ethnicity. The high rates of low income coupled with high rates of higher education suggest that despite educational attainment, sexual minority adults in our analytic sample remain economically disadvantaged. This hints at additional social constraints that are contributing to lower incomes. Previous work has found that high rates of discrimination against sexual minority adults result in slower economic mobility and lower accumulated lifetime earnings (Movement Advancement Project [MAP] and Sage 2010). This also questions whether the educational health buffer has the same “returns” for sexual minority adults since it does not provide as much of an “economic bump.” Given that recent work (e.g., Mittleman, 2022) finds substantial variation in educational achievement and ‘asymmetric’ impacts of marginalization on academic success across the intersections of gender identity, sexual identity, and race/ethnicity, future research should continue to examine the intersectional relationships between educational stratification and health outcomes and behaviors.

We also find evidence that frequent religious attendance is associated with lower odds of cigarette smoking. This aligns with previous work on heterosexual populations which found that people who frequently attend church services are less likely to smoke cigarettes (Koening et al., 1988). This finding is notable for our sample since it is composed largely of sexual minority adults of color and suggests that the health protective buffer of religious attendance may extend to sexual minority communities. Our demographic analysis, however, complicates this significant finding. We document variation in frequency of religious attendance across race/ethnicity, suggesting that while some sexual minority adults of color are integrated in religious communities, others are not. For example, black and Multiracial/Other respondents had the highest rates of frequent religious attendance relative to other racial/ethnic groups. In fact, among black respondents, there were high rates of reported frequent religious attendance across sexual orientation. There were also differences in religious attendance across sexual orientation. Among Latinx and white respondents, alternatively-identifying adults had the highest rate of frequent religious attendance relative to both their heterosexual and sexual minority counterparts. These findings reaffirm the importance of considering within- and between-identity group comparisons to construct more accurate health profiles of population groups.

Finally, our findings suggest that as level of outness increases, so does the odds of cigarette smoking. While this is not applicable to heterosexual adults, our findings suggest that for sexual minority adults, increased outness to social networks may be associated with an increased risk of cigarette smoking. This aligns with previous work that finds that since sexual minority populations have elevated rates of cigarette smoking, higher levels of outness may lead to more integration with the LGBT community and thus more opportunities to smoke (Kipke et al., 2007).

While this study contributes important information to examinations of identity and health behaviors, there are limitations in its assessment. First, while multiple recruiting strategies were used, respondents from the SJSP were heavily recruited from LGBT community events. Given the targeted recruiting, respondents may be more “out” and engaged in the sexual minority community. As a result, our findings may reflect cigarette smoking patterns of more community-integrated adults. The SJSP is valuable in that it over-sampled for LGBQ people of color and gender minority adults. While this allows for a greater number of participants from minoritized communities that are typically excluded from probability-based samples, it requires more targeted sampling techniques. While this makes the SJSP unique in that it includes information on health behaviors for a large number of LGBQ people of color, it may limit the generalizability of our findings. Given that our focus is on examining racial/ethnic disparities in smoking across sexual orientation, however, we found the SJSP to be of great value. Second, given the relatively young analytic sample, findings may reflect health behavior patterns of younger sexual minority populations. Thus, we may be capturing a snapshot of cigarette smoking at a particular point in a person’s life. More longitudinal data that tracks smoking patterns across the lifespan can provide a broader profile of smoking risk. Third, our findings are limited to cigarette smoking. Over recent years, various substitutes for cigarette smoking have developed, including vaping. Considering that more than 2.5 million youth reported using e-cigarettes in 2022, with more than a quarter of them (27.6%) reporting daily use (CDC, 2022), it is important to establish if the patterns identified in our analysis hold for other types of smoking behavior. Future research that takes into account such variability in types of smoking, as well as how they relate to intersections of key population characteristics, is warranted.

Conclusion

Overall, our findings align with previous scholarship regarding smoking risk while also providing additional insight in examining health patterns across race/ethnicity and sexual orientation. Recent work has shown that heterogeneity within social groups requires a thoughtful consideration of how intersectional advantage and disadvantage differentially harms communities. Results support a multiplicative approach to health disparities and highlight diversity within similarly marginalized groups. We underscore the need to thoughtfully consider important within- and between-group differences in assessments of health inequities across identity categories. Our study is valuable to addictions research in that it is more inclusive in its approach to sexual identity groups. In addition to including gay, lesbian, and bisexual adults, adults who identify in alternative sexual identity categories (e.g., pansexual, queer, fluid) are also included. This is important given that studies document an increase in individuals who do not identify as heterosexual or as a gay, lesbian, or bisexual person (Miranda et al., 2018). Much less is known about the experiences and health profiles of sexual minority adults who identify with these less traditional sexual identities. In including adults who identify with sexual identity categories beyond gay, lesbian, and bisexual and in paying attention to important variation in patterns of cigarette smoking across race/ethnicity and sexual orientation, our study provides a more inclusive health profile of the sexual minority community that leads to better understanding about cigarette smoking disparities.

We also emphasize the link between social factors that contribute to health inequities and identity categories, ensuring that differences in health risk are not attributed to the categories themselves. Given that our results reflect racial/ethnic differences in risk of cigarette smoking across sexual identity groups, we emphasize the importance of considering the unique type of social disadvantage that is produced from the intersection of minoritized social identities. Our results suggest that more group-specific smoking intervention programs may be warranted.

Understanding differences in cigarette smoking patterns across multiple identity groups can also better inform community members and public health organizations to mitigate smoking disparities. Regulation and campaigns to educate and discourage sexual minority populations from smoking tobacco have been created to address the documented disparities in cigarette smoking risk across sexual orientation. Often, these programs and campaigns take a “one size fits all approach.” Intersectional work reinforces the importance of rejecting “simplified” framings of uniform health disparities by highlighting the variability in health behaviors across racial/ethnic groups once additional social identities are considered. Our findings suggest, for example, that Latinx gay and lesbian adults may benefit from more targeted cessation programs as they maintain a higher likelihood of cigarette smoking compared to other racial/ethnic groups. More insight into the differences in cigarette smoking patterns across race/ethnicity and sexual orientation can inform community members and health initiatives centered on reducing tobacco smoking among multiply marginalized groups.