Differential Alcohol Use Disparities by Sexual Identity and Behavior Among High School Students

Abstract

Aims

Sexual minority youth (SMY) use alcohol at disproportionate rates compared to their heterosexual peers. However, sexual orientation is multidimensional. Analyzing alcohol use disparities only by one dimension of sexual orientation may result in critical disparities being obscured.

Methods

Data from state and local versions of the Youth Risk Behavior Survey were pooled (2009–2017), resulting in a large, diverse sample (n = 201,671). Multivariable logistic regression models were used to analyze associations between sexual identity/sexual behavior and three alcohol use outcomes among sexually active youth: age at first drink, binge drinking and current drinking.

Results

SMY, when categorized by sexual identity and sexual behavior, reported greater alcohol use than their heterosexual peers, though the magnitude of these disparities varied by subgroup. Both those who identified as bisexual and those who reported sexual behavior with both males and females reported the greatest levels of alcohol use. Decomposition analysis revealed that youth whose reported sexual behavior was not aligned with stereotypical behavioral expectations based on their sexual identity had higher odds of current drinking and drinking before 13 years as compared to other youth.

Conclusions

Results highlight the need to incorporate multiple methods of sexual orientation measurement into substance use research. Interventions based solely on identity, rather than both identity and behavior, may not be sufficient; targeted research into the causes of alcohol use disparities is needed, especially for bisexual youth, youth whose sexual behavior and sexual identity are not stereotypically aligned, and youth who report a sexual identity of ‘not sure.’

Short Summary: SMY use alcohol at higher rates; the magnitude of these disparities differs by sexual identity and sexual behavior. Bisexuals and youth who had sex with males and females reported the most alcohol use. Sexual behavior that was unaligned with expectations based on sexual identity was associated with drinking.

INTRODUCTION

Alcohol misuse remains a significant public health concern among US youth (National Institute on Alcohol Abuse and Alcoholism 2005); 29.8% of youth reported current alcohol use in the 2017 Youth Risk Behavior Survey (YRBS) (Kann et al. 2018). Adolescent alcohol use is associated with a number of deleterious health outcomes, including poor school performance, depression, suicide and condomless sex, as well as an increased risk of developing alcohol dependence in adulthood (Hingson et al. 2006).

The prevalence of alcohol use is even greater among sexual minority youth (SMY) compared to their heterosexual peers (Hughes et al. 2016). Previous research demonstrated that SMY were more likely to report heavy episodic drinking, drinking alcohol in the past month, and earlier age at first drink compared to heterosexual youth (Talley et al. 2014). One meta-analysis reported that the odds for using substances, including alcohol, were 190% higher for SMY than heterosexual youth, with bisexuals experiencing significantly higher disparities (Marshal et al. 2008). Additionally, while alcohol use is decreasing among US high school students overall, Phillips et al. discovered that most alcohol use behaviors among sexual minority males have not decreased to a statistically significant degree since 2007 and that alcohol use remained elevated in sexual minority females, possibly due to continued, pervasive individual, interpersonal, and structural level discrimination (Phillips II et al. 2019b).

Prior research has explored these disparities using the minority stress framework. Minority stress posits that stigma, prejudice and discrimination at multiple socioecological levels enable environments that are persistently hostile for those who hold minority identities, and that this environment subsequently leads to mental health issues and chronic activation of stress processes that incentivize risk behaviors as coping mechanisms, including alcohol use (Meyer 2003). In SMY, empirical studies support the minority stress theory (Lee et al. 2016; Mereish et al. 2017ax).

Sexual orientation is often conceptualized along three dimensions: attraction, behavior and identity (Patterson et al. 2017). While the investigation in this area has been limited, several prior studies have demonstrated the importance of considering multiple dimensions of sexual orientation when analyzing health disparities, as different measures of sexual orientation can alter their magnitude. For example, one study of the Massachusetts YRBS tested six different operational definitions of ‘sexual minority status’ and found that the robustness of disparities among SMY varied greatly depending on how sexual orientation was operationalized (Matthews et al. 2014). A different study found that all dimensions of sexual minority orientation (identity, attraction, behavior) were associated with increased risk for substance use, although to different extents for each dimension (McCabe et al. 2005).

Despite increasing evidence, limited research has examined the impact of operationalizing multiple dimensions of sexual orientation on health disparities among youth. Expanding this research is vital, as sexual identity and sexual behavior, especially among adolescents, are complex and multilayered constructs, and exploration and experimentation are normal parts of the process of sexual development for many youths (Mustanski et al. 2014). It is, therefore, common for youth to report a sexual identity which may not perfectly align with stereotypical expectations of one’s sexual behavior—for example, reporting a gay identity, while also reporting sex with both males and females (Phillips II et al. 2019a) may reflect the normal diversity of individual sexual experience. However, an individual may be stigmatized if their sexual behavior does not perfectly align with social expectations, therefore creating distress and possibly leading to increased levels of alcohol consumption. As such, sexual identity, sexual behavior and their inter intersection may all impact the magnitude of alcohol use disparities among SMY, though these patterns are not well understood among youth.

Thus, given the possible differential impact of different modalities of measurement on estimates of health disparities, it is essential to know how different dimensions of sexual orientation operate together to influence alcohol use patterns in order to understand who is most at risk and develop policy and interventions accordingly. While some previous research has characterized disparities in alcohol use by sexual identity and sexual behavior in SMY, no research has comprehensively compared the two using nationally representative data, nor has it considered how the interactions between identity and behavior may impact alcohol use. In order to address this paucity of research, this study explored how approaches to measuring sexual orientation may impact alcohol use disparities in US SMY, as well as how the interactions between identity and behavior associated with alcohol use. By leveraging a large pooled sample of adolescents from the YRBS, we aim to elucidate the methodological importance of multiple measures of sexual orientation, and their interactions, to demonstrate how different operationalizations of these measures change the magnitude of alcohol disparities in sexually active SMY and make clear recommendations for future research as well as future interventions in this vulnerable population.

METHODS

Participants

The YRBS is a biennial national survey conducted by the Centers for Disease Control and Prevention (CDC) since 1991 to collect health data on students in grades 9–12 (Centers for Disease Control and Prevention et al. 2013). In this study, we pooled data from local versions of the YRBS across states and large urban school districts. Jurisdictions used a two-stage cluster sample design to identify a representative sample of students (Centers for Disease Control and Prevention et al. 2013). In the first stage, schools were selected with a probability proportional to their enrollment. In the second stage, classes of a required subject or during a required period were randomly selected; all students within these classes were eligible to participate. Participation was voluntary and anonymous. A new sample was selected each year that the survey was administered; the same students were not followed over time. For a complete description of the sampling procedure, see Kann et al. (2018). This study was deemed exempt by Northwestern University’s institutional review board (IRB); individual jurisdictions are responsible for securing IRB approval for their data collection.

Analytic sample

Pooled YRBS data across multiple jurisdictions and years (biennially from 2009 to 2017) comprised the analytic sample. The entire dataset consisted of 524,473 youth. Only jurisdiction-years that assessed both sexual identity and sexual behavior were included (125 jurisdiction-years). In order to facilitate comparisons in alcohol use by sexual identity and sexual behavior, youth who reported never having had sexual contact were excluded (N = 183,082), leaving a sample of 207,472 youth. Students were further excluded if they were missing any primary demographic variables (sex: 0.81%; race/ethnicity: 3.27%; age: 0.36%, not mutually exclusive), resulting in a final sample of 201,671 students.

Measures

All measures included in these analyses were identical across all jurisdiction years unless otherwise noted.

Demographics

Sex. Sex was determined by asking participants ‘What is your sex?’ with the choices: (a) male or (b) female. The YRBS did not ask about gender identity until 2017. Therefore, we will refer only to ‘male’ or ‘female’ participants.

Race/ethnicity. Race/ethnicity was assessed by asking participants if they identified as Hispanic or Latino (yes or no) and to select all of the races that applied from the following list: American Indian or Alaska Native; Asian; Black or African American; Native Hawaiian or Other Pacific Islander; and White. Using the CDC’s classification, these variables were combined into 5 racial/ethnic groups: (a) Asian; (b) Black or African American; (c) Hispanic/Latino (regardless of reported race); (d) White; and (e) Other.

Age. Participants were asked, ‘How old are you?’ Response options included: 12 years old or younger, 13, 14, 15, 16, 17 and 18 years old or older. Students who selected the first three options were collapsed into ‘14 years old or younger.’

Sexual identity and sexual behavior

Sexual identity. Sexual identity was assessed by asking, ‘Which of the following best describes you?’ Response options included: (a) heterosexual (straight); (b) gay or lesbian; (c) bisexual; and (d) not sure.

Sexual behavior. Sexual behavior was assessed by asking, ‘During your life, with whom have you had sexual contact?’ Response options included: (a) I have never had sexual contact; (b) females; (c) males; and (d) females and males. This variable was recorded based on the students’ reported sex. Students who reported no sexual contact were excluded from the analytic sample.

Alcohol use behaviors

Alcohol use was assessed across three domains, including age at first drink, current alcohol use and binge drinking. Additionally, results for lifetime alcohol use are included in Supplementary Table S1.

Age at first drink

This variable was assessed by the question, ‘How old were you when you had your first drink of alcohol other than a few sips?’ The responses were collapsed and dichotomized as ‘Under 13 years old’ and ‘At least 13 years old.’ Participants who did not report alcohol use were excluded from analyses for this outcome.

Current alcohol use

Participants were asked, ‘During the past 30 days, on how many days did you have at least one drink of alcohol?’ Responses were collapsed and dichotomized as ‘0 days’ or ‘1 or more day.’

Binge drinking

Binge drinking was assessed by the question, ‘During the past 30 days, on how many days did you have 5 or more drinks of alcohol in a row, that is, within a couple of hours?’ Responses were collapsed and dichotomized as ‘0 days’ or ‘1 or more days.’

Statistical analysis

All data cleaning and recoding were conducted in SAS Version 9.4 (SAS Institute, Cary, NC). Analyses were carried out using SAS-Callable SUDAAN Version 11.0.1 (RTI International, Research Triangle Park, NC) to appropriately weight estimates and to account for the complex sampling design of the YRBS. The YRBS data weights adjust for student non-response and distribution of students by age, sex and race/ethnicity in each jurisdiction (Centers for Disease Control and Prevention et al. 2013).

We first conducted descriptive analyses to determine the prevalence of sexual identity, sexual behavior and other demographics among the entire sample. Multivariable logistic regression models were developed to look at the associations between sexual identity/sexual behavior and each alcohol use outcome. The first model only included sexual identity, the second model only included sexual behavior and the third model included sexual identity and sexual behavior and an interaction between the two. All logistic regression models controlled for age, race/ethnicity, region, and survey year, stratified by sex.

RESULTS

Demographics

The pooled sample was nearly equally split between male and female youth (Table 1). More than 14% of the sample identified as something other than heterosexual—8.4% bisexual, 2.9% gay/lesbian and 3.0% not sure. A similar proportion (13.5%) reported sex with a same-sex partner—8.7% had sex with both male and female partners, and 4.8% had sex only with same-sex partners.

Table 1.

Demographics of pooled jurisdictional YRBS data by Sex, 2009–2017

	Total		Female		Male
	#	%	#	%	#	%
Demographics	201,671
Sexual identity
Heterosexual	171,158	85.66	78,952	79.39	92,206	91.49
Gay/Lesbian	5971	2.95	3144	2.93	2827	2.96
Bisexual	17,672	8.36	14,267	13.81	3405	3.29
Not sure	6870	3.03	3996	3.87	2874	2.25
Sexual behavior
Never had sexual contact	–	–	–	–	–	–
Same sex only	10,635	4.89	6196	5.64	4439	4.19
Male and female	18,229	8.69	14,184	13.90	4045	3.84
Different sex only	172,807	86.42	79,979	80.46	92,828	91.97
Sex
Female	100,359	48.22	–	–	–	–
Male	101,312	51.78	–	–	–	–
Race/ethnicity
White	83,977	46.37	42,867	48.48	41,110	44.41
Black or African American	33,211	16.47	15,824	15.74	17,387	17.15
Hispanic/Latino	57,601	29.71	28,458	28.41	29,144	30.92
Asian	8190	2.51	3961	2.41	4229	2.60
Other	18,691	4.94	9249	4.96	9442	4.92
Age
14 or younger	17,725	6.78	8791	6.81	8934	6.75
15	42,433	19.47	21,400	19.07	21,033	19.85
16	54,569	25.95	27,602	26.60	26,967	25.34
17	57,563	29.08	29,192	29.87	28,371	28.34
18 or older	29,381	18.72	13,374	17.63	16,007	19.72
Region
Northeast	95,477	25.64	47,713	26.43	47,764	24.90
Midwest	25,637	21.32	12,947	21.73	12,690	20.94
South	39,281	26.94	19,387	26.63	19,894	27.23
West	35,231	26.10	17,320	25.21	17,911	26.92
Alcohol use
Age of first Drink	144,361		75,392		68,969
Under 13 years old	43,609	28.65	19,313	23.95	24,296	33.62
At least 13 years old	100,752	71.35	56,079	76.05	44,673	66.38
Current Alcohol Use	181,549		91,022		90,527
Yes	81,554	45.89	43,278	49.40	38,276	42.59
No	99,995	54.11	47,744	50.60	52,251	57.41
Binge Drinking	179,485		89,750		89,735
Yes	42,949	24.40	21,370	24.63	21,579	24.18
No	136,536	75.60	68,380	75.37	68,156	75.82

Current alcohol use

Nearly half of the students reported recent alcohol use—49.4% of female youth and 42.6% of male youth. In adjusted models for males, sexual identity was not associated with current drinking (Model B1; Table 2). However, males who reported sex with other males were significantly more likely to have drunk alcohol in the last 30 days (aOR = 1.97; 95% CI: 1.58, 2.47; Model B2). In addition to a significant interaction between identity and behavior, Model B3 showed that males who identified as not sure were significantly more likely to report current alcohol use (aOR = 0.58; 95% CI:1.31, 2.90). Upon decomposing the interaction, significant intersections were seen for not sure^*different (aOR = 0.58), heterosexual^*both (aOR = 1.95), gay^*both (aOR = 2.15), bisexual^*both (aOR = 1.77), not sure^*both (aOR = 2.48), and not sure^*same (aOR = 1.58).

Table 2.

Multivariable logistic regression models: current drinking by sexual identity and behavior, YRBS 2009–2017.

Male N = 87,962	Model B1—SEXID + covariates			Model B2—SEXUAL BEHAVIOR + covariates			Model B3—INTERACTION + covariates
	P-value	OR	CI	P -value	OR	CI	P -value	OR	CI
Sexual Identity
Heterosexual		1.00						1.00
Gay	0.7514	1.05	(0.78,1.41)				0.4843	1.21	(0.70,2.09)
Bisexual	0.1284	1.23	(0.94,1.61)				0.3447	0.80	(0.51,1.27)
Not sure	0.8436	0.98	(0.76,1.25)				0.0011	0.58	(0.42,0.80)
Sexual behavior
Same				0.6840	1.05	(0.82,1.36)	0.1966	1.24	(0.90,1.70)
Both				<0.0001	1.97	(1.58,2.47)	0.0011	1.95	(1.31,2.90)
Different					1.00			1.00
Interaction							0.0250*
Covariates
Age
14 or younger	<0.0001	0.49	(0.39,0.60)	<0.0001	0.48	(0.39,0.60)	<0.0001	0.48	(0.38,0.59)
15	<0.0001	0.52	(0.43,0.61)	<0.0001	0.51	(0.43,0.61)	<0.0001	0.51	(0.43,0.60)
16	<0.0001	0.67	(0.57,0.78)	<0.0001	0.67	(0.57,0.78)	<0.0001	0.66	(0.57,0.77)
17	<0.0001	0.81	(0.73,0.90)	0.0001	0.81	(0.73,0.90)	<0.0001	0.81	(0.73,0.89)
18 or older		1.00			1.00			1.00
Race/ethnicity
White		1.00			1.00			1.00
Black	<0.0001	0.38	(0.34,0.42)	<0.0001	0.38	(0.34,0.42)	<0.0001	0.37	(0.34,0.42)
Hispanic	0.0001	0.70	(0.59,0.84)	0.0001	0.70	(0.58,0.84)	0.0001	0.70	(0.59,0.83)
Asian	0.0008	0.64	(0.49,0.83)	0.0009	0.64	(0.49,0.83)	0.0009	0.64	(0.49,0.83)
Other	0.2523	0.90	(0.74,1.08)	0.2412	0.89	(0.74,1.08)	0.2516	0.90	(0.74,1.08)
Region
Northeast		1.00			1.00			1.00
Midwest	0.2486	0.93	(0.83,1.05)	0.2661	0.94	(0.83,1.05)	0.2730	0.94	(0.84,1.05)
South	0.7586	0.98	(0.89,1.09)	0.7982	0.99	(0.89,1.10)	0.7747	0.98	(0.89,1.09)
West	0.8868	0.98	(0.77,1.26)	0.9159	0.99	(0.77,1.26)	0.9101	0.99	(0.77,1.26)
Year
2009	<0.0001	1.67	(1.45,1.92)	<0.0001	1.67	(1.45,1.93)	<0.0001	1.66	(1.44,1.91)
2011	<0.0001	1.55	(1.36,1.77)	<0.0001	1.55	(1.36,1.77)	<0.0001	1.54	(1.35,1.76)
2013	0.0003	1.24	(1.10,1.40)	0.0004	1.24	(1.10,1.40)	0.0004	1.24	(1.10,1.39)
2015	0.3147	1.11	(0.91,1.36)	0.3133	1.11	(0.91,1.36)	0.3061	1.11	(0.91,1.36)
2017		1.00			1.00			1.00
Female N = 88555	Model B4—SEXID + covariates			Model B5—SEXUAL BEHAVIOR + covariates			Model B6—INTERACTION + covariates
	P-value	OR	CI	P -value	OR	CI	P -value	OR	CI
Sexual identity
Heterosexual		1.00						1.00
Lesbian	0.4375	1.08	(0.89,1.32)				0.2660	0.66	(0.31,1.38)
Bisexual	0.0001	1.26	(1.12,1.42)				0.6825	0.96	(0.79,1.17)
Not sure	0.3938	1.10	(0.89,1.36)				0.7848	0.96	(0.69,1.32)
Sexual behavior
Same				0.0003	0.73	(0.62,0.87)	0.0039	0.65	(0.48,0.87)
Both				<0.0001	1.63	(1.45,1.84)	<0.0001	1.51	(1.24,1.83)
Different					1.00			1.00
Interaction							0.0714
Covariates
Age
14 or younger	0.0477	0.80	(0.65,1.00)	0.0362	0.79	(0.64,0.99)	0.0399	0.80	(0.64,0.99)
15	0.0021	0.80	(0.69,0.92)	0.0028	0.80	(0.70,0.93)	0.0027	0.81	(0.70,0.93)
16	0.0227	0.84	(0.73,0.98)	0.0258	0.84	(0.73,0.98)	0.0238	0.84	(0.73,0.98)
17	0.3239	0.94	(0.83,1.06)	0.3206	0.94	(0.83,1.06)	0.3206	0.94	(0.83,1.06)
18 or older		1.00			1.00			1.00
Race/ethnicity
White		1.00			1.00			1.00
Black	<0.0001	0.55	(0.50,0.60)	<0.0001	0.55	(0.50,0.60)	<0.0001	0.55	(0.50,0.60)
Hispanic	0.0763	0.90	(0.79,1.01)	0.0818	0.89	(0.80,1.01)	0.0758	0.90	(0.80,1.01)
Asian	<0.0001	0.37	(0.26,0.53)	<0.0001	0.37	(0.26,0.53)	<0.0001	0.37	(0.26,0.54)
Other	<0.0001	0.72	(0.62,0.82)	<0.0001	0.71	(0.61,0.80)	<0.0001	0.70	(0.61,0.81)
Region
Northeast		1.00			1.00			1.00
Midwest	0.0184	0.88	(0.79,0.98)	0.0159	0.88	(0.79,0.98)	0.0162	0.88	(0.79,0.98)
South	0.0004	0.85	(0.77,0.93)	0.0002	0.85	(0.77,0.92)	0.0001	0.84	(0.76,0.92)
West	0.9725	1.00	(0.83,1.20)	0.9879	1.00	(0.83,1.21)	0.9655	1.00	(0.83,1.20)
Year
2009	0.0004	1.32	(1.13,1.54)	0.0005	1.32	(1.13,1.53)	0.0005	1.31	(1.13,1.53)
2011	0.0012	1.21	(1.08,1.36)	0.0017	1.22	(1.07,1.36)	0.0017	1.20	(1.07,1.35)
2013	0.2314	1.06	(0.96,1.17)	0.3120	1.06	(0.95,1.16)	0.3204	1.05	(0.96,1.16)
2015	0.7305	0.97	(0.84,1.13)	0.7303	0.98	(0.84,1.13)	0.7224	0.97	(0.84,1.13)
2017		1.00			1.00			1.00
*Decomposition of interaction between sexual identity and sexual behavior (males)
	Heterosexual		Gay/Lesbian		Bisexual		Not sure
	OR	CI	OR	CI	OR	CI	OR	CI
Sex with different sex partners only	1.00		1.21	(0.70, 2.09)	0.80	(0.51, 1.27)	0.58	(0.42, 0.80)
Sex with both different and same sex partners	1.95	(1.31, 2.90)	2.14	(1.24, 3.70)	1.77	(1.13, 2.81)	2.48	(1.80, 3.42)
Sex with same sex partners only	1.24	(0.90, 1.70)	0.93	(0.54, 1.61)	0.91	(0.58, 1.45)	1.58	(1.14, 2.18)

For female youth, those who identified as bisexual were significantly more likely to report current alcohol use (aOR = 1.29; 95% CI 1.12, 1.42; Model B4). By sexual behavior, females who reported having sex with both males and females were significant more likely to be currently using alcohol (aOR = 1.63; 95% CI: 1.45, 1.84; Model B5). Conversely, females reporting only having sex with females were significantly less likely to be current alcohol drinkers (aOR = 0.73; 95% CI: 0.62, 0.87; Model B5). The interaction between identity and behavior was insignificant (P = 0.07; Model B6) and neither identity nor behavior remained significantly associated with current alcohol use when these variables were included in the model.

Age at first drink

Approximately one-quarter of female students (24%) and one-third of male students (33.6%) first drank alcohol before age 13. All males who reported a sexual minority identity were significantly more likely than their heterosexual peers to have had their first drink prior to age 13 (Table 3; Model C1). Similarly, all males who reported sex with other males were significantly more likely to have begun drinking before age 13 (Model C2). When the sexual identity^*sexual behavior interaction term was added to the model, the odds of gay male drinking before the age of 13 nearly tripled; additionally, the interaction term was significant (P < 0.0001; Model C3, Table 3).

Table 3.

Multivariable logistic regression models: drinking age under 13 by sexual identity and behavior, YRBS 2009–2017

Male N = 67176	Model C1—SEXID + covariates			Model C2—SEXUAL BEHAVIOR + covariates			Model C3—INTERACTION + covariates
	P-value	OR	CI	P-value	OR	CI	P-value	OR	CI
Sexual identity
Heterosexual		1.00						1.00
Gay	0.0225	1.41	(1.05,1.89)				<0.0001	3.98	(2.39,6.63)
Bisexual	0.0003	1.53	(1.21,1.92)				0.0006	1.98	(1.34,2.92)
Not sure	<0.0001	2.70	(2.07,3.53)				<0.0001	2.46	(1.63,3.71)
Sexual behavior
Same				0.0038	1.41	(1.12,1.78)	<0.0001	2.04	(1.49,2.78)
Both				<0.0001	2.00	(1.54,2.60)	0.0013	2.72	(1.48,5.02)
Different					1.00			1.00
Interaction							<0.0001
Covariates
Age
14 or younger	<0.0001	4.39	(3.40,5.66)	<0.0001	4.41	(3.42,5.68)	<0.0001	4.39	(3.39,5.69)
15	<0.0001	2.04	(1.71,2.45)	<0.0001	2.04	(1.70,2.45)	<0.0001	2.02	(1.69,2.43)
16	0.0041	1.27	(1.08,1.49)	0.0040	1.27	(1.08,1.49)	0.0047	1.26	(1.07,1.49)
17	0.7897	1.02	(0.88,1.19)	0.7749	1.02	(0.88,1.19)	0.8407	1.02	(0.87,1.18)
18 or older		1.00			1.00			1.00
Race/ethnicity
White		1.00			1.00			1.00
Black	<0.0001	1.48	(1.32,1.67)	<0.0001	1.48	(1.32,1.67)	<0.0001	1.48	(1.31,1.66)
Hispanic	0.0010	1.33	(1.12,1.58)	0.0006	1.35	(1.14,1.60)	0.0010	1.34	(1.13,1.59)
Asian	0.8606	1.03	(0.73,1.46)	0.8203	1.04	(0.74,1.46)	0.9723	1.01	(0.71,1.42)
Other	0.0055	1.30	(1.08,1.55)	0.0042	1.30	(1.09,1.56)	0.0043	1.30	(1.09,1.57)
Region
Northeast		1.00			1.00			1.00
Midwest	0.0031	1.22	(1.07,1.39)	0.0022	1.23	(1.08,1.40)	0.0022	1.23	(1.08,1.40)
South	<0.0001	1.38	(1.22,1.55)	<0.0001	1.38	(1.23,1.56)	<0.0001	1.39	(1.23,1.57)
West	0.0404	1.28	(1.01,1.62)	0.0499	1.26	(1.00,1.60)	0.0324	1.29	(1.02,1.64)
Year
2009	0.0319	1.20	(1.02,1.43)	0.0423	1.19	(1.01,1.41)	0.0253	1.21	(1.02,1.43)
2011	0.3255	1.07	(0.93,1.23)	0.4047	1.06	(0.92,1.21)	0.3295	1.07	(0.93,1.23)
2013	0.0239	0.87	(0.77,0.98)	0.0222	0.87	(0.77,0.98)	0.0233	0.87	(0.77,0.98)
2015	0.5831	0.95	(0.81,1.13)	0.5770	0.95	(0.81,1.13)	0.5882	0.95	(0.81,1.13)
2017		1.00			1.00			1.00
Female N = 73256	Model C4—SEXID + covariates			Model C5—SEXUAL BEHAVIOR + covariates			Model C6—INTERACTION + covariates
	P-value	OR	CI	P -value	OR	CI	P -value	OR	CI
Sexual identity
Heterosexual		1.00						1.00
Lesbian	<0.0001	2.34	(1.79,3.06)				0.0730	1.90	(0.94,3.83)
Bisexual	<0.0001	1.99	(1.72,2.31)				0.0001	1.67	(1.30,2.15)
Not sure	<0.0001	2.07	(1.67,2.56)				0.0052	1.55	(1.14,2.11)
Sexual behavior
Same				<0.0001	2.54	(2.05,3.15)	<0.0001	2.77	(2.09,3.66)
Both				<0.0001	2.29	(2.01,2.62)	<0.0001	2.13	(1.68,2.69)
Different					1.00			1.00
Interaction							0.0106*
Covariates
Age
14 or younger	<0.0001	4.63	(3.71,5.79)	<0.0001	4.62	(3.69,5.79)	<0.0001	4.54	(3.64,5.67)
15	<0.0001	2.27	(1.89,2.73)	<0.0001	2.27	(1.89,2.73)	<0.0001	2.26	(1.88,2.71)
16	0.0029	1.31	(1.10,1.56)	0.0024	1.32	(1.10,1.58)	0..0025	1.32	(1.10,1.57)
17	0.6947	1.04	(0.85,1.27)	0.7079	1.04	(0.85,1.27)	0.7285	1.04	(0.85,1.27)
18 or older		1.00			1.00			1.00
Race/ethnicity
White		1.00			1.00			1.00
Black	<0.0001	1.55	(1.37,1.75)	<0.0001	1.54	(1.37,1.74)	<0.0001	1.51	(1.34,1.70)
Hispanic	<0.0001	1.40	(1.22,1.61)	<0.0001	1.41	(1.23,1.62)	<0.0001	1.39	(1.20,1.60)
Asian	0.0734	1.35	(0.97,1.86)	0.0441	1.42	(1.01,1.99)	0.0509	1.39	(1.00,1.94)
Other	0.0088	1.41	(1.09,1.81)	0.0097	1.40	(1.08,1.81)	0.0180	1.36	(1.05,1.76)
Region
Northeast		1.00			1.00			1.00
Midwest	0.0156	1.16	(1.03,1.31)	0.0155	1.16	(1.03,1.31)	0.0129	1.17	(1.03,1.32)
South	0.0214	1.13	(1.02,1.25)	0.0191	1.13	(1.02,1.25)	0.0132	1.14	(1.02,1.26)
West	0.0818	1.19	(0.98,1.46)	0.0616	1.20	(0.99,1.46)	0.0493	1.21	(1.00,1.47)
Year
2009	0.0003	1.32	(1.14,1.54)	0.0002	1.33	(1.14,1.54)	0.0001	1.36	(1.17,1.58)
2011	0.0622	1.14	(0.99,1.30)	0.0421	1.14	(1.00,1.30)	0.0259	1.16	(1.02,1.32)
2013	0.9859	1.00	(0.88,1.13)	0.8793	0.99	(0.88,1.12)	0.9349	1.01	(0.89,1.14)
2015	0.5497	0.95	(0.82,1.11)	0.6116	0.96	(0.83,1.12)	0.6350	0.96	(0.83,1.12)
2017		1.00			1.00			1.00
*Decomposition of interaction between identity and behavior (males)
	Heterosexual		Gay/Lesbian		Bisexual		Not sure
	OR	CI	OR	CI	OR	CI	OR	CI
Sex with different sex partners only	1.00		3.98	(2.39, 6.63)	1.98	(1.34, 2.92)	2.46	(1.63, 3.71)
Sex with both different and same sex partners	2.72	(1.48, 5.02)	2.55	(1.53, 4.24)	1.35	(0.91, 1.99)	3.27	(2.15, 4.90)
Sex with same Sex partners only	2.04	(1.49, 2.78)	0.96	(0.57, 1.59)	1.60	(1.09, 2.37)	3.12	(2.07, 4.71)
*Decomposition of interaction between identity and behavior (females)
	Heterosexual		Gay/Lesbian		Bisexual		Not Sure
	OR	CI	OR	CI	OR	CI	OR	CI
Sex with different sex partners pnly	1.00		3.98	(2.39, 6.63)	1.98	(1.34, 2.92)	2.46	(1.63, 3.71)
Sex with both different and same sex partners	2.72	(1.48, 5.02)	2.55	(1.53, 4.24)	1.35	(0.91, 1.99)	3.27	(2.15, 4.90)
Sex with same sex partners only	2.04	(1.49, 2.78)	0.96	(0.57, 1.59)	1.60	(1.09, 2.37)	3.12	(2.07, 4.71)

A similar pattern was seen among female youth with respect to initiating alcohol use prior to the age of 13 years: all sexual minority identities and behaviors were significantly associated with alcohol use before the age of 13 (Models C4 and C5). In Model C6, all identity and behavior associations aside from among lesbian women remained significant; the sexual behavior^*sexual identity interaction term was also significant (P = 0.0106). Interaction decompositions were shown in Table 3.

Binge drinking

Approximately one-quarter of both female students (25.9%) and male students (27.9%) reported at least one instance of binge drinking in the prior 30 days. Although there were no significant differences based on sexual identity (Table 4; Model D1), males who reported having sex with both females and males were significantly more likely to have engaged in binge drinking than those who only reported sex with females (aOR = 2.20; 95% CI: 1.70, 2.84; Model D2). In the interaction model, identifying as not sure became a protective factor for binge drinking (aOR = 0.66; 95% CI: 0.46, 0.94; Model D3), and the interaction between identity and behavior was significant (P = 0.025). The interaction decomposition revealed significant intersections (Table 4).

Table 4.

Multivariable logistic regression models: binge drinking by sexual identity and behavior, YRBS 2009–2017

Male N = 86,953	Model D1—SEXID + covariates			Model D2—SEXUAL BEHAVIOR + covariates			Model D3—INTERACTION + covariates
	P-value	OR	CI	P-value	OR	CI	P-value	OR	CI
Sexual identity
Heterosexual		1.00						1.00
Gay	0.3659	0.85	(0.63,1.19)				0.1616	0.69	(0.41,1.16)
Bisexual	0.2610	1.19	(0.88,1.60)				0.1051	0.64	(0.38,1.10)
Not sure	0.0510	1.28	(1.00,1.65)				0.0227	0.66	(0.46,0.94)
Sexual behavior
Same				0.7674	0.96	(0.73,1.26)	0.3911	1.16	(0.83,1.61)
Both				<0.0001	2.20	(1.70,2.84)	0.0031	2.00	(1.26,3.15)
Different					1.00			1.00
Interaction							0.0248*
Covariates
Age
14 or younger	<0.0001	0.42	(0.33,0.55)	<0.0001	0.42	(0.32,0.54)	<0.0001	0.42	(0.32,0.54)
15	<0.0001	0.43	(0.35,0.54)	<0.0001	0.43	(0.35,0.53)	<0.0001	0.43	(0.35,0.53)
16	<0.0001	0.63	(0.53,0.74)	<0.0001	0.62	(0.53,0.74)	<0.0001	0.62	(0.52,0.73)
17	0.0016	0.80	(0.70,0.92)	0.0011	0.80	(0.70,0.92)	0.0010	0.80	(0.70,0.91)
18 or older		1.00			1.00			1.00
Race/ethnicity
White		1.00			1.00			1.00
Black	<0.0001	0.26	(0.23,0.30)	<0.0001	0.26	(0.23,0.30)	<0.0001	0.26	(0.23,0.30)
Hispanic	<0.0001	0.60	(0.47,0.75)	<0.0001	0.60	(0.46,0.74)	<0.0001	0.60	(0.48,0.74)
Asian	<0.0001	0.51	(0.37,0.70)	<0.0001	0.51	(0.37,0.70)	<0.0001	0.51	(0.37,0.70)
Other	0.0412	0.80	(0.65,0.99)	0.0340	0.80	(0.65,0.98)	0.0354	0.80	(0.65,0.98)
Region
Northeast		1.00			1.00			1.00
Midwest	0.6610	1.03	(0.90,1.18)	0.6263	1.03	(0.90,1.19)	0.6058	1.04	(0.90,1.19)
South	0.2484	1.07	(0.96,1.19)	0.2263	1.07	(0.96,1.20)	0.2546	1.07	(0.95,1.19)
West	0.2488	1.17	(0.89,1.54)	0.2366	1.18	(0.90,1.54)	0.2320	1.18	(0.90,1.54)
Year
2009	<0.0001	2.29	(1.93,2.72)	<0.0001	2.30	(1.94,2.73)	<0.0001	2.28	(1.92,2.70)
2011	<0.0001	2.08	(1.76,2.46)	<0.0001	2.08	(1.76,2.46)	<0.0001	2.06	(1.74,2.44)
2013	<0.0001	1.52	(1.32,1.76)	<0.0001	1.53	(1.32,1.77)	<0.0001	1.52	(1.31,1.75)
2015	0.0019	1.41	(1.13,1.74)	0.0018	1.41	(1.14,1.75)	0.0017	1.41	(1.14,1.74)
2017		1.00			1.00			1.00
Female N = 87,072	Model D4—SEXID + covariates			Model D5—SEXUAL BEHAVIOR + covariates			Model D6—INTERACTION + covariates
	P-value	OR	CI	P-value	OR	CI	P-value	OR	CI
Sexual identity
Heterosexual		1.00						1.00
Lesbian	0.6680	0.95	(0.77,1.18)				0.9773	0.99	(0.46,2.12)
Bisexual	0.0260	1.18	(1.02,1.36)				0.4635	0.90	(0.67,1.20)
Not sure	0.4590	1.08	(0.88,1.34)				0.2529	0.82	(0.57,1.16)
Sexual behavior
Same				0.2095	0.86	(0.69,1.09)	0.8213	0.96	(0.69,1.34)
Both				<0.0001	1.57	(1.40,1.77)	<0.0001	1.61	(1.31,1.99)
Different					1.00			1.00
Interaction							0.7951
Covariates
Age
14 or younger	0.0011	0.72	(0.59,0.88)	0.0004	0.70	(0.58,0.86)	0.0005	0.71	(0.58,0.86)
15	0.0013	0.75	(0.63,0.89)	0.0011	0.50	(0.63,0.89)	0.0011	0.75	(0.63,0.89)
16	0.0064	0.81	(0.70,0.94)	0.0056	0.81	(0.69,0.94)	0.0060	0.81	(0.70,0.94)
17	0.0441	0.88	(0.78,1.00)	0.0344	0.87	(0.77,0.99)	0.0352	0.88	(0.77,0.99)
18 or older		1.00			1.00			1.00
Race/ethnicity
White		1.00			1.00			1.00
Black	<0.0001	0.37	(0.32,0.42)	<0.0001	0.36	(0.32,0.41)	<0.0001	0.36	(0.32,0.41)
Hispanic	0.0148	0.84	(0.73,0.97)	0.0116	0.84	(0.73,0.96)	0.0120	0.84	(0.73,0.96)
Asian	<0.0001	0.32	(0.21,0.49)	<0.0001	0.33	(0.21,0.50)	<0.0001	0.32	(0.21,0.50)
Other	0.0066	0.75	(0.61,0.92)	0.0020	0.73	(0.60,0.89)	0.0022	0.73	(0.60,0.89)
Region
Northeast		1.00			1.00			1.00
Midwest	0.9910	1.00	(0.88,1.13)	0.9887	1.00	(0.88,1.13)	0.9661	1.00	(0.88,1.13)
South	0.1069	0.91	(0.81,1.02)	0.0787	0.90	(0.80,1.01)	0.0762	0.90	(0.80,1.01)
West	0.5243	1.07	(0.87,1.32)	0.4886	1.08	(0.87,1.33)	0.4719	1.08	(0.88,1.33)
Year
2009	<0.0001	1.53	(1.29,1.81)	<0.0001	1.53	(1.30,1.81)	<0.0001	1.52	(1.29,1.80)
2011	<0.0001	1.36	(1.19,1.56)	<0.0001	1.36	(1.18,1.56)	<0.0001	1.35	(1.18,1.56)
2013	0.0522	1.13	(1.00,1.28)	0.0611	1.12	(0.99,1.27)	0.0686	1.12	(0.99,1.26)
2015	0.2379	1.10	(0.94,1.29)	0.2315	1.10	(0.94,1.29)	0.2370	1.10	(0.94,1.29)
2017		1.00			1.00			1.00
*Decomposition of interaction between identity and behavior (males)
	Heterosexual		Gay/Lesbian		Bisexual		Not sure
	OR	CI	OR	CI	OR	CI	OR	CI
Sex with different sex partners only	1.00		0.69	(0.41, 1.16)	0.64	(0.38, 1.10)	0.66	(0.46, 0.94)
Sex with both different and same sex partners	2.00	(1.26, 3.15)	2.04	(1.20, 3.43)	1.88	(1.11, 2.20)	3.84	(2.68, 5.48)
Sex with same sex partners only	1.16	(0.83, 1.61)	0.79	(0.47, 1.32)	0.84	(0.49, 1.42)	1.89	(1.32, 2.69)

For female youth, bisexual females were significantly more likely to report binge drinking (aOR = 1.18; 95% CI: 1.02, 1.36; Model D4). Additionally, female youth who reported sex with both males and females were significantly more likely to report binge drinking than those who only had sex with males (aOR = 1.57; 95% CI: 1.40, 1.77; Model D5). Finally, in Model D6, there was not a significant interaction and the association between sexual identity and binge drinking became insignificant; significance remained for female youth who had sex with both males and females (aOR = 1.61; 95% CI: 1.31, 1.99; Model D6).

DISCUSSION

Our study examined two dimensions of sexual orientation and the ability to detect alcohol use disparities impacting sexually active SMY. Limited prior research has specifically broken down sexual orientation into multiple dimensions to examine the magnitude of alcohol use disparities detected by these different dimensions and their intersections among youth; previous research with the YRBS dataset using identity only has shown that SMY is more likely to engage in alcohol use behaviors than their sexual majority counterparts (Phillips II et al. 2019b). As such, this study confirms patterns of disparities demonstrated in previous work using YRBS data. However, this is the first study to comprehensively examine the differences in disparities in multiple alcohol use outcomes measured through different dimensions of sexual orientation, as well as the first to consider how the interactions between these dimensions impact alcohol consumption.

We found that using multiple dimensions of sexual orientation to analyze alcohol use disparities influenced the outcomes we observed among SMY, underscoring the importance of using multiple measures of sexual orientation to fully understand health disparities impacting SMY. Accounting for the complex and developmental nature of sexuality is crucial, not only in order to best capture disparities in alcohol use, but to shape policies and interventions that target alcohol use among vulnerable populations. Such intervention strategies, which are traditionally based solely on identity, could benefit from including multiple facets of sexuality to target all youth at risk of adverse alcohol use outcomes. Furthermore, our results indicate that diverging from traditionally exclusive or singular models of sexuality may also contribute to creating autonomous, safe spaces for SMY. Prior research has established that safe spaces such as gay-straight alliances in high schools can help decrease substance use among SMY (Heck et al. 2014); adapting these spaces to be intentionally inclusive of diverse sexual orientations, particularly for individuals whose behaviors and identities may not align to stereotypical expectations, may help to extend this protective effect to other vulnerable students.

Such strategies may be particularly important given that, as revealed by the decomposition analysis, youth whose behavior did not align with stereotypical beliefs regarding sexual identity and behavior alignment had higher odds of current drinking and having had their first drink before the age of 13 years. We hypothesize that this pattern may be due to a high degree of social pressure to have one’s behavior and identity be in alignment to conform to social expectations of the construction of sexuality (Hammack et al. 2009). Thus, youth whose behavior and identity are perceived as being ‘non-aligned’ may experience ostracization and shame, which could lead to increased alcohol use as an attempt to cope. Researchers, policymakers and teachers should take care not to reinforce that identity and behavior must be ‘aligned,’ but rather work to communicate the complex and multifaceted nature of sexuality. This is especially true during identity development; an adolescent’s sexual identity exists as a multidimensional continuum of experience; developmental trajectories among SMY are highly varied (Glover et al. 2010). As such, the complicated process of identity development should be seen as normative, rather than stigmatized, for maximum psychological benefit. More in-depth, qualitative research is needed to understand how the pressure to adhere to expectations of socially constructed sexualities may impact adolescents.

This study further corroborates prior work which indicates that bisexual populations have higher odds of alcohol use, even compared to other SMY, suggesting that minority stress disparately impacts this group (Boyd et al. 2019; Talley et al. 2019). This could be due to widespread biphobia (Bostwick and Hequembourg 2014), which causes bisexual individuals to experience the stigma of being a sexual minority while not receiving the full protective and resiliency-buffering factors that come with being fully accepted by a social support community (Dodge et al. 2016). Bisexual populations experience stress as a result of feeling ostracized by both heterosexual and sexual minority communities, and as a result, report feeling socially isolated and marginalized, leading to disparate alcohol use (Friedman et al. 2014). Addressing the pervasive and damaging nature of biphobia is likely to significantly improve the health of this vulnerable community (Mereish et al. 2017b).

Notably, this study extends existing research on biphobia, as our results demonstrate that not only reporting a bisexual identity but also having sex with both males and females regardless of one’s specific identity place youth at similar odds of binge drinking and alcohol use. The novel finding that sex with both males and females places youth at greater risk is deserving of additional research focus. There are multiple possible interpretations. For example, this may be a result of pervasive biphobia impacting individuals on the basis of behavior as well as identity; this result may also reflect the experiences of gay/lesbian youth reacting to sexual experiences which they felt pressured into as a result of pervasive homophobia, or a combination of these two interpretations. Thus, there is a need to more rigorously explore alcohol use risk factors and motivations among sexual minority subpopulations such as bisexuals and youth who report sex with both males and females. Overall, reducing stigma including biphobia, homophobia and lesbophobia toward SMY is likely to help diminish the burden of health disparities in these groups. The targeted study will help to illuminate the causal pathways of these forms of stigma to inform alcohol use interventions for SMY.

Notably, our analyses found that students who reported a ‘not sure’ sexual identity and reported sexual behavior with partners of the same sex had some of the highest odds of drinking across all outcomes. Although we are unable to directly assess cause, we hypothesize that this pattern may be connected to the protective nature of identity and community. Prior literature has found that identity centrality and identity commitment can serve as protective factors against negative health outcomes caused by minority stress and stigma among sexual minority populations, including biphobia (Fredrick 2015; Santos and VanDaalen 2016), and may lead to higher levels of resilience and self-esteem (Shramko et al. 2018). Though we were unable to directly test identity salience within our study, those with a not sure identity may have increased rates of substance use due to having lower identity centrality, even though being unsure of ones’ identity may be a normative part of identity development for some adolescents (Rosario et al. 2011). Additionally, though having access to sexual minority spaces has been shown to buffer substance use in SMY (Coulter et al. 2016), those with a not sure identity may lack access to these spaces. Therefore, having a not sure identity despite same-sex sexual contact may expose a student to minority stress while removing the social support of a community that can act as a buffer against discrimination and stigma for same-sex sexual behavior, leading to the elevated alcohol use seen in this study. In-depth qualitative research will be required to appropriately test this theory.

Our study is not without limitations. First, all data were self-reported. Reporting alcohol use presents the possibility of social desirability bias; this is especially heightened as, given their age, alcohol use among participants was illegal, meaning there is an even higher risk that drinking was underreported. However, with the anonymous administration of YRBS, the chance of this bias is reduced. Data were also cross-sectional, preventing causal analysis. Only two dimensions of sexual orientation (identity and behavior) were assessed. The exclusion of other measures, such as sexual attraction, as well as limited options of sexual identities to choose from, may have resulted in a potential underestimation of SMY. Relatedly, the phrasing of the YRBS’s ‘not sure’ sexual orientation option presents a challenge to interpretation, as one cannot be sure whether these students are reporting that they are questioning their identity, or that they are not sure of what the question is asking. Few jurisdictions assess gender identity as part of the YRBS, resulting in an inability to account for the role of gender identity in the analysis. As transgender status has been associated with alcohol risk similarly to sexual minority status, future analyses should include gender identity (Day et al. 2017). Minority stress stemming from discrimination on the basis of race/ethnicity is also an important factor contributing to adolescent substance use; while we minimally explored the effect of race by controlling for it in analyses, more in-depth, intersectional approaches should be taken to understand how race/ethnicity impacts alcohol use outcomes, especially in those youth holding multiple marginalized identities (Bowleg 2012). Finally, variance in the context of sexual minority support (including policy, geography and resources) and alcohol use practices by local jurisdictions may provide greater insight into the causal factors for disparities in alcohol use compared to a national pool of data alone that were used for our analysis (Midanik et al. 2007). Future research should assess how context impacts youth alcohol use.

Often, research and intervention programs focus on sexual identity while failing to consider sexual behavior, which can ultimately exclude vulnerable populations who could benefit from targeted intervention. In addition to the continued exploration of drivers of alcohol use among SMY by multiple dimensions of sexuality, serious steps must be undertaken to reduce disparities in alcohol use among this population. Specifically, organizations should provide safe, supportive spaces for youth as they undergo the processes of sexual identity development. Programs should be tailored to the unique needs of diverse subpopulations of SMY, keeping in mind the multifaceted and developmental nature of sexual identity and behavior, so as to help them find healthy coping strategies as an alternative to early onset of alcohol uptake, with particular attention to bisexual and not sure youth, and to youth whose sexual identity and sexual behavior may not conform to stereotypical expectations of ‘alignment.’

Funding

This study was supported by a grant from the National Institute on Alcohol Abuse and Alcoholism (R01 AA024409, Principal Investigator: Phillips) as well as a grant from the National Heart, Lung, and Blood Institute (K12 HL143959; Principal Investigator: Beach). The study sponsors had no role in the creation of this manuscript.

Conflict of interest statement

None declared.