Multiple Diverse Drinking Trajectories among Sexual Minority Women: Unique and Joint Prediction by Minority Stress and Social Influence Risk Factors

Abstract

Background:

Sexual minority populations are at heightened risk for alcohol use disorders compared to heterosexual populations, and these disparities are particularly pronounced for sexual minority women (SMW). Little research has examined the diversity of drinking trajectories among sexual minorities, despite evidence that such trajectories have high predictive utility and are useful in understanding how risk factors may be uniquely associated with specific trajectories.

Method:

We utilized four waves of data (12 months between waves) from a sample of 1,057 SMW ages 18–25 at Wave 1. The goals were to (a) identify multiple distinct trajectories of alcohol use; (b) examine the predictive utility of these trajectories; and (c) test associations between minority stress (e.g., discrimination) and social influence (e.g., sexual minority community involvement) risk factors and alcohol trajectories.

Results:

Using growth mixture modeling, we identified five classes based on drinking patterns at Wave 1 and change over time (stable low, stable high drinking, stable high HED, low increasing, and high decreasing). These classes were differentially associated with Wave 1 levels and changes in alcohol consequences. Minority stressors uniquely predicted a low increasing trajectory, while social influences uniquely predicted a stable high trajectory. Both minority stress and social risk factors predicted high decreasing and stable high HED trajectories.

Conclusions:

Findings indicate that some drinking trajectories among SMW appear similar to those found in the general population, while others appear unique. Results provide insight into how minority stress and social influence risk factors may uniquely and jointly contribute to disparities affecting this population.

1. Introduction

Sexual minorities are at heightened risk for alcohol use disorders compared to heterosexuals, and this disparity is particularly pronounced for sexual minority women (SMW; Kerridge et al., 2017). Numerous cross-sectional (see Kidd et al., 2018) and a handful of longitudinal studies (Dyar et al., 2020; Litt et al., 2015; Wilson et al., 2016) have linked factors proposed to explain these disparities (e.g., discrimination) with alcohol use among sexual minorities. However, little research has examined the diversity of drinking trajectories in this population by using growth mixture modeling (GMM). In the general population, GMM has identified drinking trajectories associated with long-term substance use outcomes (Chassin et al., 2002; Danielsson et al., 2010; Iwamoto et al., 2018). Given their predictive utility, further research is necessary to advance our understanding of risk factors for problematic drinking trajectories in this high-risk population.

1.1. Drinking Trajectories

Most alcohol use research has utilized variable-centered approaches (e.g., regression), which may obscure within-group variability in drinking trajectories (Muthén, 2006). Person-centered approaches (e.g., GMM) allow for identification of heterogeneous drinking trajectories and risk factors unique to specific trajectories (Maggs & Schulenberg, 2004). In the general population, person-centered approaches have identified drinking trajectories differentially associated with substance use outcomes (Chassin et al., 2002; Danielsson et al., 2010; Iwamoto et al., 2018). In emerging adulthood, studies tend to identify three to six trajectories (Andrade & Järvinen, 2020; Iwamoto et al., 2018; Schuckit et al., 2014). Stable heavy and stable low drinking trajectories are found consistently (Andrade & Järvinen, 2020; Iwamoto et al., 2018; Lemoine et al., 2020). Others are found less consistently: low/medium increasing drinking (low/medium drinking at baseline, increase over time) (Bountress et al., 2021; Leggat et al., 2021; Schuckit et al., 2014); high decreasing drinking (high drinking at baseline, decrease) (Bountress et al., 2021; Greenbaum et al., 2005; Schuckit et al., 2014); and moderate stable drinking (moderate drinking at baseline; no change)(Andrade & Järvinen, 2020; Iwamoto et al., 2018; Leggat et al., 2021). Stable heavy and low/medium increasing drinking trajectories have been linked to drinking consequences and alcohol use disorder (Chassin et al., 2002; Danielsson et al., 2010; Iwamoto et al., 2018).

Despite sexual minorities’ high risk, we are aware of only two studies to utilize GMM to examine drinking in this population. Marshall et al. (2015) identified five trajectories of heavy episodic drinking (HED) among sexual minority men: no HED; stable infrequent HED; stable moderate HED; moderate increasing HED; and stable high HED. While comparison of these results to those from the general population is difficult due to differences between samples (Marshall’s sample is older and excluded women), the trajectories identified in this study appear similar to those from the general population.

Among emerging adult women of all sexual orientations, Coulter et al. (2018) identified six profiles, including non-drinkers; stable low drinkers; moderate stable; low increasing; low with steep increase in drinking; and stable heavy drinking. These classes are largely consistent with those from previously discussed studies. Mostly heterosexual and bisexual women were more likely than heterosexual women to be in classes associated with alcohol use disorders, including stable heavy, low with steep increase, and moderate stable classes. However, there were only 287 lesbian and bisexual women compared to more than 6,000 heterosexual women in their sample. This may have resulted in trajectories more representative of drinking among heterosexual than SMW. Additionally, neither study tested associations between trajectories and factors theorized to explain sexual minorities’ risk for problematic drinking (Coulter et al., 2018; Marshall et al., 2015). Further research is necessary to expand our understanding of drinking trajectories and identify risk factors for heavy drinking trajectories in this population.

1.2. Potential Predictors of Drinking Trajectories among SMW

Two theories have been proposed to explain high rates of problematic drinking among sexual minorities. Minority stress theory posits that chronic stressors experienced by sexual minorities due to stigma (e.g., discrimination) contribute to drinking problems in this population (Meyer, 2003). Hatzenbuehler (2009) proposed that minority stressors deplete sexual minorities’ coping resources, thereby increasing coping motives for drinking and rates of problematic drinking. Several longitudinal studies have indicated that when sexual minorities experience more minority stress, they also experience concurrent increases in risky drinking and consequences (Dermody et al., 2014; Dyar et al., 2020). However, these studies provide mixed evidence of prospective associations between minority stressors and substance use, with some finding a prospective association (Dermody et al., 2016; Wilson et al., 2016) while others did not (Dyar et al., 2019; Dyar et al., 2020). This may suggest that minority stress is associated with different drinking trajectories, with minority stress predicting increases in drinking for some sexual minorities but not others. Determining which drinking trajectories are associated with minority stressors will provide nuance to our understanding of associations between minority stress and drinking.

Social learning theory posits that less restrictive drinking norms and use of alcohol-centric contexts for socialization in sexual minority communities contribute to higher alcohol consumption in this population (Condit et al., 2011). A handful of cross-sectional studies have linked involvement in sexual minority community activities with binge drinking (Baiocco et al., 2010) and problematic drinking (Feinstein et al., 2017). We are not aware of any research examining longitudinal associations between community involvement and alcohol use.

1.3. Current Study

The current study aimed to address gaps in the literature by: identifying diverse drinking trajectories; examining their predictive utility; and exploring risk factors for different drinking trajectories among SMW in emerging adulthood. We focused on SMW as they experience larger disparities in alcohol use than sexual minority men (Kerridge et al., 2017) and emerging adulthood because it is a critical period for the development of problematic drinking (Arnett, 2005; Johnston, 2010).

We utilized GMM to identify subgroups of SMW based on trajectories of drinking frequency and HED across four waves of data (12 months between waves). Although GMM is a data driven approach, we expected to find the following classes based on prior research:

1. stable low drinking [low drinking at Wave 1, no change over time]

2. stable high drinking [high drinking at Wave 1, no change]

3. low/medium increasing [low/medium drinking at Wave 1, increasing]

4. high decreasing [high drinking at Wave 1, decreasing]

5. moderate stable drinking [moderate drinking at Wave 1, no change]

Next, we aimed to examine the predictive utility of these trajectories. Specifically, we tested whether the intercept (alcohol consequences at Wave 1) and slope of alcohol consequences (change in consequences from Wave 1 to 4) differed across drinking trajectories. We expected to find that stable high, low/medium increasing, and moderate stable drinking groups would experience moderate increases in consequences; high decreasing drinking would experience decreases in consequences; and stable low drinking would have low, unchanging consequences.

Further, we aimed to test associations between potential predictors and drinking trajectories, including Wave 1 measures of minority stressors (discrimination, harassment), sexual minority community involvement, and drinking motives (conformity, social, enhancement, coping). We expected that higher minority stress, community involvement, and social, enhancement, and coping motives for drinking would be associated with higher risk drinking trajectories. We expected minority stress factors would predict different trajectories than those predicted by social influence factors but did not make specific hypotheses.

2. Methods

2.1. Participants and Procedures

The current manuscript utilized data from a longitudinal study of alcohol use among emerging adult SMW (Litt et al., 2015). Participants were recruited via online advertisements. Eligible participants were U.S. residents, 18–25 at recruitment, identified as lesbian or bisexual, and were assigned female at birth.¹ Participants were required to be female-assigned at birth due to the study’s focus on alcohol consumption, which has sex-dependent cutoffs for patterns of use. 1057 eligible participants provided informed consent and completed up to four annual assessments. Participants received $25 for the Wave 1 survey and $30 for each later assessment. Retention rates were 77%, 71%, and 70% for Waves 2–4.

Participants had an average age of 20.86 years (SD=2.08). The majority identified as bisexual (59.5%), with the remainder identifying as lesbian. Participants were mostly White (67.8%), with others identifying as Black/African American (10.0%), Asian American (2.6%), multiracial (15.6%), and another race/ethnicity (3.9%). Additionally, 11.3% of the sample identified as Latinx, with the remainder identifying as non-Latinx.

2.2. Measures

2.2.1. Alcohol consumption.

HED was assessed via the item “In the last 12 months, how often did you have 4 or more drinks of wine, beer, or liquor in a single day?” on a scale of 0 (never); 1 (1–3 times a year); 2 (4–7 times a year); 3 (8–11 times a year); 4 (1–3 times a month); 5 (once or twice a week); 6 (3–4 times a week); 7 (5 times a week or more). Drinking frequency was assessed via the Daily Drinking Questionnaire (Collins et al., 1985). Participants were provided a definition of a standard drink and asked to “Consider a typical week during the past 12 months. How much alcohol, on average, do you drink on each day of a typical week?” The number of days for which participants indicated drinking was used as a measure of drinking frequency.

2.2.2. Alcohol consequences

were assessed using the Young Adult Alcohol Consequences Questionnaire (Read et al., 2006). Participants indicated which of 48 listed consequences they experienced over the past year on a scale of 0 (no) and 1 (yes).

2.2.3. Drinking motives

were assessed via the Modified Drinking Motives Questionnaire (Grant et al., 2007). Participants were asked to indicate how often they drank for the listed reasons on a scale of 1 (never/almost never) to 5 (almost always/always). Subscales included: coping (13 items; α=.95; “to reduce your anxiety”); enhancement (5 items; α=.90; “because you like the feeling”); social (5 items; α=.92; “because it improves parties and celebrations”); and conformity (5 items; α=.85; “so you won’t feel left out”).

2.2.4. Sexual orientation-based discrimination

was measured using the Everyday Discrimination Scale (Krieger et al., 2005). Participants were asked “In your day-to-day life, how often do any of the following things happen to you?” and provided a list of 9 items (“you are treated with less respect than other people are”) on a scale of 1 (never) to 4 (often). For each item endorsed (response greater than 1), participants were asked to indicate “was this due to your...” (responses: “sexual orientation” or “other factors”). Responses to items attributed to sexual orientation were summed.

2.2.5. Sexual orientation-based harassment

was measured using the harassment subscale of the Daily Heterosexist Experiences Questionnaire (Balsam et al., 2013). This 6-item subscale assesses harassment and verbal victimization based on one’s sexual orientation (“being called names such as ‘fag’ or ‘dyke’”; α=.84) on a scale of 0 (never) to 5 (almost every day).

2.2.6. SGM community involvement

was assessed via a 10 item measure (Rosario et al., 2001). Participants were asked to indicate (yes/no) whether they participated in a range of SGM community activities in the past year (e.g., “gone to LGBTQ dance clubs, bars, discos, or hung around these places”). A count of “yes” responses was created.

2.3. Data Analyses

A total of 859 observations (20.3%) were missing. Within completed assessments, less than 1% of data were missing and were handled with full information maximum likelihood. First, growth mixture modeling (GMM) was performed in Mplus 8.4 to identify classes based on linear growth curves of drinking frequency and HED. Bayesian Information Criterion (BIC), sample size-adjusted BIC, Lo-Mendell-Rubin (LMR) likelihood ratio tests, parametric bootstrapped likelihood ratio tests (BLRT), entropy, smallest class size, and class interpretability were used to select the number of classes (Nylund et al., 2007). Lower BIC values indicate the preferred model, and significant LMR or BLRT indicate a preference for the current model.

Next, we examined associations with class membership.² The first analyses examined associations between minority stressors, SGM community involvement, and drinking motives at Wave 1 and class membership. Age at Wave 1, sexual orientation, and race/ethnicity were included as covariates. The second set of analyses examined differences in latent growth curve models of alcohol consequences across classes. As alcohol consequences had an over dispersed count distribution, negative binomial regression was used to estimate the intercept, slope, and class differences in these parameters. We selected negative binomial regression because negative binomial distributions do not assume that the mean and variance of the distribution are equal to one another as Poisson does (Gardner et al., 1995).

3. Results

Correlations, means, and standard deviations for variables at Wave 1 are presented in Table 1.

Table 1.

Correlations, means, and standard deviations at wave 1.

	1	2	3	4	5	6	7	8	9	10
1. Age	–
2. HED	0.19**	–
3. Drinking Frequency	0.28**	0.63**	–
4. Drinking Consequences	0.12**	0.60**	0.57**	–
5. Discrimination	−0.09*	0.01	0.03	0.11**	–
6. Sexual Orientation Harassment	−0.09*	0.08*	0.11**	0.16**	0.59**	–
7. Community Involvement	0.14**	0.20**	0.21**	0.14**	0.10**	0.23**	–
8. Social Motives	0.08*	0.49**	0.34**	0.45**	−0.03	0.01	0.12**	–
9. Enhancement Motives	0.06*	0.56**	0.38**	0.48**	−0.02	−0.01	0.07*	0.76**	–
10. Coping Motives	0.10**	0.44**	0.39**	0.59**	0.15**	0.12**	0.04	0.56**	0.58**	–
Mean	21.37	2.57	2.29	7.95	3.66	1.11	5.51	2.79	2.62	1.96
Standard Deviation	2.09	2.00	2.10	9.25	6.05	0.98	2.54	1.19	1.20	1.00
Range	18–26	0–7	0–7	0–45	0–34	0–5	0–10	1–5	1–5	1–5

HED (heavy episodic drinking).

^*p < .05

^**p < .001.

3.1. GMM

First, we used GMM to identify classes based on intercepts and linear slopes of HED and drinking frequency. Fit indices did not point to a single model (Table 2). BIC and adjusted BIC preferred the six-class solution, LMR-LRT the four-class, and BLRT the seven-class. We proceeded to examine the interpretability of 4–6 class solutions. The five-class solution was selected as it produced the most interpretable and well differentiated classes. Solutions 4 and 6 are presented in Supplementary Materials.

Table 2.

Growth Mixture Model: Fit Indices

			LMR-LRT		BLRT
Classes	BIC	Adjusted BIC	estimate	p	estimate	p	entropy
1	25253.76	25183.89
2	25056.52	24970.76	225.58	< .001	232.06	< .001	.85
3	25026.98	24925.34	62.56	.14	64.36	< .001	.83
4	24926.68	24809.16	135.87	< .001	139.77	< .001	.83
5	24913.98	24780.59	27.73	.10	28.53	< .001	.78
6	24893.67	24744.39	42.29	.23	43.50	< .001	.80
7	24925.63	24760.47	45.70	.41	46.36	< .001	.84

Lower BIC and adjusted BIC values indicate a better fitting model. Significant LMR and BLRT indicate a preference for the current model over the model with one less class. BIC = Bayesian information criterion; LMR = Lo-Mendell-Rubin likelihood ratio test; BLRT = bootstrapped likelihood ratio test. Bold text indicates the preferred model for each index of model fit.

Class 1 (low increasing drinking; n=77) had low intercepts for HED and drinking frequency and significant increases in both (see Table 3 and Figure 1). Class 2 (high decreasing drinking; n=103) had high intercepts on HED and drinking frequency and significant decreases in both. Class 3 (stable high drinking; n=223) had high intercepts on HED and drinking frequency and no significant change. Class 4 (stable high HED; n=89) had the highest intercepts for drinking frequency and HED, no change in HED, and a decrease in drinking frequency. Class 5 (stable low drinking; n=565) had low intercepts for HED and drinking frequency, a small but significant decrease in HED, and a small, significant increase in drinking frequency.

Table 3.

Five Class GMM Parameters

			HED			Drinking Days per Week
Class	Class Name	Class Size	Intercept	Slope	p	Intercept	Slope	p
1	Increasing HED + Drinking Frequency	77	1.25	1.04	< .001	1.40	.48	< .001
2	Decreasing HED + Drinking Frequency	103	3.81	−.87	< .001	4.92	−.94	< .001
3	Stable High HED + Drinking Frequency	223	3.92	.04	.53	2.56	.12	.09
4	Stable High HED	89	4.93	.03	.73	6.20	−.53	< .001
5	Stable Low HED and Drinking Frequency	565	1.54	−.23	< .001	1.08	.09	.01

Intercepts and slopes for HED and drinking frequency by GMM class.

Figure 1.

Figures of intercepts and slopes of HED and drinking frequency by GMM class. See Table 3 for parameter estimates each class Table 4 for parameter estimates for each class.

3.2. Class Membership and Wave 1 Covariates

We examined associations between GMM class membership, demographics, and hypothesized covariates (Table 4). The “stable low drinking” group was treated as reference group. Individuals in “high decreasing drinking,” “stable high drinking,” and “stable high HED” classes were older than those in the “stable low drinking” class. Bisexual women were less likely to be in the “stable high HED” and “decreasing drinking” classes than the “stable low drinking” class. All other differences for sexual orientation, race, and ethnicity were not significant. ³

Table 4.

Associations between class membership and wave 1 covariates.

	Class 1 Increasing Drinking		Class 2 Decreasing Drinking		Class 3 Stable High Drinking		Class 4 Stable High HED		Class 5 Stable Low Drinking
Wave 1 Covariate	OR	p	OR	p	OR	p	OR	p	OR	p
Age	0.78	0.28	1.98	<0.001	1.28	0.03	1.87	<0.001	ref	ref
Sexual Orientation
Lesbian	ref	ref	ref	ref	ref	ref	ref	ref	ref	ref
Bisexual	0.52	0.06	0.54	0.04	0.69	0.11	0.51	0.02	ref	ref
Race
White	ref	ref	ref	ref	ref	ref	ref	ref	ref	ref
Racial Minority	0.85	0.70	0.64	0.25	0.72	0.25	1.21	0.56	ref	ref
Ethnicity
Non-Latinx	ref	ref	ref	ref	ref	ref	ref	ref	ref	ref
Latinx	1.83	0.28	2.20	0.09	1.72	0.15	1.41	0.51	ref	ref
Discrimination	1.65	0.001	1.23	0.14	1.00	0.97	1.19	0.21	ref	ref
Sexual Orientation Harassment	1.19	0.28	1.42	0.01	1.13	0.30	1.45	0.01	ref	ref
Community Involvement	1.04	0.85	1.92	<0.001	1.45	0.001	1.39	0.048	ref	ref
Drinking Motives
Conformity	0.96	0.88	0.70	0.06	0.80	0.15	0.61	0.01	ref	ref
Social	1.52	0.35	1.24	0.31	1.76	0.01	0.94	0.80	ref	ref
Enhancement	0.76	0.49	2.04	0.001	2.67	<0.001	2.84	<0.001	ref	ref
Coping	0.95	0.88	1.99	0.001	1.08	0.68	2.80	<0.001	ref	ref

Continuous predictors on likert scales (i.e., discrimination, community involvement, drinking motives) and age were standardized prior to analyses. Example OR interpretation: a one standard deviation higher score on discrimination is associated with being 1.65 times more likely to be in the increasing drinking class than the stable low class.

Next, we examined group differences in minority stressors, community involvement, and drinking motives at Wave 1. Higher sexual orientation-based discrimination at Wave 1 predicted a higher likelihood of being in the “low increasing drinking” class than “stable low drinking.” Higher sexual orientation-based harassment predicted higher likelihood of being in “high decreasing drinking” and “stable high HED” classes compared to “stable low drinking.” Higher SGM community involvement at Wave 1 predicted a higher likelihood of being in “high decreasing drinking,” “stable high drinking,” and “stable high HED” classes compared to “stable low drinking.”

Several differences emerged on drinking motives. Higher conformity motives predicted a higher likelihood of being in the “stable low drinking” class than the “stable high HED” class. Higher social motives predicted a higher likelihood of being in the “stable high drinking” class than “stable low drinking.” Higher enhancement motives predicted a higher likelihood of being in “high decreasing drinking,” “stable high drinking,” and “stable high HED” classes compared to “stable low drinking.” Those with high coping motives were more likely to be in the “high decreasing drinking” or “stable high HED” classes compared to “stable low drinking.”

3.3. Follow-Up: Changes in Risk Factors

High decreasing and stable high HED classes had the same risk factors despite their very different trajectories. To explore factors that may differentiate these classes, we conducted post-hoc analyses of differences in changes in risk factors over time between these two groups. These groups did not differ significantly on changes in discrimination, harassment, community involvement, or conformity motives. However, they differed on changes in social, enhancement, and coping motives. The decreasing drinking class experienced significant decreases in social (b = −.30, p < .001), enhancement (b = −.38, p < .001), and coping motives (b = −.28, p < .001), while the stable high HED class only experienced decreases in coping motives (b = −.04, p = .02).

3.4. Class Membership and Consequences

Next, we examined differences among classes in the intercept and slope of alcohol consequences (Table 5; Figure 2). “Low increasing drinking” and “stable low drinking” had the lowest alcohol consequences at Wave 1, followed by “high decreasing drinking,” “stable high drinking,” and “stable high HED.” This pattern generally indicates that classes with the highest HED and drinking frequency intercepts also had the highest alcohol consequences intercepts. Differences in slopes also emerged. Only individuals in the “low increasing drinking” class experienced increases in consequences over time. “Stable high drinking” and “stable high HED” experienced small but significant decreases in consequences. “Stable low drinking” experienced a moderate decrease in consequences, while “decreasing drinking” experienced the largest decrease.

Table 5.

Associations between class membership and alcohol consequences.

	Class 1 Increasing Drinking		Class 2 Decreasing Drinking		Class 3 Stable High Drinking		Class 4 Stable High HED		Class 5 Stable Low Drinking
Growth Curve Model	Estimate	p	Estimate	p	Estimate	p	Estimate	p	Estimate	p
Alcohol Consequences
Intercept	2.32a	0.03	15.18b	<0.001	13.06b	<0.001	19.89c	<0.001	3.19a	<0.001
Slope (RR)	1.42a	0.01	.59b	<0.001	.90c,d	0.01	.91c	0.03	.79d	<0.001

Alcohol consequences were treated as count variables. Thus, slopes are in the form of rate ratios.

Figure 2.

Figures of intercepts and slopes of alcohol consequences by GMM stable high heavy episodic drinking. class. See Table 4 for parameter estimates for each class.

4. Discussion

This study was the first to both identify diverse drinking trajectories among SMW and explore minority stress and social influence risk factors for these trajectories. Findings suggest that minority stress and social influence risk factors may increase the likelihood that SMW enter into distinct drinking trajectories, with minority stress predicting one trajectory, social influence another, and both contributing to risk for other trajectories. These findings provide nuance to our understanding of drinking trajectories among SMW and how minority stress and social influences may uniquely and jointly contribute to risky drinking.

We identified five drinking trajectories. Four were similar to those identified in prior research (Andrade & Järvinen, 2020; Coulter et al., 2018; Iwamoto et al., 2018): stable low, stable high, high decreasing, and low increasing drinking. We also found one unique class, characterized by stable high HED and a high drinking frequency that decreased. This novel class may highlight the benefits of modeling multiple measures of drinking at once as prior research has largely used a single measure. While similar classes were identified in SMW and the general population, studies that directly compare trajectories between heterosexual and SMW are needed. At this point it is unclear whether SMW and heterosexual women have similar drinking trajectories, with SMW being more likely to be in riskier classes (as found by Coulter et al., 2018) or whether distinct trajectories are experienced by SMW (e.g., unique patterns and/or steeper increases and higher intercepts) as suggested by sexual orientation differences in the intercept and slope of a single average drinking trajectory (Dermody et al., 2014; Drabble et al., 2020; Oshri et al., 2014).

The current study’s trajectory classes differed in Wave 1 alcohol consequences and changes in consequences. Consistent with prior research, we found that stable high drinking (Coulter et al., 2018; Iwamoto et al., 2018; Marshall et al., 2015) and high decreasing (Schuckit et al., 2014) classes had higher consequences at Wave 1 than stable low drinking. Our unique class, stable high HED, had the highest consequences at Wave 1 – higher even than the stable high drinking class – which may be attributable to their high HED and drinking frequency at Wave 1. Few studies have examined associations between trajectory classes and changes in alcohol outcomes. As expected, the low increasing class experienced significant increases in drinking consequences, highlighting the risk associated with this trajectory. Surprisingly, four of the five classes experienced significant decreases in consequences. For stable high HED and stable high drinking classes, decreases were relatively small, while the high decreasing class experienced substantial decreases. The presence of decreases in consequences despite the lack of change in drinking patterns may arise from the development of tolerance, which may reduce prevalence of some consequences over time (e.g., negative physiological reactions to a specific quantity of alcohol) (Elvig et al., 2021). Further research is needed to elucidate associations between trajectory classes and changes in consequences.

Our findings suggest that minority stress and social influence risk factors may contribute to different drinking trajectories. The low increasing drinking class experienced elevated discrimination at Wave 1 but no social influence risk factors, while the stable high drinking class experienced predominately social influence risk factors (i.e., community involvement, social and enhancement motives). These two classes start with substantially different drinking patterns that converge by Wave 4. This suggests that although minority stressors and social influences may contribute to different trajectories, they may ultimately result in similar problematic drinking.

Interestingly, two very different classes – high decreasing and stable high HED – were marked by elevations in both minority stress and social influence risk factors. Both classes begin with high HED and drinking days. However, they quickly diverge, with HED and drinking days decreasing substantially for the decreasing class and remaining largely stable for the other. A post-hoc examination of class differences in the slopes of risk factors suggest that significant decreases in social, enhancement, and coping motives for drinking experienced by the decreasing drinking class may help to explain the resilience of this group. However, it is also possible that reductions in drinking may be causing reductions in these motives. Therefore, further research is needed to explore factors that may explain the resilience of the decreasing drinking class, such as adaptive coping, changes in social roles, or peer drinking.

4.1. Limitations

Results should be considered in light of study limitations. First, we only included SMW and thus, it is unclear to what extent findings generalize to sexual minority men and non-binary individuals. Second, as all available waves of data were used to estimate drinking trajectories, we were not able to examine prospective associations between drinking trajectories and subsequent changes in consequences and instead examined associations with concurrent consequences. Third, for the current analyses, we examined minority stressors and social influence risk factors at baseline as predictors of drinking trajectories. We did not examine the directionality of associations between minority stressors and social influence risk factors. Future research should examine the ways in which these two sets of risk factors may influence one another. Fourth, this study examines drinking trajectories and risk factors over a moderate period (four years). Future, research should examine these trajectories over a longer period and examine related factors like, motivation to reduce drinking, attempts at reducing drinking, and substance use disorder status. Fifth, while the sexual minority sample was a major strength, the absence of a heterosexual comparison group meant that we were unable to directly examine differences in drinking trajectories between sexual minority and heterosexual women, a direction for future research. Sixth, due to small numbers of individuals from some minoritized racial groups in some of the trajectory classes, we were forced to combine all racial minority individuals into a single group for analyses. This is less than ideal given that different racial groups display different patterns of risk for heavy drinking and alcohol use disorders. Future research should utilize more diverse samples of SMW to examine racial differences in drinking trajectories in more detail. Seventh, we did not examine all variables that have been linked to HED and drinking frequency. Future research should explore the roles of additional variables (e.g., social roles) in drinking trajectories.

4.2. Conclusion

Despite some limitations, the current study substantially extends our understanding of drinking trajectories in a high risk and understudied population, SMW. Findings indicate that some drinking trajectories in this population appear to be similar to those found in the general population, while others may be unique. Further, trajectories were differentially predicted by minority stressors, social risk factors, or both, providing insight into how these two sets of risk factors may contribute to disparities affecting SMW. These findings have the potential to inform future individual and community-level preventive and treatment interventions.

• Minority stressors predicted low increasing drinking trajectory.

• Social influence factors predicted stable heavy drinking trajectory.

• Minority stressors and social influences predicted high decreasing and stable high heavy episodic drinking trajectories.