Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2014 Mar 1.
Published in final edited form as: Addict Behav. 2012 Nov 29;38(3):1824–1830. doi: 10.1016/j.addbeh.2012.11.011

Drinking Settings, Alcohol Consumption, and Sexual Risk Behavior among Gay Men

Rhonda Jones-Webb 1, Derek Smolenski 1, Sonya S Brady 1, Michael Wilkerson 1, Simon Rosser 1
PMCID: PMC3569002  NIHMSID: NIHMS428623  PMID: 23261495

Abstract

We sought to 1) describe the settings or groups of settings where men who have sex with men (MSM) consume alcohol in 16 U.S. metropolitan statistical areas (MSA); and 2) investigate whether certain drinking settings or groups of settings are associated with higher levels of alcohol consumption, problem drinking, and sexual risk behavior. Latent class analysis was used to develop our measure of drinking settings. The final latent class model consisted of four distinct classes which captured the typical settings where MSM consumed alcohol: “home” “social,” “bar/social,” and “general” drinkers (i.e., drinks in all settings). Regression models showed that “general” drinkers were more likely than “social” drinkers to engage in frequent heavy drinking. Compared to `social' drinkers, general drinkers were also more likely to engage in unprotected anal intercourse (UAIMP) and UAIMP with men met in bars and other venues (e.g., private parties, bath houses) while intoxicated. Assessment of drinking settings may be a means of identifying MSM who are at greater risk for frequent, heavy drinking and related sexual risk behavior.

Keywords: alcohol consumption, sexual risk behavior, drinking settings, gay men

Introduction

Alcohol use among men who have sex with men (MSM) is pervasive (Ostro & Stall, 2008) and a cause for concern. MSM comprise the majority of HIV infections (CDC, 2011), and studies have shown that heavy alcohol use is associated with HIV risk behaviors, HIV transmission (Baliunas et al., 2010; Celentano et al., 2006; Geibel et al., 2008; Trocki & Leigh, 1991; VanDevanter, 2011), and other health problems including cardiovascular disease, diabetes mellitus, liver cirrhosis, and several cancers (Corrao, Bagnardi, Zambon, & La Vecchia, 2004; Rehm et al., 2003; Rehm, Bondy, Sempos, & Vuong, 1997).

Drinking settings refer to the places where individuals consume alcohol, and include both public milieus (bars, restaurants, night clubs, sporting and outdoor events, parks) and private milieus (e.g., one's own home). Given the prevalence of both drinking and HIV risk behavior among MSM (Ostro & Stall, 2008, CDC, 2011) and the association between alcohol use and risk behavior (Bailiunas et al., 2010, Celentano et al., 2006), identifying the settings where MSM typically consume alcohol can elucidate the way drinking settings contribute to heavy drinking and sexual risk behavior.

Several studies have investigated the association between drinking settings and alcohol use (Casswell, Pledger, & Pratap, 2002; Curran, Harford, & Muthen, 1996; Kairouz & Greenfield, 2006), but few of these studies have focused exclusively on MSM (Greenwood, et al., 2001; Stall et al., 2001; Wong, Kipke & Weiss, 2008). These studies suggest that the social context in which drinking occurs can influence both the amount of alcohol an individual consumes as well as their subsequent health risk. Some studies have reported a positive relationship between frequency of drinking in bars and higher levels of alcohol consumption (Caswell, et al., 2002; Curran, et al., 1996; Kairouz et al, 2006; Trocki & Drabble 2008), as well as problems such as drinking and driving, injuries, and fights (Stockwell, Lang, & Rydon, 1994; Usdan, Moore, Schumacher, & Talbott, 2005). Other studies have reported that settings such as parks, sporting events, and parties are also associated with increased alcohol consumption and drinking problems (Clapp, Reed, Holmes, Lange, & Voas, 2006; Harford, Seabring, & Wechsler, 2002; Jones-Webb et al., 1997; Wells, Graham, Speechley, & Koval, 2005). Still, other studies have found the relationship between drinking settings and alcohol use differs by important demographic variables such as age, gender, and education (Clapp et al., 2006; Harford et al., 2002; Wells et al, 2005; Usdan, et al., 2005).

Studies of drinking settings and alcohol use among MSM have primarily examined gay bar settings (Greenwood, et al., 2001; Wong, et al., 2008). This is not surprising given gay bars historically have served as an important social institution in gay community life (Wolitski, Stall, & Valdiserri, 2008). Wong and colleagues (2008) examined the relationship between frequency of gay bar attendance and alcohol consumption among 526 young adult MSM in Los Angeles and found that gay bar attendance was associated with greater frequency of drinking and binge drinking. Greenwood et al., (2000) also reported that frequency of gay bar attendance was associated with heavy alcohol use among MSM and bisexual men living in San Francisco. In contrast, Trocki and colleagues (2005) reported that drinking in bars was not associated with heavy drinking among gay, bisexual, and heterosexual men; however this study did not focus exclusively on gay bars or MSM. While studies of gay bar attendance and alcohol use have provided valuable insight into the social context surrounding heavy alcohol use among MSM, these studies have a number of limitations. First, studies in this area have focused solely on gay bars to the exclusion of other drinking settings where heavy drinking might occur such as parties. Thus, little is known about the settings where MSM consume alcohol and if certain settings are associated with higher levels of alcohol consumption than other settings. Second, despite high rates of alcohol use and HIV infection among MSM (Stall et al., 2001; CDC, 2011), studies of gay bar attendance and alcohol use have not explored whether drinking settings are associated with sexual risk behavior. Lomba and colleagues (2009) examined the relationships between drinking settings, alcohol use, and sexual risk behavior among teens and young adults in Portugal. About 3% of Lomba's sample included respondents who self-identified as gay or lesbian. Lomba found that frequency of attending night clubs/bars was associated with higher levels of alcohol consumption and that higher levels of alcohol consumption were associated with sexual risk behaviors, such as drinking while engaging in sex. The study by Lomba and colleagues (2009) did not directly test whether drinking settings were associated with sexual risk behavior. To date, no studies have investigated whether certain drinking settings or groups of settings are associated with sexual risk among MSM or other populations.

The aims of our study were to 1) describe the settings or groups of settings where MSM consume alcohol in 16 U.S metropolitan statistical areas (MSAs), and 2) investigate whether certain drinking settings or groups of settings (e.g., bars, restaurants, parties) are associated with higher levels of alcohol consumption, problem drinking, and sexual risk behavior.

Methods

Data

The present study is part of a larger, five-year study (SILAS) examining the effects of state laws related to homosexuality on alcohol use, unsafe sexual behavior, and alcohol-related unsafe sex among MSM (Horvath, 2010). The SILAS sample was comprised of MSM who were recruited online from 16 moderately sized metropolitan statistical areas (MSAs) in the U.S, e.g., Minneapolis, Albany, Cleveland, and Portland. Data for this analysis were collected in 2008.

Recruitment

In 2008, SILAS participants were recruited online from the two of the largest gay/bi social networking sites, Manhunt.net and Gay.com. Eligibility criteria included being 1) age 18 years or older, 2) a male who self-reported having sex with a man, 3) a resident in one of the 16 MSAs, and 4) a new study participant. Banner advertisements promoting a “Sex and Alcohol Survey” were displayed on the websites. Upon clicking on the banner advertisement, potential participants were brought to the log-in page. Over 3,370 invitations were issued to individuals who clicked on the banner advertisements and who provided contact information. Of the 2,305 eligible enrollees, 1,725 (90.55%) completed an online survey questionnaire, yielding an overall completion rate of 74.83%.

We screened for duplicate respondents prior to and following the survey. Prior to taking the full study survey, respondents were asked to complete a screener survey. The screener gathered qualifying demographics such as age and location, along with logistical information such as an email for Amazon.com payments, and a preferred email where the SILAS team would send qualifying candidates a survey activation link. When surveys were submitted, the IP address of qualifying candidates was recorded and compared to their stated location. This information was also compared against the database of existing users. Potential duplicates were identified based on IP address and location, and a simple algorithm assigned candidates a possible duplicate score. Once determined to be eligible, study subjects provided their numerical age at the beginning of the survey, and a birth date (dd/mm/yy) at the end of the survey. Age, birth date, other elements of demographic information, and time taken to complete the survey were analyzed using our deduplication algorithm, and once again subjects were assigned a possible duplicate score. Throughout all phases of the study, duplicate respondents were identified and not allowed to advance. If respondents were identified as duplicates after completing the survey, their first survey would be kept, and all subsequent surveys were removed from the dataset.

Data Collection and Measures

The questionnaire included items and scales on a variety of topics including typical Internet use, sexual and alcohol behavior with partners met online versus in bars/clubs, versus in other locations, current alcohol and drug use, and sexuality and relationship measures. Data were collected between June 18, 2008 and October 1, 2008, and participants were compensated $30 for their participation. On average, participants completed the online survey in 71 minutes.

Outcome variables

Dependent variables were alcohol consumption, problem drinking, and sexual risk behavior. We assessed alcohol consumption by combining two items to construct a quantity and frequency drinking typology that ranged from abstinence to frequent heavy drinking (see categories in Table 5). The first question examined frequency: “In the past 3 months, on average, how often did you drink any kind of alcoholic beverage?” (never, < 1 a month, about once a month, 2–3 times a month, 1–2 times a week, 2–3 times a week, nearly every day, at least once a day), and the second, quantity: “In the past 3 months, on a day when you drank some alcohol, how many drinks did you usually have?” (Greenfield, Midanik & Rogers, 2000). We assessed problem drinking using the 4-item CAGE questionnaire (Ewing, 2008; Bush, Shaw, Cleary, Delbanco & Aronson, 1987). Items on the CAGE assess the individual's perceived need to cut down on alcohol use, feelings of annoyance by people criticizing one's drinking, feeling bad or guilty about drinking, and need for a drink first thing in the morning to steady one's nerves or to cure a hangover (i.e., “eye-opener”); items are scored yes or no. A total score of 2 or more “yes” responses was used to classify participants as screening positive for problem drinking.

Table 5.

Distribution and multinomial regression of alcohol consumption and CAGE outcomes on assigned latent class

Alcohol consumption
Class Infrequent, light/moderate Row% Infrequent, heavy Row% Frequent, light/moderate Row% Frequent, heavy Row% CAGE ≥ 2 Row%
Home 43.54 8.37 32.06 16.03 26.07
Sociala 34.25 7.67 36.71 21.37 26.98
Social & Bar 31.82 9.38 42.61 16.19 27.04
General 10.49 5.85 53.66 30.00 33.73
POR (95% CI) POR (95% CI) POR (95% CI) POR (95% CI) POR (95% CI)
Home Ref. 0.96 (0.53,1.74) 0.70 (0.45,1.07) 0.59 (0.37, 0.94) 0.84 (0.65,1.08)
Social & Bar Ref. 1.41 (0.90, 2.21) 1.20 (0.83,1.74) 0.81 (0.53,1.25) 0.85 (0.63,1.15)
General Ref. 2.98 (1.42, 6.23) 4.63 (2.76, 7.75) 4.86 (2.45, 9.63) 1.27 (0.82,1.98)
Open in a new tab

Note: POR = prevalence odds ratio; CI = confidence interval. All POR estimates adjusted for age, education, income, race, HIV status, gay identification, and long-term relationship status. All regression models are weighted for latent class estimation uncertainty and employ a robust standard error estimate that accounts for MSA-based sampling.

a

Reference category for the exposure variable in the multinomial regression.

We assessed sexual risk behavior across three environments. Participants were asked about their sexual behavior in the last 3 months separately with men met online, men met in bars, and men met in other places (e.g., private parties, bathhouses). We asked men to estimate the number of men with whom they had unprotected anal intercourse with a male partner (UAIMP) both overall and specifically while intoxicated; these served as our primary measures of sexual risk behavior. We analyzed both the UAIMP and UAIMP while intoxicated variables separately by environment as well as in an aggregate form across the three environments to examine overall risk behavior.

Exposure variable

Drinking settings served as our key exposure variable and was measured by asking the following question, “Where do you typically drink alcohol?” Response categories were in restaurants, gay bars/clubs, social gatherings, others' homes, own home, outdoor leisure, concerts/sporting events. Respondents could check all that apply. A drinking setting variable was constructed using latent class analysis (LCA) to develop a measure that captured the settings or groups of settings in which MSM consumed alcohol. The results of the LCA is described in detail in the data analysis section below.

Covariates

Demographic covariates included age, education level, income, race/ethnicity, long term relationship status, gay identification, and HIV status. These variables were selected because they have been shown to be related to drinking and/or sexual risk behavior (U.S. Department of Health and Human Services, 2010; Ostrow & Stall, 2008). Relationship status was measured by asking, “Do you have a relationship with a man who you would describe as your long-term boyfriend, domestic partner or spouse (yes/no)?” Gay identification was assessed by asking whether respondents identified as gay/homosexual, bisexual, heterosexual, or other. HIV status was measured by the following question: “Have you ever been diagnosed with HIV (yes/no)?”

Data Analysis

First, we used a latent class model to identify homogeneous groups based on their response patterns to the drinking setting item, “Where do you typically drink alcohol?” Latent class analysis (LCA) is designed to identify a small number of classes or groups that exhibit similar patterns of responses, with the result being independence among indicators within each class (Collins & Lanza, 2010). Because variability between each indicator is modeled as residual error, a latent variable is considered “error-free” and a more precise estimate of the unobserved construct than any individual indicator or simple sum of multiple indicators. In general, studies on drinking settings and alcohol use have been limited by their reliance on individual indicators of drinking settings (e.g., frequency of drinking in bars, last location alcohol was consumed prior to an injury or accident (Lang & Stockwell, 1991; MacDonald et al, 2006; McLean & Connor, 2009; Usdan et al, 2005). Our approach to measuring drinking settings thus provides an alternative method to standard approaches of measuring drinking settings which can improve statistical inference. For example, our approach to measuring drinking settings permits us to identify distinct patterns in the drinking settings of MSM that might not be apparent using an individual indicator (e.g., 1–2 settings, 3–4 settings, etc).

Given the large sample size, we randomly divided participants into one training and two validation subpopulations prior to latent class modeling. For the training subpopulation, we estimated iterative latent class models with each additional model containing one additional class. We used standard indices of absolute and relative model fit to identify the optimal solution (Collins & Lanza, 2010). We examined the absolute fit of the model using the G2 statistic, with a non-significant (p>.05) p-value indicating a non-significant discrepancy between the model implied contingency table and the observed contingency table of the data. Since multiple models can have an acceptable absolute fit, we also examined general relative fit indices including the Akaike information criterion (AIC), the Bayesian information criterion (BIC), and the Lo-Mendel-Rubin likelihood ratio test (LMR LRT). For the AIC and BIC, lower values indicated an improvement in model fit. The likelihood ratio test was used to identify a stopping point for the iterative models since a non-significant (p>.05) p-value indicates no further improvement in model fit resulting from an additional estimated latent class. Additional considerations in arriving at the final training model included parsimony and substantive interpretation of the latent classes. We used a multi-group latent class analysis to determine the equivalence of the item thresholds and the population distribution across the three randomly-derived subpopulations (Collins & Lanza, 2010). Non-significant differences in the G2 statistic between freely-estimated and constrained models were used to support an inference of model replicability. The final solution was modeled using the entire analytic sample.

Next, we compared demographic variables across the identified latent classes using a multinomial regression model with the latent classes specified as the outcome variable. For the alcohol outcomes, we used a multinomial regression model with the latent classes specified as a categorical explanatory variable. We included estimates for UAIMP and UAIMP while intoxicated, both overall and stratified by environment (meeting a partner in a bar, online, or in other places), to examine the association between the drinking classes and risk behavior. For the overall variables, we had sufficient variability in the data to employ a negative binomial regression to examine the difference in the prevalence rates between the classes. For the environment-specific analyses, the counts were quite sparse, so we used a logistic regression model to examine the association between any risk behavior and the drinking venue classes. Models in which alcohol was the outcome of interest excluded non-drinkers since these participants, by definition, are not eligible for either the exposure or outcome used in the model; models in which the outcome was sexual risk behavior included both non-drinkers and drinkers, since the “non drinker” designation does not preclude sexual risk behavior. All regression models adjusted for age, education, income, race/ethnicity, long term relationship status, gay identification and HIV status; sexual risk models also adjusted for alcohol consumption (US. Department of Health and Human Services, 2010; Ostro & Stall, 2008). For all regression models, we used the probability of latent class assignment as a weight variable to account for the probabilistic assignment of participants to their most likely latent class (Clark & Muthén, 2009). We used robust variance estimates to account for nonindependence of observations based on the selection of study participants from 16 defined MSAs. The primary motivation for this approach to variance estimation is to account for the nonindependence of individual observations introduced by selecting study participants from defined catchment areas. Participants within a defined MSA may be more similar to one another across a broad range of characteristics than to those selected from another catchment area. Using the robust variance estimator allows us to account for potential nonindependence to reduce bias in statistical inference. The latent class models were estimated using Mplus, Version 6.0 (Muthén & Muthén, 2009), and the regression models were estimated using Stata, Version 11.2 (Stata, 2012).

Results

Descriptive

Sample characteristics

Table 1 describes the demographic characteristics of the sample. In general, the sample was white (74%), young (56% ≤ age 34 years), and highly educated (51% with 4+ years of college). Most respondents were single (71%) and 14% of respondents were living with HIV.

Table 1.

Demographic characteristics of SILAS respondents (n=1,725)

Variable n %
Age (years)
 18–24 420 24.36
 25–34 554 32.13
 35–44 441 25.58
 45–54 233 13.52
 55 and above 67 3.89
 Missing 9 0.52
Education
 High school/Technical school 245 14.21
 Some college 593 34.40
 College 578 33.53
 Post-graduate 308 17.87
Annual income ( $US)
 <20,000 458 26.57
 20,000 – 39,999 512 29.70
 40,000   59,999 336 19.49
 60,000 – 79,999 154 8.93
 80,000 – 99,999 72 4.18
 ≥100,000 84 4.87
 Missing 108 6.26
Race/ethnicity
 White, non-Hispanic 1,269 73.61
 Black, non-Hispanic 62 3.60
 Hispanic 245 14.21
 Other 121 7.02
 Missing 27 1.57
Sexual orientation
 Gay 1,487 86.35
 Bisexual 207 12.01
 Heterosexual 13 0.75
 Other 15 0.87
 Missing 2 0.12
Relationship status
 Single 1,218 70.65
 In a long-term relationship 505 29.29
 Missing 1 0.06
HIV-status
 HIV-negative 1,471 85.32
 HIV-positive 244 14.15
 Missing 9 0.52
Open in a new tab

Drinking settings

The model fit indices for several iterative latent class models in the training subsample are reported in Table 2. The four-class solution was the first solution to have a non-significant G2 value, and had AIC and BIC values lower than those for the two- and three-class solutions. The five- and six-class models also had good absolute fit to the data, but the BIC values were higher than those for the four-class solution, and the decreases in the AIC were modest, leading us to prefer the four-class solution based on statistical criteria. Review of the class structure for the four- and five-class models also favored the four-class model in terms of improved interpretability and overall model parsimony. Multi-group assessment of the four-class model between the training and two validation subsamples indicated that the four-class model thresholds and distributions replicated across the three groups (Table 2).

Table 2.

Fit indices for iterative latent class models in the training sample (n = 518), and multi-group analysis of the final model solution between the training and validation split samples

Training model
No. classes AIC BIC G2,df p-value Entropy LMR LRT p-value
2 4536.92 4609.17 368.81, 238 <.001 0.82 11067.37 <.001
3 4454.22 4564.72 238.11, 229 .039 0.75 9065.22 <.001
4 4414.98 4563.73 210.87, 220 .659 0.72 8821.89 <.001
5 4400.67 4587.67 168.98, 210 .983 0.83 8683.37 <.001
6 4402.66 4627.91 152.97, 201 .995 0.82 8602.99 <.001
Multi-group model
ΔG2, df p-value
Parameters freely estimated 17024.53 17597.59 663.83, 660 .451 - -
Constrained thresholds 16978.29 17208.58 745.59, 724 .281 81.76, 64 .067
Constrained distribution 16974.08 17172.24 753.38, 730 .267 7.79, 6 .254
Open in a new tab

Note: AIC = Akaike information criterion; BIC = Bayes information criterion; G2 = likelihood ratio test of model fit, LMR LRT = Lo-Mendel-Rubin likelihood ratio test.

Conditional item endorsement probabilities (i.e., the likelihood of favoring a particular setting) shown in Table 3 guided our naming of the four classes. Participants in the first class, `home' drinker had a greater probability of indicating drinking in their own home than other venues, such a social gatherings. The second class, `social' drinker consisted of higher endorsements of drinking in restaurants, at social gatherings, and at their own and others' homes than other venues. The third class, “social/bar” included participants who endorsed the two items specific to alcohol consumption in bars in addition to the other social venues. Finally, participants in the last class, `general' had a high probability of indicating they drank at all venues included in the model. The distribution of participants into assigned latent classes was approximately uniform. Classification accuracy for each class was high, with “home” (89.5%) and the “all” (88.5%) having the highest values. The overall quality of the model, or entropy, was adequate at 0.68.

Table 3.

Distribution of item endorsements within the identified latent classes (n = 1565)

Class
Venue Home (27.35%) p Social (23.45%) p Social and bar (22.67%) p General (26.52%) p
Restaurants 0.34 0.59 0.90 0.97
Gay bars/clubs 0.47 0.40 0.87 0.92
Non-gay bars/clubs 0.12 0.22 0.69 0.94
Social gatherings 0.19 0.81 0.95 0.99
Others' homes 0.15 0.77 0.69 0.95
Your own home 0.55 0.83 0.64 0.93
Outdoor leisure 0.01 0.42 0.08 0.86
Concert/sporting events 0.01 0.27 0.12 0.80
Average classification accuracy 89.5% 76.7% 74.4% 88.5%
Open in a new tab

Note: p=proportion of respondents in a class who endorse the particular item. Proportions in bold indicate a greater likelihood of endorsement for the particular item. Overall entropy of the model classification was 0.68.

Drinking settings by demographic characteristics

Non-drinkers and `home' drinkers were somewhat similar in terms of demographic characteristics (Table 4). Compared to `social' drinkers, non drinkers and `home' drinkers were older, and `home' drinkers more likely to report being HIV positive. The `social/bar' and `general' drinkers were younger and more likely to self-identify as gay as compared to `social' drinkers; the `general' drinkers also had a greater odds of earning $40,000 or more per year as compared to `social' drinkers.

Table 4.

Descriptive statistics and multinomial regression of non-drinkers and drinkers, by assigned latent class, on demographic variables

Non-drinker (n=159) M (SD) Home (n = 428) M (SD) Social (n = 367) M (SD) Social/Bar (n = 355) M (SD) General (n = 415) M (SD)
Age, years 37.9 (11.8) % 35.5 (10.9) % 33.2 (11.9) % 32.0 (9.5) % 31.7 (9.2) %
College education 52.2 47.9 45.8 58.3 53.7
Income ≥$40,000 41.8 38.9 40.2 35.6 44.0
Non-Hispanic White 16.4 18.2 16.4 15.8 13.5
Not gay-identified 15.1 13.1 18.0 11.0 12.1
Long-term relationship 29.6 31.4 30.5 26.2 28.7
Living with HIV 19.2 20.5 12.3 11.6 9.9
POR (95% CI) POR (95% CI) POR (95% CI) POR (95% CI) POR (95% CI)
Age, years 1.04 (1.02, 1.07) 1.03 (1.01, 1.04) Ref. 0.99 (0.98, 1.01) 0.99 (0.98, 1.00)
College education 1.39 (0.80, 2.42) 1.19 (0.84, 1.68) Ref. 1.83 (1.39, 2.41) 1.48 (1.12, 1.95)
Non-drinker POR (95% CI) Home POR (95% CI) Social POR (95% CI) Social/Bar POR (95% CI) General POR (95% CI)
Income ≥$40,000 1.16 (0.78, 1.74) 1.05 (0.81, 1.37) Ref. 0.91 (0.63, 1.31) 1.28 (1.07, 1.52)
Non-Hispanic White 1.02 (0.67, 1.57) 1.16 (0.89, 1.50) Ref. 1.02 (0.65, 1.61) 0.82 (0.62, 1.08)
Not gay-identified 0.76 (0.40, 1.46) 0.65 (0.40, 1.07) Ref. 0.50 (0.31, 0.81) 0.59 (0.42, 0.82)
Long-term relationship 0.97 (0.67, 1.40) 1.04 (0.74, 1.46) Ref. 0.82 (0.53, 1.25) 0.94 (0.66, 1.32)
Living with HIV 1.73 (0.97, 3.10) 1.91 (1.13, 3.23) Ref. 0.96 (0.55, 1.67) 0.80 (0.48, 1.31)
Open in a new tab

Note: POR=prevalence odds ratio; Cl = confidence interval

Multivariate

Drinking settings and alcohol use

Drinking setting was associated with level of alcohol consumption. `General' drinkers had a significantly greater prevalence of frequent and/or heavy drinking as compared to `social' drinkers. In contrast, `home' drinkers were less likely to report frequent, heavy drinking; they were similar to `social' drinkers with respect to likelihood of infrequent, heavy drinking and frequent, light/moderate drinking (Table 5). There were no differences identified between social and social/bar and social drinkers in terms of frequency and intensity of alcohol use. Also, there were no observed differences between the drinking classes in terms of the CAGE assessment.

Drinking settings, and sexual risk behavior

Drinking setting was also associated with sexual risk behavior. Table 6 provides a more detailed analysis of the relationship between drinking setting and sexual risk behavior across and for each of the three environments examined in our UAIMP measures (i.e., men met online, bar, other). As expected, `general' and `social/bar' drinkers had a greater odds of reporting UAIMP with partners met in bars as compared to “social” drinkers. Of note, both the `general' and `social/bar' classes include drinking in bars as a defining characteristic. There was no association between drinking setting and UAIMP among MSM who met men online or in other venues. In terms of UAIMP while intoxicated, the pattern of associations was similar to UAIMP in general. For example, `general' drinkers had the highest prevalence of UAIMP with partners met through either bars or other venues while intoxicated.

Table 6.

Logistic regression of any unprotected anal intercourse male partners (UAIMP) and any UAIMP while intoxicated on assigned latent class, both overall and by sex environment

UAIMP UAIMP while intoxicated
Drinking class All environments M (SD) Online % Bar % Other % All environments % Online % Bar % Other %
Non-drinker 36.5 34.0 3.8 8.8 - - - -
Home 36.9 31.5 8.9 8.4 12.2 5.6 2.3
Sociala 38.7 34.9 5.7 10.4 14.2 5.2 4.1
Social & Bar 35.8 30.4 11.6 7.3 11.8 7.3 2.8
General 35.9 30.1 10.4 9.4 14.2 8.9 6.3
POR (95% CI) POR (95% CI) POR (95% CI) POR (95% CI) POR (95% CI) POR (95% CI) POR (95% CI) POR (95% CI)
Non-drinker 0.87 (0.52, 1.44) 0.91 (0.52, 1.59) 0.50 (0.20, 1.23) 0.75 (0.37, 1.55) - - - -
Home 0.90 (0.68, 1.20) 0.83 (0.59, 1.16) 1.47 (0.93, 2.34) 0.74 (0.47, 1.18) 1.16 (0.86, 1.57) 0.78 (0.47, 1.31) 1.00 (0.64, 1.57) 0.65 (0.34, 1.24)
Social & Bar 0.93 (0.64, 1.35) 0.84 (0.52, 1.37) 2.04 (1.31, 3.20) 0.75 (0.36, 1.57) 1.40 (0.89, 2.20) 0.88 (0.49, 1.57) 1.37 (0.84, 2.23) 0.68 (0.28, 1.66)
General 0.98 (0.69, 1.38) 0.85 (0.58, 1.25) 2.06 (1.12, 3.78) 1.02 (0.65, 1.58) 1.80 (1.21, 2.67) 1.07 (0.68, 1.68) 1.95 (1.11, 3.45) 1.77 (1.06, 2.97)
Open in a new tab

Note: POR=prevalence odds ratio; CI = confidence interval. All PORs adjusted for age, education, income, race, long-term relationship status, gay identification, and HIV status. All regression models are weighted for latent class estimation uncertainty and employ a robust standard error estimate that accounts for MSA-based sampling.

a

Referent class for negative binomial regression models of UAIMP and UAIMP while intoxicated outcome variables.

Discussion

We identified four distinct patterns in the drinking settings of MSM: `home' (27%), `social' (23%), `social/bar' (23%), and `general', i.e., drank in all settings (27%). `General' drinkers (i.e., drank in all settings) consumed significantly more alcohol than `social' drinkers. Because general drinkers drank in more venues than `social' drinkers, they likely had greater opportunities to drink. `General' drinkers also had higher incomes than `social' drinkers and therefore had greater financial resources to drink in more settings. Non-drinkers and `home' drinkers were more likely to be living with HIV than other types of drinkers. The potential experiences of discrimination may partially explain why `home' drinkers typically drank at home and to a lesser degree at gay bars/clubs.

Compared to `social' drinkers, 'general' drinkers were also significantly more likely to engage in UAIMP with male partners met at bars and other venues while intoxicated. Availing oneself of alcohol across a variety of settings may lead to more opportunities to engage in risky sexual practices among MSM. Our findings on the relationship between drinking settings and sexual risk behavior have not been previously reported.

Drinking settings were not significantly associated with problem drinking as measured by the CAGE. While the CAGE has good internal reliability, responses to the CAGE are not stable over time (Dawe & Mattick, 1997). Thus, our null finding on drinking settings and problem drinking may reflect certain limitations of the CAGE. Alternatively, our findings may suggest that drinking settings are not associated with outcomes such as alcoholism.

Limitations

Several limitations of our study should be noted. First, we did not ask participants to indicate how much they drank in each setting they checked; previous studies have done so (Cassell et al, 2002; Kairouz & Greenfield, 2006). Future studies may wish to compare the validity of different approaches to measuring drinking settings among MSM. Latent class analysis offers one possible approach to measuring drinking settings. Second, we did not examine the characteristics of the drinking settings of MSM, and which characteristics were most associated with heavy drinking. Future studies should explore if certain characteristics of drinking venues are associated with heavy drinking more than others among MSM (e.g., size, location, presence of music/dancing, establishment practices and policies, presence of other roommates in private settings). Third, results from our study are generalizable to MSM who live in select MSA's. Future studies should replicate our study in other cities, especially large cities such as New York and Los Angeles. Finally, our data are cross-sectional. Although our findings are consistent with a model in which drinking in a variety of settings leads to increased drinking and sexual risk behavior, we cannot assume that this is the case. Future studies in this area will require stronger analytical study designs to confirm results from our study.

Implications

Despite the limitations of our study, our findings have important implications for designing alcohol interventions that target MSM. Our findings suggest that alcohol interventions should target MSM who drink in multiple settings, not only gay bars. Internet-based interventions may provide a vehicle for reaching such men (Bowen, Williams, Daniel, & Clayton, 2008; Levy & Strombeck, 2002; Rhodes, 2004). Internet-based interventions reach large numbers of individuals and offer a degree of anonymity which may be appealing to MSM who are heavy drinkers (Cooper, 1998). Also, the flexibility of the Internet may make it easier to integrate health messages regarding heavy drinking and HIV risk behavior. Individual level alcohol interventions are most effective when complemented by environmental strategies that seek to alter aspects of the drinking environment, e.g., drinking settings (Toomey & Lenk, 2011). Internet based interventions that target MSM who drink in multiple settings should therefore be complemented by strategies that alter aspects of drinking settings that encourage heavy drinking, e.g., over service, drinking games, seeing others intoxicated (Zakelstaka, Mundt, Balousek, Wilson, & Fleming, 2009; Clapp et al, 2006).

Conclusion

Our results build on and expand the current research on drinking settings and alcohol consumption among MSM (Greenwood et al., 2001; Stall et al., 2001; Wong et al., 2008). Results suggest that MSM consume alcohol in a variety of settings and there are distinct patterns in where MSM consume alcohol. Further, our results indicate that drinking in a multiple settings may increase alcohol consumption as well as UAMIP with men met in bars and other venues while intoxicated. Internet-based interventions might be one avenue for reaching these men.

Research Highlights

  • There are distinct patterns in where MSM consume alcohol

  • Drinking in a variety of settings is associated with increased alcohol consumption and sexual risk behavior

  • Highlights the importance of focusing on MSM who drink in multiple settings

Acknowledgements

This grant was supported by grant funding from the National Institute of Alcohol Abuse and Alcoholism, B.R.S. Rosser, Principal Investigator.

References

  1. Baliunas D, Rehm J, Shuper P. Alcohol consumption and is of incident human immunodeficiency virus infection: A meta-analysis. International Journal of Public Health. 2010;55:159–166. doi: 10.1007/s00038-009-0095-x. [DOI] [PubMed] [Google Scholar]
  2. Bowen A, Williams M, Daniel C, Clayton S. Internet based HIV prevention strategy targeting rural MSM: Feasibility, acceptability, and preliminary efficacy. Journal of Behavioral Medicine. 2008;31:463–477. doi: 10.1007/s10865-008-9171-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bush B, Shaw S, Cleary PD, Delbanco TL, Aronson MD. Screening for alcohol abuse using the CAGE questionnaire. Journal of Family Practice. 1987;44:151–160. doi: 10.1016/0002-9343(87)90061-1. [DOI] [PubMed] [Google Scholar]
  4. Caswell S, Pledger M, Pratap S. Trajectories of drinking form 18–26 years: Identification and prediction. Addiction. 2002;97:1427–1437. doi: 10.1046/j.1360-0443.2002.00220.x. [DOI] [PubMed] [Google Scholar]
  5. Celentano DD, Valleroy LA, Sifakis F, MacKellar DA, Hylton J, Thiede H, Torian LV. Associations between substance use and sexual risk among very young men who have sex with men. Sexually Transmitted Diseases. 2006;33:265–271. doi: 10.1097/01.olq.0000187207.10992.4e. doi: 10.1097/01.olq.0000187207.10992.4e/ [DOI] [PubMed] [Google Scholar]
  6. Centers for Disease Control [Accessed 1/5/2012];HIV and AIDS among gay and bisexual men. 2011 http://www.cdc.gov.nchstp/newsroom/docs/astfacts-msm-final508comp.pdf.
  7. Clapp JD, Reed MB, Holmes MR, Lange JE, Voas RB. Drunk in public, drunk in private: The relationship between college students, drinking environments and alcohol consumption. The American Journal of Drug and Alcohol Abuse. 2006;32:275–285. doi: 10.1080/00952990500481205. [DOI] [PubMed] [Google Scholar]
  8. Collins LM, Lanza ST. Latent class and latent transition analysis for the social, behavioral, and health sciences. Wiley; New York: 2010. [Google Scholar]
  9. Cooper A. Sexuality and the internet: Surfing into the new millennium. Cyber Psychology and Behavior. 1998;1:187–193. [Google Scholar]
  10. Corrao G, Bagnardi V, Zambon A, LaVecchia C. A meta-analysis of alcohol consumption and risk of 15 diseases. Preventive Medicine. 2004;38:613–619. doi: 10.1016/j.ypmed.2003.11.027. [DOI] [PubMed] [Google Scholar]
  11. Curran P, Harford T, Muthen B. The relationship between heavy alcohol use and bar patronage: A latent growth model. Journal of Studies on Alcohol. 1996;57:410–418. doi: 10.15288/jsa.1996.57.410. [DOI] [PubMed] [Google Scholar]
  12. Dawe S, Mattick R. Review of diagnostic screening instruments for alcohol and other drug use and other psychiatric disorders. The Australian Government Publishing Service. 1997 Publication Number 1834. [Google Scholar]
  13. Ewing JA. The CAGE questionnaire for detection of alcoholism. Summary of the original article. [Accessed 11/30/2011];Journal of the American Medical Association. 2008 300:2054–2056. doi: 10.1001/jama.2008.570. http://www.jama.ama-assn.org/content/300/17/2054.full. [Google Scholar]
  14. Geibel S, Luchters S, King'ola N, Esu-Williams E, Rinyiru A, Waimar T. Factors associated with self-reported unprotected anal sex among male sex workers in Mombasa, Kenya. Sexually Transmitted Diseases. 2008;35:746–752. doi: 10.1097/OLQ.0b013e318170589d. doi: 10.1097/OLQ.0b013e318170589d. [DOI] [PubMed] [Google Scholar]
  15. Greenwood GL, White EW, Page-Shafe K, Bein E, Osmond DH, Paul J, Stall RD. Correlates of heavy substance use among young gay and bixexual men: The San Francisco Young Men's Health Study. Drug and Alcohol Dependence. 2001;61:105–112. doi: 10.1016/s0376-8716(00)00129-0. [DOI] [PubMed] [Google Scholar]
  16. Greenfield T, Midanik L, Rogers JD. A 10-year national trend study of alcohol consumption, 1984–1995: Is the period of declining over? American Journal of Public Health. 2000;90:47–52. doi: 10.2105/ajph.90.1.47. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Harford T, Seibring M, Wechsler H. Attendance and alcohol use at parties and bars in college: A national survey of current drinkers. Journal of Studies on Alcohol. 2002;56:726–733. doi: 10.15288/jsa.2002.63.726. [DOI] [PubMed] [Google Scholar]
  18. Horvath KJ, Weinmeyer RM, Rosser BRS. Should it be illegal for HIV-positive persons to have unprotected sex without disclosure?: An examination of attitudes among US men who have sex with men and the impact of state law. AIDS Care. 2010;22:1221–1228. doi: 10.1080/09540121003668078. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Jones-Webb R, Short B, Wagenaar A, Toomey T, Murray D, Wolfson M, Forster J. Journal of Drug Education. 1997;27:67–82. doi: 10.2190/RJYG-D5C3-H2F0-GJ0L. [DOI] [PubMed] [Google Scholar]
  20. Kairouz S, Greenfield TK. A comparative multi-level analysis of contextual drinking in American and Canadian adults. Addiction. 2006;102:71–80. doi: 10.1111/j.1360-0443.2006.01655.x. [DOI] [PubMed] [Google Scholar]
  21. Lang E, Stockwell T. Drinking locations of drink-drivers: A comparative analysis of accident and nonaccident cases. Accident Analysis and Prevention. 1991;23:573–584. doi: 10.1016/0001-4575(91)90022-w. [DOI] [PubMed] [Google Scholar]
  22. Levy R, Strombeck R. Health benefits and risks of the Intenet. Journal of Medical Systems. 2002;26:495–510. doi: 10.1023/a:1020288508362. [DOI] [PubMed] [Google Scholar]
  23. Lomba L, Apóstolo J, Mendes F. Drugs and alcohol consumption and sexual behaviours in night recreational settings in Portugal. Adicciones. 2009;21:309–326. [PubMed] [Google Scholar]
  24. Macdonald S, Cherpitel CJ, DeSouza A, Stockwell T, Borges G, Giesbrecht N. Variations of alcohol impairment in different types, causes and contexts of injuries: Results of emergency room studies from 16 countries. Accident Analysis and Prevention. 2006;38:1107–1112. doi: 10.1016/j.aap.2006.04.019. [DOI] [PubMed] [Google Scholar]
  25. McLean R, Connor J. Alcohol and injury: A survey in primary care settings. The New Zealand Medical Journal. 2009;122:21–28. [PubMed] [Google Scholar]
  26. Muthén L, Muthén B. Mplus. Version 6.1 Muthen & Muthen; Los Angeles: 2009. . StataCorp . Version 11.2 Stata Corporation; College Station, TX: 2009. . Clark S, Muthén B. Relating latent class analysis results to variables not included in the analysis. Working paper. 2009
  27. Ostro D, Stall R. Alcohol, tobacco, and drug use among gay and bisexual men. In: Wolitski R, Stall R, Valdiserri R, editors. Unequal opportunity: Health disparities affecting gay and bisexual men in the United States. Oxford, University Press; New York: 2008. pp. 121–158. [Google Scholar]
  28. Rehm J, Room R, Monteiro M, Gmel G, Graham K, Rehn, Jernigan D. Alcohol as a risk factor for global burden of disease. European Addiction Research. 2003;9:157–164. doi: 10.1159/000072222. [DOI] [PubMed] [Google Scholar]
  29. Rehm J, Bond S, Sempos C, Vuong CV. Alcohol consumption and coronary heart disese morbidity and mortality. American Journal of Epidemiology. 1997;146:495–501. doi: 10.1093/oxfordjournals.aje.a009303. 1997. [DOI] [PubMed] [Google Scholar]
  30. Rhodes S. Hookups or health promotion? An exploration study of a chat room-based HIV. AIDS Education and Prevention. 2004;16:315–327. doi: 10.1521/aeap.16.4.315.40399. [DOI] [PubMed] [Google Scholar]
  31. Stall R, Paul J, Greenwood G, Pollack L, Bein E, Crosby M, Catania J. Alcohol use, drug use and alcohol-related problems among men who have sex with men: The Urban Men's Health Study. Addiction. 2001;96:1589–1601. doi: 10.1046/j.1360-0443.2001.961115896.x. [DOI] [PubMed] [Google Scholar]
  32. Stockwell T, Lang E, Rydon P. High risk drinking settings: the association of serving and promotional practices with harmful drinking. Addiction. 1993;88:1519–1526. doi: 10.1111/j.1360-0443.1993.tb03137.x. [DOI] [PubMed] [Google Scholar]
  33. Toomey T, Lenk K. A review of environmental-based community interventions. Alcohol Research & Health. 2011;34:163–166. [PMC free article] [PubMed] [Google Scholar]
  34. Trocki KF, Leigh BC. Alcohol consumption and unsafe sex: A comparison of heterosexuals and homosexual men. Journal of Acquired Immune Deficiency Syndromes. 1991;4:981–986. [PubMed] [Google Scholar]
  35. Trocki K, Drabble L. Bar patronage and motivational predictors of drinking in the San Francisco Bay area: Gender and sexual identity differences. Journal of Psychoactive Drugs. 2008;(Supplement 5):345–356. doi: 10.1080/02791072.2008.10400662. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Trocki K, Drabble L, Midanik L. Use of heavier drinking contexts among heterosexuals, homosexuals and bisexuals: Results from a national household probability survey. Journal of Studies on Alcohol. 2005;66:105–110. doi: 10.15288/jsa.2005.66.105. [DOI] [PubMed] [Google Scholar]
  37. Usdan SL, Moore CG, Schumacher JE, Talbott LL. Drinking locations prior to impaired driving among college students: Implications for prevention. Journal of American College Health. 2005;54:69–75. doi: 10.3200/JACH.54.2.69-75. [DOI] [PubMed] [Google Scholar]
  38. U.S. Department of Health and Human Services . National Survey on Drug Use and Health (2010): Volume 1, Summary of National Findings. Office of Applied Studies; Rockville, MD: 2010. Prepared by the U.S. Department of Health and Human Services, Substance Abuse and Mental Health Services Administration. [Google Scholar]
  39. VanDevanter N, Duncan A, Burrell-Piggott T, Bleakley A, Birnbaur J, Siegel K, Ramjohn D. Aids Patient Care and STDs. 2011;25:113–121. doi: 10.1089/apc.2010.0100. doi: 10:1089/apc.2010.0100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Wells S, Graham K, Speechley M, Koval JJ. Drinking patterns, drinking contexts and alcohol-related aggression among late adolescent and young adult drinkers. Society for the Study of Addiction. 2005;100:933–944. doi: 10.1111/j.1360-0443.2005.001121.x. doi: 10.1111/j.13600443.2005.01121.x. [DOI] [PubMed] [Google Scholar]
  41. Wong CF, Kipke MD, Weiss G. Risk factors for alcohol use, frequent use, and binge drinking among young men who have sex with men. Addictive Behaviors. 2008;33:1012–1020. doi: 10.1016/j.addbeh.2008.03.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Zakletskaia LI, Mundt MP, Balousek SL, Wilson EL, Fleming MF. Alcohol-impaired driving behavior and sensation-seeking disposition in a college population receiving routine care at campus health services centers. Accident Analysis and Prevention. 2009;41:380–386. doi: 10.1016/j.aap.2008.12.006. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES