Metrics of Sexual Behavior Stigma Among Cisgender Men Who Have Sex With Men in 9 Cities Across the United States

Abstract

Men who have sex with men (MSM) in the United States are stigmatized for their same-sex practices, which can lead to risky sexual behavior, potentiating risk for human immunodeficiency virus (HIV) infection. Improved measurement is necessary for accurately reporting and mitigating sexual behavior stigma. We added 13 sexual behavior stigma items to local surveys administered in 2017 at 9 sites in the Centers for Disease Control and Prevention’s National HIV Behavioral Surveillance system, which uses venue-based, time-sampling procedures to survey cisgender MSM in US Census Metropolitan Statistical Areas. We performed exploratory factor analytical procedures on site-specific (Baltimore, Maryland; Denver, Colorado; Detroit, Michigan; Houston, Texas; Nassau-Suffolk, New York; Portland, Oregon; Los Angeles, California; San Diego, California; and Virginia Beach-Norfolk, Virginia) and pooled responses to the survey items. A 3-factor solution—“stigma from family” (α = 0.70), “anticipated health-care stigma” (α = 0.75), and “general social stigma” (α = 0.66)—best fitted the pooled data and was the best-fitting solution across sites. Findings demonstrate that MSM across the United States experience sexual behavior stigma similarly. The results reflect the programmatic utility of enhanced stigma measurement, including tracking trends in stigma over time, making regional comparisons of stigma burden, and supporting evaluation of stigma-mitigation interventions among MSM across the United States.

Abbreviations

HIV

human immunodeficiency virus

MSM

men who have sex with men

NHBS

National HIV Behavioral Surveillance

Worldwide, gay, bisexual, and other men who have sex with men (MSM) continue to be stigmatized for their same-sex practices (1, 2). Sexual behavior stigma, which can be anticipated, perceived, internalized, or enacted (3–5), has been linked to multiple risk factors for human immunodeficiency virus (HIV) infection and transmission, including condomless sex, low uptake of HIV testing and prevention, and low uptake of and nonadherence to treatment strategies (6–12). In the United States, where MSM make up a majority of people living with HIV and in 2018 were estimated to account for more than two-thirds of incident HIV infections (13), sexual behavior stigma remains highly prevalent in urban areas (4), where the bulk of incident HIV infections also occur (14). Research conducted by the Centers for Disease Control and Prevention’s National HIV Behavioral Surveillance (NHBS) system in 2011 and 2017 documented a high prevalence of enacted stigma among MSM in US Census Metropolitan Statistical Areas across the United States: 32–34% of MSM reported verbal discrimination; 16–23% reported discrimination in work, school, and health-care settings; and 8% reported physical assault because someone knew or assumed they were attracted to men (15, 16). Such stigma has been linked to past-year condomless anal sex, exchange sex, having 4 or more male sexual partners, and poor adherence to HIV treatment medication (15, 17, 18).

Additionally, sexual behavior stigma has been associated with physical, mental, and substance-use–related morbidity (19–24); interpersonal discord and isolation (19); and decreased access to employment/income opportunities and health care (19). Given this context, improved understanding of the nature and measurement of sexual behavior stigma across diverse US contexts is paramount for accurately documenting the burden of such stigma, tracking trends in stigma over time, and developing and evaluating stigma-mitigation interventions to improve general health and well-being. Improved sexual behavior stigma measurement can also further the goal of reducing new HIV infections as outlined in the federal government’s plan for ending the HIV epidemic in the United States (25).

Moreover, research supports the need for a more comprehensive measure of sexual behavior stigma. In a recent systematic review, Fitzgerald-Husek et al. (26) found that scales used to assess sexuality-related stigma among MSM commonly invoked sexual identity rather than sexual behavior as the stigmatized attribute of focus, inadvertently excluding MSM who engage in same-sex practices but lack a minority sexual identity. Furthermore, most of the reviewed studies on sexuality-related stigma among MSM focused exclusively on internalized stigma, with half or fewer including measurement of enacted, perceived, or anticipated stigma (26).

We sought to address these gaps by evaluating the utility of a sexual behavior stigma scale administered to MSM in the United States. The proposed 13-item scale assesses multiple types of sexual behavior stigma, including perceived, anticipated, and enacted stigma (27). The scale’s factor structure and internal consistency have been previously assessed among MSM across sub-Saharan Africa and in a nationwide, online sample of MSM across the United States (4, 27). We aimed to further assess the utility of the scale by exploring whether or not its factor structure and other relevant psychometric properties vary by US region, which will build on extant evidence to support the scale’s implementation across contexts (24). Given the differences in sampling between our prior work with this scale and the present study—a nationwide sample of MSM recruited online versus an urban, primarily coastal sample of MSM recruited in places where MSM gather socially—we performed exploratory rather than confirmatory factor analyses to allow for the possibility that the factor structure might vary from previous findings.

METHODS

Data source

We invited administrators at all 22 NHBS sites to include 13 sexual behavior stigma items in their 2017 surveillance efforts (Table 1). Along with the standard set of questions included in all NHBS surveys, individual sites can include additional questions on topics of local interest, which is the mechanism through which sites could include our sexual behavior stigma items in their surveys. Eleven NHBS sites included the items, and 9 shared their data: Baltimore, Maryland; Denver, Colorado; Detroit, Michigan; Houston, Texas; Los Angeles, California; Nassau-Suffolk, New York; Portland, Oregon; San Diego, California; and Virginia Beach-Norfolk, Virginia. Eight sites (n = 3,614) included all 13 items, and 1 site (Baltimore; n = 472) included 8 items. Data-use agreements were signed by all parties, and an institutional review board at each site approved the study. This secondary analysis of deidentified data from all sites received an exemption from Johns Hopkins University’s institutional review board.

Table 1.

Sexual Behavior Stigma Survey Items Administered to Cisgender Men Who Have Sex With Men in 9 US Metropolitan Statistical Areas, 2017

Stigma Item	Stigma Type
1. Have you ever felt excluded from family activities because you have sex with men?	Perceived
2. Have you ever felt that family members have made discriminatory remarks or gossiped about you because you have sex with men?	Perceived
3. Have you ever felt rejected by your friends because you have sex with men?	Perceived
4. Have you ever felt afraid to go to health-care services because you have sex with men?	Anticipated
5. Have you ever avoided going to health-care services because you have sex with men?	Anticipated
6. Have you ever heard health-care providers gossiping about you (talking about you) because you have sex with men?	Enacted
7. Have you ever felt that you were not treated well in a health center because you have sex with men?	Perceived
8. Have you ever felt that the police refused to protect you because you have sex with men?	Perceived
9. Have you ever felt scared to be in public places because you have sex with men?	Anticipated
10. Have you ever been verbally harassed and felt it was because you have sex with men?	Enacted
11. Have you ever been blackmailed by someone because you have sex with men?	Enacted
12. Has someone ever physically hurt you (pushed, shoved, slapped, hit, kicked, choked or otherwise physically hurt you)? [AND] Do you believe any of these experiences of physical violence was/were related to the fact that you have sex with men?	Enacted
13. Have you ever been forced to have sex when you did not want to? (By forced, I mean physically forced, coerced to have sex, or penetrated with an object, when you did not want to). [AND] Do you believe any of these experiences of sexual violence were related to the fact that you have sex with men?	Enacted

Sampling procedures and participants

State and local health departments implemented a standardized NHBS protocol provided by the Centers for Disease Control and Prevention and used venue-based, time-sampling procedures to survey cisgender MSM in Metropolitan Statistical Areas. To recruit MSM at each site, NHBS staff identified venues and events (e.g., bars, retail establishments, cafes/restaurants, bathhouses, parks, pride events) attended by MSM and days and times during which MSM were likely to be present at those venues. Each month, staff randomly selected venues, days, and time periods for recruitment. Eligibility criteria for participation in the 2017 NHBS survey included age ≥18 years, assigned male sex at birth, current self-identification as male, residence in the respective Metropolitan Statistical Area, any lifetime history of oral or anal sex with another man, and ability to complete the survey in English or Spanish. Trained interviewers administered standardized surveys in person via handheld computer-assisted personal interview and offered HIV testing. Participants received separate compensation for survey completion and HIV testing.

Measures

Participants responded to 13 sexual behavior stigma survey items (Table 1) that had been developed using a socioecological framework through prior research with MSM in sub-Saharan Africa and that have since been used with MSM in the United States (4, 27). Additional details on item development have been published elsewhere (27). Items assessed experiences of perceived, anticipated, and enacted sexual behavior stigma in social, health-care, and community contexts. Response options included “yes, in the last 6 months”; “yes, but not in the last 6 months”; and “no.” For analysis, responses were dichotomized by collapsing the affirmative responses. Items 12 and 13 assessed enacted stigma in the form of physical and sexual violence, and each included 2 questions: whether the participant had experienced physical or sexual violence and, among those who responded affirmatively, whether they believed the violence was related to their having sex with men. Endorsement of both the experience of violence and the belief that it was related to having sex with men was coded as affirmative, versus no experience or experience without attribution to sexual behavior. In addition, sociodemographic characteristics (age, race/ethnicity, education) and sexuality disclosure to family, health-care providers, and non–gay/lesbian/bisexual friends were ascertained.

Analyses

For each site, item-level missingness was assessed, and descriptive statistics were calculated for sociodemographic characteristics and stigma items. The Kaiser-Meyer-Olkin test of sampling adequacy (28), which measures the proportion of variance in variables (e.g., stigma items) that may be driven by underlying factors, was performed to assess suitability for factor analysis; a Kaiser-Meyer-Olkin score greater than or equal to 0.50 indicates adequate sampling to detect underlying factors and suitability for factor analysis. A principal components analysis was then conducted on a tetrachoric correlation matrix given dichotomous response options for items. Next, a scree plot was generated, and a parallel analysis was performed. The results of the principal components analysis, scree plot, parallel analysis, and scientific interpretation were weighed to determine the number of factors to extract in exploratory factor analysis with geomin factor rotation (due to expected item cross-loadings and factor correlations) and robust weighted least-squares estimation (due to fewer convergence problems) (29–31).

After extracting factors for each site, we examined item loadings for each factor, assessing the strength of each loading (≥0.40), low- or cross-loading items (items loaded strongly and similarly on multiple factors (i.e., with a difference of ≤0.20 and with at least 1 loading ≥0.40)), and factor interpretability (32). We compared factor solutions across sites and examined the following fit statistics for each: root mean squared error of approximation < 0.05; comparative fit index > 0.90; Tucker-Lewis index > 0.90; and standardized root mean squared residual < 0.08 (33–36). Final model selection for each site was based on the number of strongly loading items per factor, interpretability, parsimony, and fit indices.

Next, data from all sites except Baltimore were pooled, since 5 items from the sexual behavior stigma scale were excluded at that site, and the procedures described above were repeated. Cronbach’s α (≥0.60 was considered adequate) and inter-item correlations were calculated to assess the internal consistency of each factor in the pooled and site-specific data sets. No data were imputed. All analyses were conducted in Stata, version 15 (37), and Mplus, version 8 (30).

RESULTS

Sample characteristics

Participants were from the West (Denver, Portland, San Diego, and Los Angeles; 2,056/4,086 (50%)), South (Baltimore, Virginia Beach-Norfolk, and Houston; 1,382/4,086 (34%)), Midwest (Detroit; 507/4,086 (12%)), and Northeast (Nassau-Suffolk; 141/4,086 (4%)) Census regions of the United States. The median age was 32 years, and median ages ranged from 30 years to 35 years across sites. Approximately 38% (1,567/4,086) of participants reported having attended some college, ranging from 32% (178/554) in Denver to 48% (175/367) in Virginia Beach-Norfolk. Participants reporting White race comprised 49% (2,004/4,086) of the sample, ranging from 17% (81/472) in Baltimore to 75% (298/397) in Portland. About one-quarter (948/4,086) reported being of Hispanic, Latino, or Spanish origin, ranging from 5% (23/472) in Baltimore to more than 40% (214/524) in Los Angeles. Over 80% of participants had disclosed their sexuality to family members (3,517/4,086), health-care providers (3,360/4,086), and non–gay/lesbian/bisexual friends (3,630/4,086). Disclosure was highest in Portland and San Diego and lowest in Baltimore and Detroit (Table 2).

Table 2.

Characteristics of Cisgender Men Who Have Sex With Men in 9 US Metropolitan Statistical Areas, by Data Set, 2017

	National HIV Behavioral Surveillance Site
	Pooled Data(All 9 Sites) (n = 4,086)		Baltimore, Maryland (n = 472)		Denver, Colorado (n = 554)		Detroit, Michigan (n = 507)		Houston, Texas (n = 543)		Los Angeles, California (n = 524)		Nassau-Suffolk, New York (n = 141)		Portland, Oregon (n = 397)		San Diego, California (n = 581)		Virginia Beach-Norfolk, Virginia (n = 367)
Characteristic	No.	%	No.	%	No.	%	No.	%	No.	%	No.	%	No.	%	No.	%	No.	%	No.	%
Age, yearsa	32 (26–44)		33 (28–45)		32 (26–44)		32 (25–48)		33 (27–43)		30 (26–37)		32 (25–47)		34 (28–46)		35 (28–46)		30 (25–39)
Education
High school or less	1,013	24.8	196	41.5	116	20.9	208	41.0	114	21.0	107	20.4	28	19.9	60	15.1	80	13.8	104	28.3
Some college	1,567	38.4	154	32.6	178	32.1	202	39.8	204	37.6	202	38.5	67	47.5	134	33.8	251	43.2	175	47.7
Bachelor’s degree	1,028	25.2	88	18.6	179	32.3	73	14.4	155	28.5	151	28.8	28	19.9	131	33.0	167	28.7	56	15.3
Graduate school	474	11.6	34	7.2	79	14.3	22	4.3	70	12.9	64	12.2	18	12.8	72	18.1	83	14.3	32	8.7
Missing data	4	0.1	0	0.0	2	0.4	2	0.4	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0
Race
American Indian or Alaska Native	132	3.2	2	0.4	27	4.9	4	0.8	21	3.9	41	7.8	3	2.1	7	1.8	25	4.3	2	0.5
Asian	89	2.2	10	2.1	11	2.0	7	1.4	9	1.7	16	3.1	1	0.7	10	2.5	22	3.8	3	0.8
Black/African-American	1,291	31.6	332	70.3	43	7.8	311	61.3	160	29.5	130	24.8	14	9.9	16	4.0	58	10.0	227	61.9
Native Hawaiian or other Pacific Islander	37	0.9	1	0.2	8	1.4	4	0.8	0	0.0	6	1.1	4	2.8	2	0.5	8	1.4	4	1.1
White	2,004	49.0	81	17.2	407	73.5	147	29.0	320	58.9	216	41.2	104	73.8	298	75.1	337	58.0	94	25.6
Multiracial	325	8.0	46	9.7	47	8.5	25	4.9	23	4.2	43	8.2	6	4.3	46	11.6	53	9.1	36	9.8
Missing data	208	5.1	0	0.0	11	2.0	9	1.8	10	1.8	72	13.7	9	6.4	18	4.5	78	13.4	1	0.3
Ethnicity
Hispanic/Latino/Spanish	948	23.2	23	4.9	142	25.6	26	5.1	178	32.8	214	40.8	51	36.2	62	15.6	221	38.0	31	8.4
Missing data	2	0.1b	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0	1	0.2	1	0.3
Sexuality disclosure
Non–gay/lesbian/bisexual friends	3,630	88.8	365	77.3	513	92.6	396	78.1	489	90.1	484	92.4	125	88.7	370	93.2	549	94.5	339	92.4
Family members	3,517	86.1	355	75.2	490	88.4	392	77.3	477	87.8	470	89.7	122	86.5	361	90.9	526	90.5	324	88.3
Health-care providers	3,360	82.2	342	72.5	480	86.6	365	72.0	445	82.0	446	85.1	107	75.9	353	88.9	520	89.5	302	82.3
No one/never disclosed	238	5.8	74	15.7	19	3.4	60	11.8	29	5.3	14	2.7	11	7.8	10	2.5	11	1.9	10	2.7
Missing data	4	0.1	0	0.0	1	0.2	2	0.4	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0	1	0.3

Abbreviation: HIV, human immunodeficiency virus.

^a Values are expressed as median (interquartile range).

^b Value was rounded up from 0.049.

Item endorsement

Per participant, missingness of data for any stigma scale item was 1% of participants (n = 40), with the majority of those with any missing scale response missing only 1 item (n = 32). Per stigma item, missingness was less than 1%. Two-thirds of participants endorsed 1 or more lifetime stigma experiences. Endorsement was highest for gossip/discriminatory remarks by family (44%), reaching 50% in Los Angeles, and verbal harassment (43%), reaching 60% in Portland. Endorsement was lowest for having been gossiped about by health-care providers, which was less than 6% overall and across sites (see Web Table 1, available at https://doi.org/10.1093/aje/kwab240).

Exploratory factor analyses

Three sites (Nassau-Suffolk, Detroit, and San Diego) required removal of items to run principal components analysis on the tetrachoric correlation matrix due to cells with small values, and 2 sites (Denver and Nassau-Suffolk) required removal of items to increase the Kaiser-Meyer-Olkin score to more than 0.50. Nassau-Suffolk was subsequently left with only 4 items and consequently was not subjected to individual factor analysis, though the data set was included in the pooled analysis.

Although a 2-factor model was indicated for Baltimore, where 5 scale items were not administered, a 3-factor model was the most commonly indicated solution across the remaining 7 sites (Web Table 2) and demonstrated good fit (Table 3). Items 1 and 2 (excluded by family, gossiped about/heard discriminatory remarks by family) loaded on factor 1 at all sites but Houston. Items 4 and 5 (feared/avoided going to health-care services) loaded on factor 2 at all sites but Houston. Items 8–10 (felt police refused to protect, feared being in public, verbally harassed) and 12 and 13 (experienced physical/sexual violence) tended to load on factor 3 at most sites, except for Houston and Los Angeles. Items 3 (rejected by friends), 6 (felt mistreated in a health center), 7 (gossiped about by providers), and 11 (blackmailed) had low loadings, cross-loaded, did not consistently load on the same factor, or had been excluded previously because of small cell values (Web Table 3).

Table 3.

Fit Statistics for 3-Factor Exploratory Factor Analyses in a Study of Sexual Behavior Stigma Among Cisgender Men Who Have Sex With Men in 8 US Metropolitan Statistical Areas, by Data Set, 2017

			Test of Model Fit			RMSEA b
Data Set a	TotalNo. of Participants	Excluded Variable	ꭓ 2 Value	Degrees of Freedom	P Value	Estimate	90% CI	Comparative Fit Index	Tucker-Lewis Index	SRMR
Pooled data (8 sites)	3,614	None	228.507	42	<0.001	0.035	0.031, 0.040	0.988	0.977	0.047
Baltimore, Maryland	472	None	6.842	7	0.446	0.000	0.000, 0.056	1.000	1.000	0.030
Denver, Colorado	554	Item 11	44.450	33	0.088	0.025	0.000, 0.042	0.995	0.989	0.052
Detroit, Michigan	507	Item 11	45.023	33	0.079	0.027	0.000, 0.045	0.993	0.986	0.061
Houston, Texas	543	None	62.314	42	0.023	0.030	0.012, 0.045	0.992	0.986	0.059
Los Angeles, California	524	None	67.877	42	0.007	0.034	0.018, 0.040	0.984	0.970	0.065
Portland, Oregon	397	None	49.038	42	0.212	0.021	0.000, 0.041	0.997	0.994	0.063
San Diego, California	581	Item 7	31.307	33	0.552	0.000	0.000, 0.028	1.000	1.002	0.046
Virginia Beach-Norfolk, Virginia	367	None	56.667	42	0.065	0.031	0.000, 0.050	0.991	0.983	0.067

Abbreviations: CI, confidence interval; RMSEA, root mean squared error of approximation; SRMR, standardized root mean squared residual.

^a After preliminary factor-analytical procedures, Nassau-Suffolk, New York (n = 141) was left with only 4 items and consequently was not subjected to individual factor analysis. Because 5 of the 13 items from the sexual behavior stigma scale were not included in Baltimore’s local survey, the Baltimore data set was excluded from the pooled analysis.

^b Thresholds used to assess fit: RMSEA, <0.05; comparative fit index and Tucker-Lewis index, >0.90; SRMR, <0.08.

In the pooled data set, sampling adequacy was meritorious (Kaiser-Meyer-Olkin score = 0.85), indicating suitability for factor analysis. We extracted 3 factors (Web Table 2), and a 3-factor solution exhibited good fit (Table 3). Items 1 and 2 loaded on factor 1, which was named “stigma from family,” as the items comprising it assessed sexuality-based exclusion from family activities and gossip/discriminatory remarks by family. Items 4 and 5 loaded on factor 2, which was named “anticipated health-care stigma,” as the items comprising it assessed anticipatory fear and avoidance of health care due to worry that providers would learn about one’s sexuality. Items 8–10 and items 12 and 13 loaded on factor 3, which was named “general social stigma,” as the items comprising it assessed a range of negative social encounters (e.g., felt police refused to protect, verbally harassed, experienced physical/sexual violence) and were less context-specific (e.g., in public; unspecified perpetrators of harassment/violence). Significant correlations were found between “stigma from family” and “anticipated health-care stigma” (r = 0.41), “stigma from family” and “general social stigma” (r = 0.69), and “anticipated health-care stigma” and “general social stigma” (r = 0.44). Similar to site-specific analyses, loadings for items 3, 7, and 11 were less than 0.40 for all factors, and item 6 cross-loaded on factors 2 and 3 (Web Table 3).

Internal consistency

For “stigma from family,” internal consistency was adequate in the pooled sample (α = 0.70) and at all sites (α = 0.64–0.75). For “anticipated health-care stigma,” internal consistency was adequate in the pooled sample (α = 0.75) and at all sites (α = 0.61–0.80). For “general social stigma,” internal consistency was adequate in the pooled sample (α = 0.66) and at 8 sites (α = 0.62–0.68); the internal consistency of “general social stigma” was not assessed in the Baltimore data set because of the limited number of items included (Table 4).

Table 4.

Cronbach’s α and Average Inter-Item Correlations for a 3-Factor Model of Sexual Behavior Stigma Among Cisgender Men Who Have Sex With Men in 9 US Metropolitan Statistical Areas, by Data Set, 2017

	Factors
		Stigma From Family (Items 1 and 2)		Anticipated Health-Care Stigma (Items 4 and 5)		General Social Stigma (Items 8–10 andItems 12 and 13)
Data Set	Total No. of Participants	Cronbach’s α	IIC	Cronbach’s α	IIC	Cronbach’s α	IIC
Pooled data (8 sites)a	3,614	0.70	0.54	0.75	0.60	0.66	0.28
Baltimore, Marylandb	472	0.72	0.57	0.79	0.65
Denver, Colorado	554	0.72	0.56	0.72	0.56	0.65	0.27
Detroit, Michigan	507	0.75	0.60	0.77	0.63	0.67	0.29
Houston, Texas	543	0.72	0.56	0.80	0.67	0.65	0.27
Los Angeles, California	524	0.67	0.51	0.74	0.59	0.62	0.25
Nassau-Suffolk, New York	141	0.75	0.60	0.61	0.44	0.64	0.38
Portland, Oregon	397	0.74	0.59	0.73	0.58	0.66	0.28
San Diego, California	581	0.64	0.47	0.74	0.59	0.68	0.30
Virginia Beach-Norfolk, Virginia	367	0.65	0.48	0.80	0.28	0.66	0.28

Abbreviation: IIC, inter-item correlation.

^a The Baltimore data set was excluded from the pooled analysis here; see “Sensitivity Analyses” section of the text for Cronbach’s α calculated with the Baltimore data set included.

^b Cronbach’s α was not calculated for “general social stigma” using the Baltimore data set, as the majority of survey items comprising this factor were not administered there.

Sensitivity analyses

Patterns in factor loadings seen in the 3-factor solutions of other sites were not reflected in the 3-factor solutions of Los Angeles or Houston. Therefore, 4-factor solutions were examined at these sites. In Los Angeles, items 1 and 2 loaded on 1 factor and items 9, 10, 12, and 13 loaded on 1 factor resembling “stigma from family” and “general social stigma” from the main analysis, respectively. Items 3–5 loaded on 1 factor, approximating the “anticipated health-care stigma” factor from the main analysis; item 3’s loading with items 4 and 5 may reflect rejection-focused expectations and perceptions (38). Items 6 and 7 comprised their own factor, indicating that perceived and enacted health-care stigma comprises a distinct stigma domain for MSM in Los Angeles. In Houston, items 1 and 2 loaded on 1 factor and items 9, 10, and 12 loaded on 1 factor, resembling “stigma from family” and “general social stigma” from the main analysis, respectively. The “anticipated health-care stigma” factor did not emerge, as all health-care stigma items loaded on 1 factor, suggesting a broader “health-care stigma” construct. Context rather than type of stigma may be driving the constitution of that factor. Items 4 and 9 cross-loaded to form a fourth factor, suggesting that fear-related stigma may constitute an important stigma domain for MSM in Houston (Web Table 4).

We performed exploratory factor analyses on the pooled data set with the Baltimore data set included. We extracted 3 factors (Web Table 2), a 3-factor solution exhibited good fit, and the same 3-factor structure emerged: “stigma from family” (α = 0.70), comprised of items 1 and 2; “anticipated health-care stigma” (α = 0.76), comprised of items 4 and 5; and “general social stigma” (α = 0.66), comprised of items 8–10 and items 12 and 13. Complete-case analyses including and excluding Baltimore revealed comparable factor structures (Web Tables 2, 5, and 6).

Finally, we performed age-stratified exploratory factor analyses to explore potential variation in the factor structure of the sexual behavior stigma items across the life course. The same factor structure was approximated across age strata (18–29, 30–39, 40–49, and ≥50 years), with a few exceptions. In the age stratum 18–29 years, item 7 loaded on the “anticipated health-care stigma” factor. Similarly, in the age stratum ≥50 years, items 6 and 7 loaded on the “anticipated health-care stigma” factor, while they loaded on the “general social stigma” factor among persons aged 30–39 years. In the age stratum 40–49 years, item 11—which did not load meaningfully on any factor in any other analysis—loaded on the “general social stigma” factor (Web Tables 7–9).

DISCUSSION

We explored the factor structure and psychometric properties of a 13-item sexual behavior stigma scale and assessed its utility for use with cisgender MSM at 9 NHBS sites across the United States. We identified 3 factors underlying sexual behavior stigma—“stigma from family,” “anticipated health-care stigma,” and “general social stigma”—comprised of 9 sexual behavior stigma items that exhibited a more consistent factor structure than the original 13 items. These findings are consistent with the 3-factor structure found previously with this scale from a nationwide online convenience sample of more than 3,000 MSM (39). Slight dissimilarities between the studies’ findings (discussed below) may reflect differences in sampling and sociodemographic characteristics. Specifically, NHBS participants resided in urban, primarily coastal settings in the West or South, were recruited from venues of MSM social congregation, and were more racially/ethnically diverse (4), possibly illustrating how structural and contextual influences in such settings shape stigma experiences.

“Stigma from family” features items reflecting the social context in which stigma was experienced, demonstrating that negative encounters with family represent a distinct stigma domain for MSM. “Stigma from family” also features items reflecting the same type of stigma (perceived), though this does not preclude the possibility of the stigma’s having actually been enacted. Perceived stigma in health-care settings and perceived stigma in public places did not load on “stigma from family,” mirroring our prior US work (24). In addition, perceived stigma from friends did not strongly or distinctly load with “stigma from family,” which contrasts with our prior US work (27). Instead, and despite being below the 0.40 threshold at several sites, perceived stigma from friends loaded comparably on “stigma from family” and “anticipated health-care stigma” across pooled analysis and most individual analyses. Many MSM, especially MSM of Color, have clearly defined families of choice comprised of nonbiological relationships (40, 41); therefore, such similar, overlapping social circles may have influenced perceived stigma from friends to load on “stigma from family.” The comparable loading on “anticipated health-care stigma” may reflect rejection sensitivities (38), with perceived stigma from friends being explicitly rejection-focused and “anticipated health-care stigma” being implicitly or indirectly rejection-focused. Perceived stigma from friends may hold less salience for MSM in our NHBS sample as compared with our prior online sample, suggesting differential experiences of sexual behavior stigma by geography, urbanicity, race/ethnicity, and/or sampling method (4, 42–44). Stigma from friends may be less relevant to measure in urban, primarily coastal settings or for MSM who frequent venues of MSM social congregation, since both contexts may provide greater access to networks of other sexual minority men and the creation of alternative family structures without risk of sexuality-based rejection.

Items comprising “anticipated health-care stigma” also reflect both the context of the stigma (health care) and the type of stigma experienced (anticipated), replicating our prior US work (27). “Anticipated health-care stigma” illustrates how individuals may become primed to anxiously anticipate stigma-related rejection, which may drive them to avoid situations, such as health-care encounters, where such rejection is possible (38). However, given the racial/ethnic diversity of our sample, the extent to which “anticipated health-care stigma,” as measured here, is distinct from the medical mistrust commonly reported by racial/ethnic minority MSM may be important to consider in future research (45, 46). Dropped perceived and enacted health-care stigma items—feeling mistreated in a health center (item 6), being gossiped about by health-care providers (item 7)—partially loaded on “anticipated health-care stigma,” a reflection of similarity in context, and partially loaded on “general social stigma,” a reflection of similarity in content. In our prior US work, the same cross-loading pattern was observed for item 6 but not for item 7, which prominently loaded on factor 3 (27). This may reflect a lower risk of experiencing enacted health-care stigma following sexuality disclosure in urban, primarily coastal settings (7), since enacted health-care stigma was lower (5% vs. 7%) despite sexuality disclosure’s being higher (82% vs. 72%) in the present study relative to our prior US work (27). While none of the health-care stigma items were commonly endorsed, compared with anticipated health-care stigma, perceived and enacted health-care stigma may lack particular salience for MSM in urban, primarily coastal settings, where stigma-mitigation interventions in health-care contexts have likely been implemented, and may therefore be poor indicators of sexual behavior stigma there.

Items comprising “general social stigma” spanned multiple social contexts (interpersonal, community) and included multiple stigma types (perceived, anticipated, enacted). Items reflecting perceived/anticipated stigma (feared being in public places, felt police refused to provide help) may have covaried with those reflecting enacted stigma (physical/sexual violence, verbally harassed) due to their common relationship to victimization, as such experiences may lead to fear of public places and warrant requests for police assistance, providing an opportunity for negative police encounters. Factor loadings and endorsement of “general social stigma” items varied widely, demonstrating how site-specific social and structural characteristics may shape stigma experiences. For example, MSM of Color in our sample, who contend with structural and interpersonal racism on top of sexual behavior stigma, may experience many of the social stigmas more frequently and intensely than non-Hispanic White MSM (47–50). Additionally, endorsement of “general social stigma” items was highest in Portland, where disclosure was also highest, and lowest in Detroit, where disclosure was also lowest, possibly reflecting how sexuality disclosure can lead to enacted stigma (7). Overall, the factor loading for item 11 (blackmailed) on “general social stigma” was comparable to our prior US work (27), which we deemed low in the current study, but appeared to be marginally salient for MSM aged 40–49 years. Blackmail motivated by homonegativity has been widely documented in sub-Saharan Africa (51) but may be less commonly encountered by MSM in the United States, though differences by urbanicity have been found (4). Blackmail motivated by homonegativity may be a poor indicator of sexual behavior stigma among MSM in the United States.

Limitations

Our available-case approach and complete-case sensitivity analyses could have biased our findings. However, any potential bias would have been negligible given the very low level of missingness (52). Second, 4 site-specific samples required removal of some items in order to conduct preliminary factor analytical procedures, preventing full-scale analysis at those sites. Third, though the survey items were crafted to assess sexual behavior stigma, some participants may have responded in reference to their sexual identity or attraction instead. Particularities related to sexual identities, such as their politicization, may result in different stigma experiences than sexuality-related stigma centered on sexual behavior. Similarly, “stigma from family” was comprised of items that, while referencing one’s family as the stigma source, did not actually define “family.” This could have resulted in misclassification bias, since MSM could have considered either their biological family or their chosen family (if applicable) when responding to the items. Fourth, despite high internal consistency, “stigma from family” and “anticipated health-care stigma” were each comprised of only 2 items, which may weaken factor reliability. Future research could explore whether additional experiences related to these constructs could be assessed to improve the scale’s sensitivity and discrimination between individuals who experience different levels of these types of stigma. Fifth, variability in sociodemographic characteristics and sexuality disclosure patterns across sites may have driven differential experiences of stigma, requiring additional research to inform how stigma may be shaped by these factors. Similarly, MSM recruited from different venues (e.g., parks vs. pride events) could experience stigma differently, potentially resulting in selection bias. Sixth, selecting the best-fitting structure to apply across all sites may have masked unique differences in factors at some sites, especially Houston, possibly driven by its unique sociopolitical, geographical context relative to the other sites. Likewise, differences in stigma factors by age strata may have been masked by our choice to examine within-lifetime experiences of stigma only. Finally, all sites were urban and most were coastal, raising concerns about generalizability to rural interior US areas. Future research should assess the scale’s construct validity and examine how/whether this may vary by urbanicity, race/ethnicity, age stratum, or geography.

Conclusion

Taken together with prior US work on this sexual behavior stigma scale, findings from the current study indicate that MSM across the United States encounter sexual behavior stigma similarly, though perceived stigma from friends and enacted health-care stigma may have less salience or be experienced differently by MSM in urban, primarily coastal settings sampled in venues, relative to MSM in mixed urban and nonurban Middle-American settings sampled online. Such stigmas may therefore be more or less relevant to measure and mitigate depending on the context and sampling method. The otherwise congruent factor structure documented across diverse contexts and sampling strategies provides evidence for the scale’s reliable measurement of sexual behavior stigma, lends support for the scale’s wider implementation, and fills the need for improved measurement of sexual behavior stigma experienced by MSM in the United States. Findings reflect the programmatic utility of enhanced measurement of stigma, including tracking trends over time, making regional comparisons of stigma burden, and supporting the evaluation of stigma-mitigation interventions among MSM across the United States.