Differences in HIV Risk and Prevention among Cisgender Latino Sexual Minority Men by Language of Online Survey Completion: Analysis of National and Washington State Data

Abstract

While online or internet-based surveys can increase access to non-English speaking individuals, it is unclear whether the availability of completing online surveys in Spanish enhances the diversity of research participants and supports the inclusion of individuals at higher risk for HIV among Latino sexual minority men (SMM). We sought to examine how language of online survey completion among cisgender Latino SMM was associated with indicators of HIV risk and prevention. We analyzed national and Washington State data using the Understanding New Infections through Targeted Epidemiology (UNITE) Cohort Study (2017 and 2018) and the Washington HIV/STI Prevention Project (WHSPP) survey (2017 and 2018/2019), respectively. While many indicators of HIV prevention and risk were similar by language of survey completion, our analyses indicated several important differences. Latino SMM who completed online surveys in Spanish differed from those who completed surveys in English across several sociodemographic characteristics. After adjusting for sociodemographic characteristics and HIV-related risk factors, Spanish language respondents in UNITE were less likely to have tested for HIV in the past year and those in WHSPP were more likely to report a recent STI diagnosis. Findings suggest that Latino SMM who complete surveys in Spanish comprise a unique subgroup that may have a specific HIV health and risk behavior profile. Our results suggest a need for increased and tailored efforts to recruit and include Spanish speaking Latino SMM for local and national internet-based studies.

INTRODUCTION

Latino gay, bisexual, and other sexual minority men (SMM) are disproportionately affected by HIV in the United States (U.S.) (1). Among all new HIV diagnoses in 2018, one in five occurred in Latino SMM (2). Despite major advancements in HIV treatment and prevention, Latino SMM have not experienced declines in HIV diagnosis rates over the past decade (3). Language is one factor that may shape access to HIV prevention among Latino SMM (4). Specifically, language barriers can pose challenges to obtaining adequate and appropriate HIV prevention information and can also limit opportunities for engaging with providers and receiving care (5, 6).

While language has been frequently cited as a barrier to accessing health and HIV prevention services (4, 7–9), it is often confounded by sociodemographic characteristics (i.e. poverty, education) or other factors such as health insurance, when predicting HIV prevention outcomes. Further, language is closely related to social processes such as acculturation or assimilation, which have been shown to influence HIV infection risk (10, 11). Given that an important component of tailoring HIV services and programs involves delivery in individuals’ preferred language (12), overlooking language points to a missed opportunity for enhancing HIV prevention efforts for Latino SMM (13).

A valuable step towards meeting the linguistic needs of populations who are at high risk for HIV is ensuring that HIV research and data collection efforts that inform the development of prevention programs are conducted in multiple languages. Given language limitations and challenges to recruiting and training bilingual staff for in-person studies, both large and small research studies and datasets in the U.S. have tended to exclude non-English speaking people (14–16). The availability and use of online or internet-based surveys, however, has reduced such challenges while also potentially increasing access to individuals who speak a language other than English. Notably, internet-based research has become increasingly common and has demonstrated to be effective at reaching SMM populations (17–19). As approximately 73 percent of Latinos in the U.S. speak Spanish at home (20), conducting online surveys in Spanish may be critical for public health planning and evaluation of trends in HIV behavior and prevention for Latino SMM.

Yet, whether providing an option to complete online surveys in Spanish does indeed increase the diversity of research participants and supports the inclusion of individuals at higher HIV risk is unclear. Further, given the costs of translating survey instruments and recruitment materials and procedures, greater understanding of whether language of online survey completion is associated with determinants of HIV risk and prevention is needed to inform strategies for improving both research and HIV prevention strategies for Latino SMM. The present study sought to address this gap and examine how cisgender Latino SMM who completed online surveys in Spanish vs. English differed across sociodemographic characteristics, HIV-related risk behaviors, and HIV prevention behaviors.

METHODS

Data source and population

We used data from the 2017 and 2018/2019 Washington HIV/STI Prevention Project (WHSPP) and baseline data (2017 and 2018) from the Understanding New Infections through Targeted Epidemiology (UNITE) Cohort Study. Separate analyses of these datasets were conducted to address our study aims using a local (WHSPP) and national (UNITE) perspective.

The WHSPP is a cross-sectional online survey that collected HIV prevention data from SMM in Washington State (21, 22). Respondents were recruited through banner and text-based pop-up advertisements on social media, geospatial sexual networking applications (apps), and general lesbian, bisexual, transgender, and queer (LGBTQ)-interest websites and online apps. Recruitment targeted male and transgender individuals who were 16 years of age or older who reported having ever had sex with a man. The first round of the WHSPP was conducted from January to February 2017 with a sample of 1236, and the second was conducted from November 2018 to January 2019 with a sample of 1218. In the 2017 round, respondents were randomly offered either a $10 Amazon gift certificate, $10 donation to a charitable organization, or no monetary incentive for participation. The $10 Amazon incentive arm was discontinued early in the course of data collection due to fraudulent responses. In 2018/2019, participants were offered the opportunity to select a charitable organization to which the study would donate $5 to as an incentive. We combined the 2017 and 2018/2019 WHSPP data for analyses and removed participants in the 2018/2019 round who indicated that they participated in the 2017 survey. Additional details about the WHSPP methodology can be found elsewhere (21–23).

The UNITE study is a national online cohort study designed to examine factors that influence HIV seroconversion among HIV negative SMM (24). UNITE collected online survey responses from SMM in the U.S. between November 2017 through September 2018 (n=7956). Internet-based recruitment strategies across diverse virtual venues, including geospatial sexual networking apps, social media sites, and email blasts were used. Individuals were eligible if they were at least 16 years of age, reported HIV negative or unknown status, identified as male/masculine, identified as a sexual minority (i.e., gay, queer, bisexual), had a mailing address where packages could be received in the U.S. or Puerto Rico, were recruited from geospatial sexual networking apps or reported a history of using geospatial networking apps to meet partners in the past six months, reported willingness to complete at-home HIV and STI testing, allowed contact information to be used for mailing test kits and compensation delivery, and indicated for PrEP based on Centers for Disease Control and Prevention (CDC) guidelines (25). Specifically, participants had to report the presence of at least one of the following in the past 6 months: 1) self-reported diagnosis of a sexually transmitted infection (STI); 2) having been prescribed post-exposure prophylaxis (PEP); 3) reported condomless anal sex with a casual male partner of any HIV status or with an HIV-positive or unknown status main partner; 4) reported condomless anal sex with a main partner in the context of a non-monogamous partnership where the partner was having condomless anal sex with other partners (25). Individuals who reported being on PrEP were only eligible if they indicated suboptimal adherence to PrEP as this would place them at risk for HIV. Suboptimal PrEP adherence was defined as either missing four or more days of dosing in a row or self-report of fair, poor, or very poor adherence.

Study measures

Both the WHSPP and UNITE projects posted advertisements in English and Spanish, and participants could change the survey language after clicking on an advertisement. For both samples, we classified participants according to the language of survey completion (Spanish or English) and their sociodemographic characteristics, including age, race/ethnicity, sexual identity, education, annual household income, and health insurance.

While both surveys measured similar HIV-related behaviors, there were some differences in the recall period and measures used. For WHSPP, we examined the following behaviors reported by participants in the past 12 months: number of male sex partners, condomless anal sex with unknown status or HIV positive partners, drug use (poppers, meth, or injecting drugs), and exchange of drugs or money for sex. For UNITE, we assessed behaviors that were reported in the past three months, which included: number of casual male sex partners, engaging in condomless anal sex with unknown or HIV positive status casual partners (receptive, insertive, or any), drug use (poppers, meth, injecting drugs), and exchange of sex for goods (e.g. money, drugs, shelter, food).

For both WHSPP and UNITE, we assessed HIV testing history (never tested, tested in last year, and tested more than 1 year ago), pre-exposure prophylaxis (PrEP) use (never, past, and current), testing for STIs (in past 6 months for WHSPP and in past 3 months for UNITE), and bacterial STI diagnosis (gonorrhea, chlamydia, syphilis; in past 12 months for WHSPP and in past 6 months for UNITE). For WHSPP, we also assessed whether participants had heard of PrEP.

Analyses

Given our study’s focus on language and HIV prevention among Latino SMM, we restricted analyses of WHSPP and UNITE data to participants who reported being Latino/Hispanic, cisgender male, HIV-negative, and reported oral or anal sex with a man in the past 12 months. We conducted separate analyses of the WHSPP and UNITE data. First, we calculated descriptive statistics for sociodemographic variables, HIV risk behaviors, and prevention behaviors in WHSPP and UNITE. We used chi-square, Fisher’s exact, or Wilcoxon Mann-Whitney tests to determine whether sociodemographic characteristics and HIV risk and prevention behaviors differed by language of survey completion (Spanish vs. English). Then, we conducted multivariable Poisson regression with robust error variance to examine whether language was associated with select HIV risk and prevention outcomes: 1) having tested for HIV in the past 12 months, 2) current PrEP use (vs. never having used PrEP), and 3) any bacterial STI diagnosis (in past 12 months for WHSPP and in past 6 months for UNITE). For each outcome, we evaluated two models—one adjusting for select sociodemographic characteristics (age, education, health insurance) and recruitment platform (geospatial sexual networking app/social media versus other website or app) (Model A), and a second model that also adjusted for HIV-related risk factors (Model B). The sociodemographic characteristics included in the models were selected a priori as potential confounding variables of the association between language and HIV risk and prevention outcomes. The HIV-related risk factors included in Model B varied slightly between WHSPP and UNITE due to the differences in reference periods and variable definitions noted above —for WHSPP, we adjusted for the total number of male sex partners in the past 12 months, having had any condomless anal sex with unknown or HIV positive status partners, and reported drug use. For UNITE, HIV-related risk factors included the total number of casual male sex partners in the past 3 months, having had any condomless anal sex with unknown or HIV positive status casual partners, and reported drug use. From these models, we report adjusted risk ratios (aRR) and 95% confidence intervals (CIs). Statistical analyses for WHSPP were performed using R version 3.5.3 (26) and STATA version 13.1 (27), and analyses for UNITE were performed using SPSS v. 23 (28).

RESULTS

Sociodemographic characteristics by language of survey completion

Sociodemographic characteristics by language of online survey completion for WHSPP and UNITE are presented in Table 1. There were a total of 317 and 2099 participants in the WHSPP and UNITE surveys, respectively, that met our study’s inclusion criteria and were included in analyses. For WHSPP, 28% of included Latino participants completed surveys in Spanish, compared to 6% of Latino participants in UNITE. Across both samples, Spanish-language respondents were older, less likely to identify their race as White. In WHSPP, Spanish-language SMM respondents were less likely to have attained an education beyond high school or a GED (Table 1).

Table I:

Characteristics of Latino MSM participating in the Washington HIV/STI Prevention Project (WHSPP) (2017 and 2018/2019 rounds) and the UNITE Survey (2017–2018) by language of online survey completion

	WHSPP (n=317a)				UNITE (n= 2099)
	Spanish (n= 90a)	English (n= 227a)	Test statistic	P-valuec	Spanish (n= 117a)	English (n=1982a)	Test statistic	P-valuec
	N (%b)	N (%b)			N (%b)	N (%b)
Age (years) (M, IQR)d	35 (29, 40)	28 (22, 36)	W=13708	<0.001	32 (27, 41)	27 (23,34)	W=2042448	<0.001
Race			Fisher’s exact	<0.001			Fisher’s exact	0.023
White	34 (37.8)	118 (52.0)			42 (35.9)	877 (44.2)
Black/Afro-Latino	0 (0)	15 (6.6)			9 (7.7)	76 (3.8)
Asian and Pacific Islander	3 (3.3)	4 (1.8)			0 (0)	18 (0.9)
Native American/Alaska Native	4 (4.4)	5 (2.2)			0 (0)	66 (3.3)
Multiracial	7 (7.8)	24 (10.6)			33 (28.2)	496 (25.0)
Other/No race specifiede	42 (46.7)	61 (26.9)			33 (28.2)	449 (22.7)
Sexual Identity			Fisher’s exact	0.415			Fisher’s exact	0.132
Heterosexual or straight	1 (1.1)	1 (0.4)			-	-
Homosexual or gay	70 (79.5)	182 (80.5)			104 (88.9)	1621 (81.8)
Bisexual	17 (19.3)	38 (16.8)			10 (8.5)	299 (15.1)
Queer or some other identityf	0 (0)	5 (2.2)			3 (2.6)	62 (3.1)
Education			χ2 (3) = 19.887	<0.001			χ2 (3) = 13.351	0.004
Less than high school	8 (9.0)	10 (4.5)			2 (1.7)	76 (3.8)
High school or GED	31 (34.8)	33 (14.9)			16 (13.7)	255 (12.9)
Some college	22 (24.7)	86 (38.7)			41 (35.0)	973 (49.1)
Bachelor’s degree or higher	28 (31.5)	93 (41.9)			58 (49.6)	678 (34.2)
Annual Household Income			χ2 (3) = 19.734	<0.001 g			χ2 (3) = 4.429	0.220
Less than $20,000	26 (36.1)	35 (17.5)			54 (46.2)	775 (39.6)
$20,000–$49,999	27 (37.5)	58 (29.0)			50 (42.7)	828 (42.3)
$50,000–$999,999	7 (9.7)	66 (33.0)			10 (8.5)	291 (14.9)
$100,000+	10 (13.9)	29 (14.5)			3 (2.6)	65 (3.3)
Prefer not to answerg	2 (2.8)	12 (6.0)			-	-
Health Insurance			Fisher’s exact	<0.001			χ2 (2) = 4.010	0.155
Yes	48 (56.5)	190 (85.6)			79 (67.5)	1449 (73.1)
No	35 (41.2)	26 (11.7)			36 (30.8)	464 (23.4)
I don’t know	2 (2.4)	6 (2.7)			2 (1.7)	69 (3.5)
Recruitment Platform			Fisher’s exact	0.449			Fisher’s exact	<0.001
Geospatial sexual networking apps	21 (23.3)	57 (25.1)			86 (73.5)	1872 (94.5)
Social Media	69 (76.7)	165 (72.7)			27 (23.1)	27 (1.4)
Other website or app	0 (0)	5 (2.2)			4 (3.4)	83 (4.2)

MSM= men who have sex with men

^aDenominators may be smaller for some variables due to missing data

^bColumn percentages are reported

^cChi-Square p-values unless otherwise stated

^dM= Median, IQR= Interquartile Range

^eFor WHSPP, this category includes 35 respondents who selected “I prefer not to answer” for race (18 in Spanish, 17 in English), 64 who selected “Other” and typed in responses including “Hispanic”, “Latino”, “Mexican”, or “None of the above” (24 in Spanish, 40 in English), and 4 who selected “Other” and did not specify any particular race in the text box (all English). For UNITE, this category includes responses in the “other” category related to country of origin

^fFor WHSPP, this includes write-in responses such as pansexual, asexual, fluid, etc

^gThe “prefer not to answer” category was excluded for statistical testing.

Participants who completed the UNITE survey in English were more likely to be recruited via geospatial sexual networking apps (94.5%) compared to those who completed the survey in Spanish (73.5%), and were less likely to be recruited through social media (1.4% vs. 23.1%) (p<.001). Overall, a lower percentage of WHSPP respondents than UNITE respondents were recruited via geospatial sexual networking apps, but there were no statistically significant differences in recruitment platform by language for WHSPP.

In the WHSPP survey, Spanish-language respondents were significantly less likely to report having health insurance than English-language respondents (56.5% vs. 85.6%; p<.001) and were more likely to have an annual household income of less than $20,000 (36.1% vs. 17.5%; p<.001). UNITE survey respondents did not significantly differ by language with respect to annual household income or having health insurance.

HIV risk and prevention behaviors by language of survey completion

Table 2 presents comparisons of HIV risk and prevention behaviors among Latino SMM in the WHSSP and UNITE surveys by language of survey completion. In the WHSPP survey, 8.8% of English-language respondents reported injection drug use while no Spanish-language respondents reported injection drug use (p=0.005). Further, 11.6% of English-language respondents reported meth use compared to 1.4% of Spanish-language respondents (p=0.016) in WHSPP. Similarly, a greater percentage of English-language respondents in the UNITE survey reported meth use (7.7%) compared to Spanish-language respondents (2.6%) (p=0.019). For UNITE, a greater percentage of English-language respondents reported having tested for HIV in the past year (78.6%) relative to Spanish-language respondents (64.1%) (p<.001) and having tested for STIs in the past 3 months (50.9% vs. 39.3%, p=.015). A greater percentage of English-language respondents reported having been diagnosed with gonorrhea in the past 6 months (8.1%) than Spanish-language respondents (2.6%) in UNITE (p=.031) while a greater percentage of Spanish-language respondents reported having been diagnosed with syphilis in the past 12 months (13.6%) compared to English-language respondents (5.0%) (p=.017) in the WHSPP survey. There were no significant differences by language in the number of sex partners, condomless anal sex, or PrEP use in either survey.

Table II:

HIV risk and prevention behaviors among Latino MSM participating in the Washington HIV/STI Prevention Project (WHSPP) (2017 and 2018/2019) and the UNITE Survey (2017/2018) by language of online survey completion^a

	WHSPP (n=317b)				UNITE (n= 2099b)
	Spanish (n= 90b)	English (n= 227b)	Test statistic	P-valued	Spanish (n= 117b)	English (n= 1982b)	Test statistic	P-valued
	N (%c)	N (%c)			N (%c)	N (%c)
Total number of male sex partners in the past 12 months M(IQR)e	3.0 (1.0, 6.0)	3.0 (1.0, 6.3)	W=9618	0.818	-	-
Total number of casual male sex partners in the past 3 months M(IQR)b	-	-			4.0 (2.0, 7.0)	5.0 (3.0, 8.0)	W=103165	0.219
Any condomless anal sex with unknown or HIV positive status partners	23 (30.7)	86 (41.1)	χ2 (1) = 2.14	0.144	-	-
Condomless anal sex with unknown or HIV positive status casual partners
Receptive	-	-			87 (74.4)	1448 (73.1)	χ2 (1) = 0.095	0.830
Insertive	-	-			89 (76.1)	1456 (73.5)	χ2 (1) = 0.387	0.590
Any	-	-			101 (86.3)	1675 (84.5)	χ2 (1) = 0.279	0.592
Injection drug use	0 (0)	18 (8.8)	Fisher’s exact	0.005	1 (0.9)	45 (2.3)	χ2 (1) = 1.033	0.250
Poppers use	17 (23.0)	43 (21.6)	χ2 (1) = 0.006	0.938	51 (43.6)	751 (37.9)	χ2 (1) = 1.520	0.221
Meth use	1 (1.4)	23 (11.6)	χ2 (1) = 5.793	0.016	3 (2.6)	153 (7.7)	χ2 (1) = 4.268	0.019
History of exchange sexf	1 (1.3)	4 (2.0)	Fisher’s exact	>0.999	13 (13.7)	231 (22.2)	χ2 (1) = 3.751	0.053
Tested for HIV			χ2 (2) = 2.893	0.235			χ2 (2) = 19.075	<0.001
Never tested	15 (16.9)	57 (25.3)			7 (6.0)	130 (6.6)
Tested in last year	54 (60.7)	128 (56.9)			75 (64.1)	1558 (78.6)
Tested more than 1 year ago	20 (22.5)	40 (17.8)			35 (29.9)	294 (14.8)
Heard of PrEP	70 (77.8)	170 (74.9)	χ2 (1) = 0.156	0.693	-	-
PrEP use			Fisher’s exact	0.427			χ2 (2) = 0.820	0.664
Never used	65 (76.5)	175 (79.2)			101 (86.3)	1649 (83.2)
Used in the past	6 (7.1)	8 (3.6)			11 (9.4)	220 (11.1)
Current use	14 (16.5)	38 (17.2)			5 (4.3)	113 (5.7)
Tested for STIs	50 (57.5)	125 (56.6)	χ2 (1) = 0.000	0.986	46 (39.3)	1009 (50.9)	χ2 (1) = 5.938	0.015
Diagnosed with any of the following STIsg
Gonorrhea	19 (21.6)	38 (17.1)	χ2 (1) = 0.569	0.451	3 (2.6)	160 (8.1)	χ2 (1) = 4.680	0.031
Chlamydia	8 (9.1)	35 (15.8)	χ2 (1) = 1.825	0.177	7 (6.0)	116 (5.9)	χ2 (1) = 0.003	0.954
Syphilis	12 (13.6)	11 (5.0)	χ2 (1) = 5.709	0.017	8 (6.8)	83 (4.2)	χ2 (1) = 1.871	0.171
Any bacterial STI	27 (30.7)	54 (24.3)	χ2 (1) = 1.011	0.315	16 (13.7)	273 (13.8)	χ2 (1) = 0.001	0.976

MSM= men who have sex with men, STI= sexually transmitted infection, PrEP= pre-exposure prophylaxis

^aUnless otherwise specified, WHSSP variables were assessed for the past 12 months and UNITE variables were assessed for the past 3 months

^bDenominators may be smaller for some variables due to missing data

^cColumn percentages are reported

^dChi-Square p-values unless otherwise stated

^eM = Median, IQR = Interquartile range

^fIn WHSSP, this is defined as receiving things like money or drugs for sex in the past 12 months. In UNITE, this is defined as having you ever exchanged sex for receiving goods (e.g., money, drugs, shelter, food). This variable was added to the survey after initial implementation and as such the sample size is 857

^gSTI diagnoses were measured with a past 6 month recall window in UNITE and a past 12 month recall window for WHSPP.

Multivariable associations with HIV risk and prevention outcomes

Table 3 presents the multivariable associations with having tested for HIV in the past 12 months in the WHSPP and UNITE surveys. In Model A, which adjusted for age, education, health insurance, and recruitment platform, Spanish language survey completion was not associated with HIV testing in the WHSPP, but in the UNITE sample it was associated with lower likelihood of having tested for HIV in the past 12 months (aRR= 0.83, 95%CI: 0.73, 0.95; p=.007). This association remained significant in the UNITE sample in Model B, which also adjusted for HIV-related risk factors (aRR= 0.82, 95%CI: 0.71, 0.94; p=0.005).

Table III.

Multivariable associations with having tested for HIV in the past 12 months among Latino MSM

	WHSPP				UNITE
	Model A (n=299a)		Model B (n=257a)		Model A (n=2099a)		Model B (n=2006a)
	aRR (95% CI)	P-value	aRR (95% CI)	P-value	aRR (95% CI)	P-value	aRR (95% CI)	P-value
Spanish language	1.18 (0.95, 1.46)	0.140	1.20 (0.97, 1.50)	0.100	0.83 (0.73, 0.95)	0.007	0.82 (0.71, 0.94)	0.005
Age	1.00 (0.99, 1.01)	0.937	1.00 (0.99, 1.01)	0.983	1.00 (0.99, 1.00)	0.591	1.00 (0.99, 1.00)	0.708
Education		<0.001		<0.001		<0.001		<0.001
Less than high school	Ref		Ref		Ref		Ref
High school or GED	1.35 (0.69, 2.64)		1.25 (0.63, 2.47)		1.15 (0.95, 1.40)		1.11 (0.91, 1.35)
Some college	1.32 (0.68, 2.55)		1.30 (0.66, 2.54)		1.25 (1.05, 1.50)		1.23 (1.02, 1.47)
Bachelor’s degree or higher	2.18 (1.15, 4.16)		1.94 (1.01, 3.73)		1.38 (1.15, 1.65)		1.35 (1.13, 1.62)
Has health insurance	0.98 (0.73, 1.31)	0.881	0.97 (0.72, 1.30)	0.831	1.14 (1.07, 1.21)	<0.001	1.14 (1.07, 1.21)	<0.001
Recruited from geospatial sexual networking appsb	0.95 (0.76, 1.18)	0.620	1.00 (0.81, 1.23)	0.987	1.19 (1.06, 1.35)	0.004	1.22 (1.07, 1.38)	0.003
Total number of male sex partners in the past 12 months (aRR per 1 partner)			1.00 (1.00, 1.01)	0.141			-
Total number of casual male sex partners in the past 3 months (aRR per 1 partner)			-				1.00 (0.99, 1.00)	0.515
Any condomless anal sex with unknown or HIV positive status partners			1.19 (0.99, 1.43)	0.069			-
Any condomless anal sex with unknown or HIV positive status casual partners			-				0.99 (0.94, 1.05)	0.798
Drug use (injection drugs, meth, or poppers)			1.23 (1.02, 1.47)	0.027			1.07 (1.02, 1.12)	0.003

^aObservations with missing data on one or more variables were dropped from regression analyses

^bFor regression analyses, recruitment through social media was grouped with recruitment through other websites/apps and used as the referent category.

Language of survey completion was not statistically associated with current PrEP use (vs. never) in either multivariable regression model that examined PrEP use in either survey (Table 4). Sensitivity analyses examining the associations with ever PrEP use (current or past) vs. never demonstrated similar results. In the multivariable regression models that assessed for associations with having had any bacterial STI diagnosis in the past 6 months for UNITE (Table 5, language of survey completion was not statistically associated with reporting an STI diagnosis across models. In WHSPP, Spanish-language respondents had a significantly higher likelihood of reporting an STI diagnosis in the past 12 months than English-language respondents in Model B (aRR=1.63, 95% CI 1.03, 2.57; p=0.038).

Table IV.

Multivariable associations with current PrEP use vs. never use of PrEP

	WHSPP				UNITE
	Model A (n=277a)		Model B (n=240a)		Model A (n=1868a)		Model B (n=1782a)
	aRR(95% CI)	P-value	aRR (95% CI)	P-value	aRR (95% CI)	P-value	aRR (95% CI)	P-value
Spanish language	1.49 (0.83, 2.65)	0.178	1.16 (0.61, 2.18)	0.652	0.73 (0.29, 1.88)	0.518	0.73 (0.28, 1.87)	0.506
Age	1.02 (1.00, 1.04)	0.103	1.03 (1.01, 1.05)	0.013	1.00 (0.98, 1.02)	0.811	0.99 (0.98, 1.01)	0.207
Education		0.029		0.185		0.002		0.001
Less than high school	Ref		Ref		Ref		Ref
High school or GED	0.52 (0.05, 5.39)		0.47 (0.05, 4.36)		3.53 (0.47, 26.81)		3.28 (0.43, 24.98)
Some college	2.16 (0.29, 15.83)		1.59 (0.24, 10.64)		3.46 (0.48, 24.66)		3.16 (0.44, 22.70)
Bachelor’s degree or higher	3.33 (0.47, 23.77)		2.04 (0.31, 13.51)		6.72 (0.94, 48.06)		6.35 (0.88, 45.80)
Has health insurance	3.73 (1.04, 13.37)	0.043	2.91 (0.81, 10.46)	0.102	2.33 (1.35, 4.03)	0.003	2.42 (1.41, 4.15)	0.001
Recruited from geospatial sexual networking appsb	1.30 (0.78, 2.16)	0.316	1.21 (0.71, 2.04)	0.479	1.56 (0.62, 3.94)	0.255	1.45 (0.57, 3.71)	0.440
Total number of male sex partners in the past 12 months			1.01 (1.01, 1.02)	0.001			-
Total number of casual male sex partners in the past 3 months			-				1.01 (1.00, 1.02)	0.001
Any condomless anal sex (CAS) with unknown or HIV positive status partners			0.99 (0.58, 1.70)	0.985			-
Any condomless anal sex (CAS) with unknown or HIV positive status casual partners			-				1.38 (0.84, 2.28)	0.205
Drug use (injection drugs, meth, or poppers)			1.12 (0.66, 1.89)	0.683			2.01 (1.42, 2.86)	<0.001

^aObservations with missing data on one or more variables were dropped from regression analyses

^bFor regression analyses, recruitment through social media was grouped with recruitment through other websites/apps and used as the referent category.

Table V.

Multivariable associations with any bacterial STI diagnosis (in the past 12 months for WHSPP and in the past 6 months for UNITE)

	WHSPP				UNITE
	Model A (n=300a)		Model B (n=259a)		Model A (n=2099)		Model B (n=2006a)
	aRR(95% CI)	P-value	aRR (95% CI)	P-value	aRR (95% CI)	P-value	aRR (95% CI)	P-value
Spanish language	1.24 (0.80, 1.94)	0.337	1.63 (1.03, 2.57)	0.038	1.23 (0.76, 1.98)	0.399	1.22 (0.74, 2.00)	0.441
Age	1.00 (0.98, 1.02)	0.786	1.00 (0.98, 1.02)	0.903	0.99 (0.98, 1.00)	0.119	0.99 (0.97, 1.00)	0.019
Education		0.007		0.209		0.950		0.932
Less than high school	Ref		Ref		Ref		Ref
High school or GED	1.30 (0.46, 3.68)		1.59 (0.46, 5.43)		0.94 (0.50, 1.76)		0.88 (0.48, 1.60)
Some college	1.10 (0.40, 3.03)		1.60 (0.48, 5.36)		0.96 (0.54, 1.68)		0.90 (0.53, 1.53)
Bachelor’s degree or higher	2.39 (0.92, 6.22)		2.32 (0.73, 7.31)		1.02 (0.57, 1.83)		0.95 (0.55, 1.65)
Has health insurance	0.79 (0.46, 1.36)	0.395	0.96 (0.54, 1.71)	0.901	1.32 (1.01, 1.72)	0.043	1.39 (1.06, 1.80)	0.016
Recruited from geospatial sexual networking appsb	0.66 (0.40, 1.11)	0.116	0.59 (0.35, 0.99)	0.046	2.64 (1.31, 5.29)	0.006	3.09 (1.39, 6.88)	0.006
Total number of male sex partners in the past 12 months			1.01 (1.00, 1.01)	<0.001			-
Total number of casual male sex partners in the past 3 months			-				1.01 (1.00, 1.02)	0.002
Any condomless anal sex (CAS) with unknown or HIV positive status partners			2.05 (1.35, 3.11)	0.001			-
Any condomless anal sex (CAS) with unknown or HIV positive status casual partners			-				0.92 (0.70, 1.21)	0.543
Drug use (injection drugs, meth, or poppers)			2.23 (1.53, 3.27)	<0.001			2.05 (1.64, 2.55)	<0.001

^aObservations with missing data on one or more variables were dropped from regression analyses

^bFor regression analyses, recruitment through social media was grouped with recruitment through other websites/apps and used as the referent category.

DISCUSSION

Analyses of the WHSPP and UNITE datasets illustrated how including Spanish-speaking Latinos in internet-based research is critical to understanding the HIV risk and prevention needs of cisgender Latino SMM. Using Washington State and national data, our findings demonstrated that cisgender Latino SMM who completed online surveys in Spanish differed from those who completed surveys in English across several sociodemographic characteristics and indicators of HIV risk. We found that Latino SMM who completed online surveys in Spanish were older, had different educational experiences, may experience additional barriers to STI/HIV testing, and were less likely to report specific substance use behaviors than those responding in English. Encouragingly, though, there were no differences by survey language with respect to current PrEP use, although use was low across all Latino SMM, particularly in UNITE.

As participants who completed the WHSPP survey in Spanish were more likely to be uninsured, have lower levels of education, and have lower annual household incomes than English language participants, Latino SMM with lower levels of English proficiency may experience greater barriers to accessing health and HIV prevention services in Washington State. While Spanish-language respondents in UNITE also reported lower annual household incomes than English-language respondents, a greater percentage of Spanish-language respondents in UNITE had a bachelor’s degree or higher than English-language respondents. Despite these differences in educational attainment, UNITE data suggest that Spanish-language respondents may still experience significant socioeconomic challenges to accessing health services.

Our finding that Spanish-language respondents tended to be older than English-language respondents in both WHSPP and UNITE reflects national trends in the Latino population in which the younger population is more likely to be English proficient (29). These data suggest that Spanish-dominant Latino SMM may have unique generational and age-related needs with regards to HIV prevention. Further, as Latino SMM who completed the WHSPP and UNITE surveys in Spanish were less likely to identify as White compared to those who completed the surveys in English, Spanish-dominant Latino SMM may be more likely to experience racism or discrimination faced by non-White minority populations, which can affect access to HIV prevention services. However, these differences in reports of race may also be attributable to different understandings of race and ethnicity in the U.S. compared to countries in Latin America as suggested by a larger proportion of Spanish-language respondents reporting another or no race (30).

In addition to differences across sociodemographic characteristics, our findings indicated that Latino SMM who completed surveys in Spanish differed from Latino SMM who completed surveys in English across select HIV risk and prevention behaviors. Spanish-language respondents were less likely than English-language respondents to have tested for HIV in the past year and were less likely to have tested for STIs in the past 3 months in UNITE, which underscore gaps in testing services for Spanish-language Latino SMM respondents. Further, completing the survey in Spanish was significantly associated with decreased likelihood of having tested for HIV in the past 12 months even after adjusting for sociodemographic characteristics and risk behaviors in the UNITE survey. This finding underscores that national efforts to increase HIV testing in the Latino SMM population must take their literacy and language needs into account. The lack of linguistically appropriate HIV/STI prevention services in the U.S. may exacerbate the language barriers that contribute to inequities in HIV prevention (13). The lack of a significant association between language and HIV testing in the WSHPP sample may reflect local, concerted efforts in Washington State to increase HIV testing through community-based organizations. Specifically, the End AIDS Washington Campaign launched in 2014 was one of the first in the country, and focused on developing partnerships between the Department of Health with local public health and community-based organizations to increase the percent of people diagnosed with HIV (31). The Campaign’s emphasis on community engagement may have facilitated access to Spanish speaking individuals for HIV testing (31).

Our findings that language of survey completion was associated with reporting an STI diagnosis in the past 12 months in the multivariable model and that a greater percentage of Spanish-language respondents reported a syphilis diagnosis in the past year relative to English-language respondents in the WHSPP survey may reveal challenges in STI prevention at a local level. Specifically, Washington State experienced a 279% increase in cases of primary and secondary (P&S) syphilis from 2010 to 2018 (32). As SMM accounted for approximately three-fourths of all P&S syphilis cases, attention to diagnosis rates according to SMM’s language needs may guide allocation of treatment and prevention resources to more effectively reach affected subpopulations.

Study results also pointed to differences between Spanish and English-language respondents in reports of meth use. Across both surveys, a greater percentage of English-language respondents reported meth use (in past 12 months for WHSPP and in past 3 months for UNITE) than Spanish-language respondents. A greater percentage of English-language respondents also reported injection drug use than Spanish-language respondents in the WHSPP. Yet, poppers use was similar across language of survey completion. Our findings reflect important differences in the social and behavioral contexts between the two groups, which may contribute to distinct opportunities for drug use and access to particular drugs.

While study results pointed to some significant differences between Spanish and English-language respondents in HIV risk and prevention, many of the behaviors were similar across the two groups. Bivariate analyses indicated that there were no significant differences between Spanish and English-language respondents in PrEP use. Results from the multivariable models demonstrated that this lack of association between language and PrEP use persists even after adjusting for potential confounding variables. However, the regression analyses suggest that education, health insurance, and sexual risk behaviors are associated with PrEP use. Recruitment platform, sexual risk behaviors, and drug use are associated with reporting an STI diagnosis. Hence, approaches to reach cisgender Latino SMM for online surveys or for HIV/STI prevention interventions must not only address their language needs but also the educational and health system barriers that impact their access to services.

Overall, Latino SMM, regardless of language, may experience challenges to accessing and using PrEP. The majority of Latino SMM in WHSPP and UNITE reported never having used PrEP, which aligns with existing data that demonstrate disparities in PrEP uptake in Washington State and the U.S. among Latino SMM (21, 22, 33). Given the elevated risk of HIV observed among Latino SMM (34), these findings suggest that targeted interventions to increase PrEP uptake in this population at both local and national levels are necessary to reach the goals of Ending the HIV Epidemic (EHE) in the U.S. (35).

There are several study limitations to the current study. The sample in WHSPP and UNITE may not be representative of all cisgender Latino SMM in Washington State or the U.S., respectively. The UNITE survey only recruited high-risk SMM who were eligible for PrEP or taking PrEP with suboptimal adherence. The self-report measures used in both surveys may be subject to bias. Additionally, the small sample sizes of Spanish-language respondents in WHSPP and UNITE limit the generalizability of the findings. The data were collected using online convenience samples and only included participants who had access to the internet. Notably, approximately 86 percent of adult Latinos in the U.S. reported using the internet in 2015 (36), demonstrating that online surveys may reach the vast majority of this population. While the proportion of Latinos who do not have access to the internet tend to be Spanish dominant speakers and foreign-born, the digital divide between Spanish dominant and English dominant speakers has narrowed in recent years with nearly three-fourths of Spanish dominant Latinos reporting internet use (37).

Further, given differences in recruitment strategies and in the inclusion and exclusion criteria of WHSPP and UNITE, findings are not necessarily comparable. Specifically, UNITE focused on recruitment of SMM who reported one or more indicators of HIV risk. Additionally, our finding that English-language respondents were more likely to be recruited for UNITE through sexual networking apps than Spanish-language respondents is important given prior research indicating that SMM who use particular sexual networking apps may differ in their sociodemographic characteristics, HIV risk behaviors, and substance use behaviors from men who do not use those apps (38). Hence, Spanish and English group differences in UNITE may be attributable to the study’s recruitment strategy and the differences in users of particular social media platforms and sexual networking apps. Further, as the recruitment strategies for UNITE and WHSPP did not differ for Spanish and English-language respondents apart from the language of the advertisements, efforts to recruit Spanish-speaking Latino SMM for online surveys may warrant tailored approaches to reach them through applications, websites, and other digital platforms that are unique to this population.

Our analyses did not assess whether participants were born in or outside of the U.S.—nativity is an important factor that is linked to language, stigma, discrimination, and social networks, which can all influence access to HIV prevention resources. National data suggest that approximately 90% of U.S.-born Latinos are English proficient compared to 36% of foreign-born Latinos (39). Hence, although we were unable to account for nativity, Spanish-language respondents may include more foreign-born participants than English-language respondents across both datasets.

CONCLUSIONS

Our study results illustrate that language of online survey completion among cisgender Latino SMM is associated with key sociodemographic characteristics and behaviors that can shape access to HIV prevention services. Latino SMM who are not English proficient comprise a unique subgroup that may have a specific HIV health and risk behavior profile. Hence, Latino SMM who complete online surveys in Spanish may have different HIV and STI prevention needs. There is a clear need to provide tailored and linguistically appropriate HIV prevention services for the Latino SMM community. Further, increased and tailored efforts to recruit and include Spanish-speaking Latino SMM for local and national internet-based studies are necessary to enhance understanding of their barriers and opportunities for HIV prevention.