Racial and ethnic sexually minoritized men (SMM) and gender-expansive communities (sexual and gender minorities; SGM), including such groups as Black and Latine SGM, face a health disparity–driven HIV epidemic in New York City (NYC), as approximately 80% of new diagnoses in NYC among people assigned male at birth are among Black and Latine SMM (Centers for Disease Control and Prevention [CDC], 2021b; HIV Epidemiology Program & New York City Department of Health and Mental Hygiene, 2019). More broadly in the U.S., it is estimated that one in five Latine SMM and one in three Black SMM are diagnosed with HIV by age 49 (Hess et al., 2017; Matthews et al., 2016; Singh et al., 2024), and 62% of Black transgender women are estimated to be living with HIV (CDC, 2021a). A milieu of under-addressed structural and intersecting discrimination, fueled by historical oppression, has made Black and Latine SGM subject to these HIV inequities (Bauer, 2014; Bowleg, 2008, 2012, 2013; Callander et al., 2019; Crenshaw, 2017; Duncan et al., 2019, 2020, 2021; English et al., 2020; Feelemyer et al., 2021; Quinn et al., 2019).

Social network characteristics may have utility in linking Black and Latino SGM to HIV prevention services: a previous study found that Black SGM were more likely to discuss preventing HIV/STIs with network members who were family members and older than them (Shrader, Duncan, Driver, et al., 2024). However, social networks may also worsen the HIV-related disparity among Black SGM: social network norms endorsing HIV-related risk behavioral profiles were associated with individual-level HIV-related behaviors, such as sex-drug use (Shrader et al., 2023). Further, sexual networks may predispose Black and Latine SMM to increased vulnerability to HIV as well: Tieu et al. (2016) found that Black and Latine SMM differed in terms of their sexual networks; although there were no differences in sexual network size, Black SMM preferred sexual partners of the same race relative to White and Latine SMM; however, Black and Latine SMM were more likely than White SMM to report having sexual partners living with HIV or of an unknown HIV serostatus (Tieu et al., 2016). As a 2024 study found that 61% of Black SMM and transgender women reported sex-drug use (Shrader, Duncan, Knox, et al., 2024), and SMM who use stimulants are at a three- to six-fold greater risk of HIV infection (Koblin et al., 2006; Oldenburg et al., 2015; Ostrow et al., 2009; Plankey et al., 2007), further investigation into the social networks of Black and Latine SGM in NYC is warranted to address these health disparities. As individuals do not exist in a vacuum, social network studies facilitate the examination and advancement of HIV equity among Black and Latine SGM (Maulsby et al., 2014; Millett et al., 2007, 2012).

To understand how networks can influence the dissemination of information, behaviors, and attitudes within a network, various social network methodologies have been proposed by Latkin and colleagues (2013), including social network recruitment and interventions approaches. One of these recruitment approaches is respondent-driven sampling (RDS), which has increased in popularity as a sampling technique to recruit individuals from underreached communities. Unlike snowball sampling, which is subject to recruitment biases, RDS provides an opportunity to incorporate a statistical weighting to adjust for these biases and produce more representative samples. RDS offers significant improvements over other network recruitment methods such as peer referral (PR) and multiplicity sampling. Additionally, RDS enhances multiplicity sampling by utilizing dual incentives (for both the referrer and referent) and a coupon recruitment system, thereby improving data integrity and diversity (Heckathorn & Cameron, 2017).

RDS addresses the limitations of traditional sampling techniques in reaching marginalized and minoritized populations (Heckathorn & Cameron, 2017). More specifically in the U.S., RDS has successfully recruited Black or African Americans (Kogan et al., 2011; MacQueen et al., 2015), Latines (Garcini et al., 2022; Kanamori et al., 2019; Montealegre et al., 2013; Shrader et al., 2021), and Asian Americans (Merli et al., 2022). RDS was designed and has since consistently been utilized for HIV surveillance research, particularly recruitment and outreach among populations considered most at risk that lack sampling frames, including people who inject drugs (Heckathorn, 2002; Heckathorn & Cameron, 2017; Latkin et al., 2010, 2013; Rudolph et al., 2011). Additionally, RDS methodology has been widely used as an effective methodology to capture representative data from vulnerable populations that are considered hidden or hard to reach due to stigma and marginalization (Johnston et al., 2013). RDS has also been shown to effectively recruit sexually minoritized (Maulsby et al., 2013; Middleton et al., 2022) and gender-expansive communities (Arayasirikul et al., 2015, 2016). Further, multiple U.S.-based studies have shown success utilizing RDS to recruit socially networked immigrant populations, largely participants from Latin American countries (Garcini et al., 2022; Kanamori et al., 2019, 2021; Shrader et al., 2021).

Thus, RDS also has the potential to identify pre-existing sociocentric networks and is one of few recruitment methods that can be used to recruit community-based networks of groups of friends, or sociocentric friendship networks, for an intervention. Sociocentric networks have utility in identifying potential change agents to disseminate HIV prevention information among SMM communities and to examine structural patterns of network interactions (Chung et al., 2005; Schneider, McFadden, et al., 2012; Schneider et al., 2015). Studies using mixed-method network sampling to recruit community-based samples of SGM have demonstrated the effective use of sociocentric approaches for people living in urban areas (Hequembourg & Panagakis, 2019). For example, Kanamori and colleagues were able to recruit 11 sociocentric groups of 13 Latine SMM in Miami, Florida in 2018 using RDS (Kanamori, Shrader, Johnson, et al., 2022). However, it is uncertain if this strategy would work in the “post-pandemic” world due to the disruption of social networks as posited by the Big Events theory (Friedman et al., 2021), or in a major city like New York, New York (NYC).

As different sampling strategies can influence sample compositions, it is important to select the most appropriate strategy for the population being studied and the research aims (Meyer & Wilson, 2009). As sampling can play the most pivotal role in conducting research, each method must carefully be considered before selecting the most appropriate method for the population under investigation. For example, while direct recruitment can more efficiently identify SGM living with HIV with unsuppressed viral load, RDS may be more efficient in reaching SGM who have low income and are not open to their sexuality. Thus, there may be utility in supplementing RDS with direct recruitment to enlist a more diverse sample experiencing intersectional stigma.

Although there is ample research using RDS to recruit SGM to expand the scientific knowledge base of HIV prevention among SGM (Duncan et al., 2019; Fuqua et al., 2012; Kanamori et al., 2021; Murrill et al., 2016; Schneider, Michaels, et al., 2012; Wei et al., 2012), there is scant information on the feasibility of RDS to enroll groups of Black, Latine, and Caribbean SGM who (a) live in an urban area and (b) are positioned in an existing sociocentric friendship network (Shelton et al., 2019). In addition, there is much heterogeneity within the Black and Latine communities in NYC, with over 20% of NYC’s population attributing their country of birth to the Caribbean or Latin America (New York City Comptroller & Bureau of Budget, 2017). Thus, we aimed to conduct a pilot study to assess the feasibility of RDS, relative to PR, to recruit Black and Latine SGM in NYC, a major U.S. city, and to assess the feasibility of using sociocentric network interventions for HIV prevention implementation research among this sample.

We recruited a total of 20 participants to serve as seeds for the sRDS network; however, of these, only 6 seeds and 45 network members enrolled in the study (n = 51). Of 51 sRDS participants, 29 met inclusion criteria, provided consent, and completed the quantitative assessments. Those participants who were unable to recruit their networks were provided the opportunity to participate in the PR network. For the PR networks, a total of 70 individuals completed the screener and of these, 15 participants were enrolled as seeds and 15 were referred into the study and enrolled as PR network members. One of these participants did not provide complete data and their information was excluded from analyses. These enrollment networks are visualized in Figure 1. While the sociocentric networks were composed of six unique networks, the peer referral networks were composed of 15 unique networks, the majority of which were isolated nodes. Our sample size of 30 for each group suffices to establish feasibility as determined by previous research (Browne, 1995; Teresi et al., 2022).

In this pilot study among a focused sample of Black, Latine, and Caribbean SGM, we found that this group did not believe it would be feasible to use sRDS for recruiting their friends into sociocentric groups and receive an HIV prevention intervention in a group setting. We also found that there were differences between those participants recruited using PR and sRDS. Participants’ perceptions surrounding sRDS varied, with some participants viewing sRDS and sociocentric interventions favorably in theory, but most believing it would not be feasible, acceptable, or appropriate in practice. Our qualitative findings suggest that when designing sampling methods such as sRDS or PR with this community, researchers should prioritize intersectional or invisible considerations, perceived facilitators and barriers to participation, characteristics of the individuals who are being referred, the message being disseminated, and logistical strategies and dissemination.

Our sRDS recruitment was unsuccessful in the sense that only six out of 20 seeds were able to recruit their networks and 58% (29/51) participants from the sRDS enrolled into the study despite being recruited. We originally planned to recruit and enroll 13 participants into each sRDS network; however, only one of six networks met this criterion. The PR networks were more successful, with participants who shared information about the study on their social media pages recruiting more participants. Participants born in the U.S., relative to the Caribbean, were generally more open about their sexual identity and felt comfortable sharing flyers about drug use and serving as a point of contact and resource about this study. A prior study with Latina women found that when sRDS was used to recruit sociocentric groups of Latina seasonal farmworkers vulnerable to HIV for an HIV prevention intervention, HIV prevention behaviors were sustained for at least one year (Kanamori et al., 2019). Unlike the present study, which recruited only through CBOs and street outreach, Kanamori’s study actively partnered with a CBO to conduct all enrollment activities. Additionally, Kanamori et al. (2019) did not enroll groups of friends, but rather a mix of strangers and friends. Another study focusing on Latine SMM successfully used sRDS to recruit and enroll 11 groups of 13 Latine SMM each to identify how sociocentric network characteristics were associated with discussions about HIV prevention information. Kanamori et al. (2021) also partnered with a CBO, and the CBO staff conducted all recruitment activities. sRDS seems to be successful when enrolling individuals who have established networks; however, when networks are fragmented or sparse, sRDS is not a feasible study design or recruitment strategy. Although this suggests that sRDS may not work for group-based interventions, it may still be successful as a sampling method for epidemiological studies.

Alternatively, our findings suggest that the COVID-19 pandemic may have disrupted networks, and thus reduced the success of sRDS recruitment for the present study. Participants did not explicitly discuss the role of Big Events (Friedman et al., 2021); however, upon evaluation of themes such as the ideal referrer and referent and logistical strategies for dissemination, we posit that COVID-19 may partially have been responsible for the fragmentation or reconfiguration of social networks and, instead, fostered an increased reliance on social media for news and social networking, as previously reported in other groups (Kovacs et al., 2021). The COVID-19 pandemic not only normalized virtual interaction over in-person, physical interaction, but it also disrupted the U.S.’s public health infrastructure, including social services. Thus, network-based interventions can consider how to account for this shift in social structures and the health infrastructure. Balancing big events and changes to networks and norms could also induce stress because of long-standing intersectional discrimination (Bauer, 2014; Bowleg, 2008, 2012, 2013; Callander et al., 2019; Crenshaw, 2017; Duncan et al., 2019, 2020, 2021; English et al., 2020; Feelemyer et al., 2021; Quinn et al., 2019).

Additionally, we found that relative to using PR, when using sRDS we were able to recruit a higher proportion of foreign-born participants, participants with ancestry from the Caribbean, and participants living with HIV. When considering inclusion criteria for a study, it is important to carefully consider the type of sampling strategy. These findings are consistent with previous research suggesting that RDS can recruit a different participant demographic (Beyrer et al., 2021; Meyer & Wilson, 2009). Across both sampling groups, most participants reported unemployment, low income, and an arrest history, and shared how the incentive was a way to earn additional income. However, participants in our sRDS group were more likely to be immigrants and living with HIV, suggesting that this group may experience more marginalization and vulnerability than the PR group. NYC has numerous immigrant enclaves, and there is substantial heterogeneity across these communities. As a result, the invisible and intersectional identities and experiences of people from different countries of origin must be considered when planning recruitment.

Participants generally held a belief that their friends had competing interests that hindered their enrollment and participation in sRDS, which included socioeconomic barriers. As our incentive for providing a referral was a $10 gift card, future research can consider providing more money within ethical considerations and cash rather than gift cards. Studies conducted in 2018–2023 with similar participant demographics incentivized participants with amounts ranging from $10 (Duncan et al., 2019) to $20 (Kanamori, Shrader, Flores-Arroyo et al., 2022; Kanamori, Shrader, Johnson, et al., 2022) for referring members of their social networks. However, due to a lack of transparency about participant incentives in research (Brown et al., 2016), it is difficult to ascertain a fair rate for compensation despite global incentive databases having been proposed by Brown et al. (2018). A global incentive database would be helpful for benchmarking compensation amounts and ensuring that participants are fairly compensated, without having undue inducement.

Our study also supports various social network theories, including the structural holes theory (Burt, 1995). For example, participants discussed how the ideal participant for an sRDS recruitment method would be someone who has a large and diverse network, which can increase access to novel information (i.e., low constraint). However, participants also discussed the importance of individuals being embedded in tightly knit groups, which can promote trust and comfort in a study (i.e., high constraint). Our study findings suggest that while low constraint may be important to spread information about studies, high constraint could potentially be more effective in recruiting and enrolling individuals into a study. These findings also support that of a previous study examining selection of peer change agents using a sociocentric approach. Schneider et al. (2015) expanded on the structural holes theory to emphasize that individuals with low constraint may also be innovators in HIV preventive behaviors, emphasizing the importance of these individuals in HIV prevention interventions beyond group-based interventions. Our findings also support the theory of homophily, which posits that individuals are more likely to be attracted to and connect to people who are similar to them (McPherson et al., 2001), as participants in this study referred other participants who were similar to them.

Overall, we found that recruitment of participants from pre-existing sociocentric friendship groups for a group-based HIV prevention intervention was not feasible, acceptable, or appropriate for BLC SGM. Additionally, we found that relative to using PR, when using sRDS we were able to recruit a higher proportion of foreign-born participants, participants with ancestry from the Caribbean, and participants living with HIV, emphasizing the strength of RDS as a sampling method in recruiting hidden populations. Future investigation with this population should carefully consider an appropriate recruitment and sampling strategy based on the intersecting experiences their priority population may experience. Additional research should focus on the efficacy of non-sociocentric/group-based social network interventions such as popular opinion leader interventions (Pagkas-Bather et al., 2020). Our findings are relevant because social networks have become a critical means of addressing persistent health inequities among Black and Latine SGM in the U.S., particularly in the absence of structural change.