Applying Social Network Interventions in Men Who Have Sex With Men: A Systematic Review

Communication 623 Final Paper

Presented in partial fulfillment of the requirements for Health Psychology Seminar at Annenberg School for Communication, University of Pennsylvania

by Qijia Chen May 2014

Instructor: Professor John B. Jemmott, Annenberg School for Communication

Abstract

The current paper provides a comprehensive review of the literature on social network intervention in HIV prevention among MSM. I performed a systematic literature research in multi-lingual databases and located a relatively exhaustive collection of articles on social network interventions in MSM. There is strong evidence that such interventions are efficacious in generating changes in behavioral and psychological constructs that might lead to risk reduction in HIV infection and transmission. This paper identified several urgent issues in the current state of research development, particularly, a lack of theoretical development, an insufficient numbers of methodologically rigorous studies, and an absence of clear conceptual distinctions and implementary recommendations. The review also proposes potential solutions and suggestions for future research.

Introduction Men who have sex with men (MSM) comprise the largest proportion of new HIV infections in the USA (CDC, 2011). Efforts to curtail the incidence of HIV infection require that at-risk individuals adopt effective behavioral changes. Previous research and prevention programs targeting risk behaviors among MSM often have focused on individualistic models of health behavior, but sexual behaviors are inherently determined by constituents of individuals’ social networks. Recently, there has been considerable research interest in studying HIV risk and prevention among MSM from network perspectives because network perspectives offer the promise of a way to move the analytic focus above individual-level variables to consider the patterns through which individuals are connected in larger sociometric networks. Social network intervention has been successfully implemented in various target groups. In a systematic literature review, Wang et al. (2011) synthesized 11 network-based condom promotion intervention studies. All of the studies found substantial improvement in condom use or biological outcomes when comparing intervention and control conditions. Although only two out of the 11 included in Wang et al. (2011)’s review targeted MSM, the strong evidence of the effectiveness of social network interventions suggests that such interventions should have comparable, if not greater, effects among MSM. The adoption of network sciences in the public heath arena is relatively recent and multiple, often confusing, terms have been employed to refer to network constructs. To better explicate the scope and goals of the current analysis, I will first briefly clarify the definitions of some of the terms. A risk-potential network is defined as a pattern of multiple ties between two people that can spread infection if the infectious agent is present (Friedman, 1999). In the current analysis, risk-potential networks are conceptualized as a preexisting state determined by local

epidemiological and demographic features, rather than a vehicle for intervention. An egocentric network considers only the direct linages of a given person (“ego”). Operationally, an egocentric network relies solely on ego's self-reported network (“alters”) (Friedman, 2001). Therefore, by definition, an egocentric network might not include all the existing risk-potential linkages. Sociometric networks (or "social network") consist of a set of people and the entire pattern of linkages among them (Rothenberg et al., 1998). In sociometric networks, multidirectional linkages typically connect members (Wasserheit & Aral, 1996). The studies selected in the present paper assembled either egocentric or sociometric networks through volunteering naming by seeds or peer leaders or observation by ethnographic researchers. In some cases, researchers might name the egocentric networks identified by participants as "sociometric network", even though the networks might not reflect all necessary linkages between members. The current study focuses on social network intervention that includes all type of network “alters” (e.g., kin, friends, colleague’s, community members, or sex partners) rather than sexual relationship only. Although the latter might provide greater insights or better capture the riskpotential network, the emphasis of this paper is on social networks (egocentric and sociometric), particularly how researchers could employ the structural characteristics of such networks in HIV interventions with MSM. Network characteristics can include network size, network density, and perceived ego-alter closeness, and the effectiveness of social network interventions will be subject to the impact of interactions of these characteristics. The current analysis will focus on two forms of impact of social networks: social support and norms. Social network interventions utilize natural social networks to facilitate behavioral changes. In most cases, researchers determine the leaders (“seeds”) of each network and engage them to regularly communicate tailored HIV prevention messages to their personal social

network members. Inherently, because of the existing network connections, the seeds, once trained in HIV prevention promotion, can provide their at-risk network members with not only emotional and material support, but also instrumental support (House, Kahn, McLeod, & Williams, 1985; Hogan, 2002). If behavior change recommendations come from the primary influence leader of a social network, the messages might have considerable impact on the network members’ risk behaviors because of the personal nature and emotional bonds. Social norms define another aspect of social network interventions. There are two types of norms: descriptive and injunctive. Descriptive norms typically refer to the perception of others' behavior, in other words, it concerns what behaviors are generally performed by ones’ friends, peers, or others in their social circles. Injunctive norms, on the other hand, refer to perceptions of which behaviors are typically approved or disapproved by one’s social circles. People rely on both descriptive and injunctive norms to determine what is acceptable and unacceptable social behavior. Past research indicates that networks activate normative beliefs through mutual consultation, resource sharing, advice offering, and tacit or explicit restrictions (Faber et al., 2002). Other research has also suggested that social networks are a source of sexual behavior norms (Latkin et al., 2003; Peterson et al., 2009; Wang et al., 2011). Social networklevel interventions carry a number of potential benefits. They allow prevention to be focused on networks that may be at particularly elevated risk and in need of services. In addition, social networks at very high HIV risk are often hidden, marginalized and difficult to access by "outside" professionals. The natural leaders of these networks can be taught to deliver tailored risk-reduction messages to network members who they know personally and who would otherwise be difficult to reach with prevention services. Members of HIV-vulnerable networks may attach particular credibility to HIV prevention communications that come directly from

trusted leaders within their own existing circle of friends, and leaders of networks may derive personal satisfaction from taking on roles as AIDS prevention advocates. The current paper provides a comprehensive review of the literature on social network intervention in HIV prevention among MSM. I systematically searched the research literature using multi-lingual databases to identify a relatively exhaustive collection of articles on social network interventions in MSM. The nine selected studies were analyzed and compared based on their network structure characteristics, sampling and recruitment methods, theoretical frameworks, and reported outcomes. The current paper will identify effective strategies, review the theoretical and methodological features of past studies, and offer directions and suggestions for future research. Method Literature Search I searched PsycINFO, ProQuest, Communication Abstracts, PubMed, and the Education Resources Information Center (ERIC) for studies with network intervention on HIV transmission targeting men who have sex with men. Search terms included “social network”, “egocentric network”, “peer”, “peer-led”, “HIV/AIDS”, “MSM”, “men who have sex with men”, and “intervention”. For example, on PsycINFO I searched for empirical studies with the following search string: (social network OR egocentric network OR friendship groups OR peer) AND (men who have sex with men OR MSM) AND HIV NOT (women OR female). In addition, I also systematically searched the China Knowledge Resource Integrated Database using the same keywords to identify articles published in Chinese. Inclusion Criteria

Research reports were included if at least one condition in a study met the following eligibility criteria: 1. Empirical Study: Due to the relatively small search results, I did not restrict study types. In the current analysis, I included randomized controlled trials, quasi-experimental studies, pre- and post-intervention studies without control groups, and serial cross-sectional intervention studies. 2. HIV: Reports must be relevant to HIV prevention, risk reduction or treatment. The target of the randomized controlled trial must be behavioral changes (e.g., self-reported condom use) or epidemiological measures (e.g., incidence and prevalence rate of HIV). Therefore, randomized controlled trials with HIV-infected populations, but not about HIV prevention intervention or HIV-related outcome measures were excluded. For example, a randomized controlled trial that examined the effectiveness of mindfulness-based stress reduction in affective symptoms management and quality of life improvement in gay men living with HIV (Carrico, Antoni, Duran, Ironson, Penedo, Fletcher, 2006) would not be selected. 3. Population: Studies needed to have MSM as their target population. 4. Outcome Measures: Studies were required to measure a behavioral or psychological outcome such as self-reported unprotected intercourse, unprotected intercourse with multiple partners, knowledge about HIV and AIDS, attitudes towards condom use, self-efficacy in risky reduction behavioral changes, descriptive and injunctive norms. Studies without appropriate behavioral measures were therefore excluded. For example, Young, Zhao, Teiu, Kwok, Gill, & Gill (2013)’s paper on social-media based HIV prevention intervention using peer leaders was not included because the authors did not report outcome measures. A total of 25 citations were identified using the search algorithm across all databases.

After evaluating the articles, I only included titles mentioning social/peer networks and support with relation to HIV intervention. Articles that focused solely on impacts of sexual partner networks were excluded. A total of 16 articles were excluded. Results Study Location and Population. Two studies were conducted in Eastern Europe (Russia and Bulgaria) (Amirkhanian et al, 2003; 2005). Five studies were conducted in China (Lau et al., 2013; Liu et al., 2009; Zhang et al., 2009; Zhang et al., 2010; Zhu et al., 2008). The other two selected studies were conducted in Los Angeles, United States (Choi et al., 2013; Young et al., 2013). With the exception of two studies (Amirkhanian et al, 2003; Choi et al., 2013), each study had between 100 and 400 individuals. Study population consisted exclusively of MSM in urban settings. Network Identification. Although in two studies (Amirkhanian et al, 2003; 2005) the authors claimed that the networks were sociometrically identified, all studies included in the current analysis were in fact egocentric networks. In Amirkhanian et al. (2003; 2005), the seeds (i.e., indexes) were selected based on ethnographic observations in MSM nightclubs, pubs and street venues. However, the networks were consisted of the seeds and their egocentric MSM network members. Similarly, the seeds or leaders in other studies were also selected based on their personality characteristics (e.g., friendliness) and social status (e.g., well-respected in their communities). One study (Zhang et al., 2009) did not specify the selection process of the seeds. The mean network size (i.e., number of total network members/number of seeds) ranged from 3 to 29, with a median of 7 (1 seed corresponds to 7 network members). The detailed information was listed in Table 2. Study Design. Among the nine interventions, two were designed without a control group

(Choi et al., 2013; Liu et al., 2009);), and seven had some form of comparison group (Amirkhanian et al., 2003; 2005; Lau et al., 2013; Young et al., 2013; Zhang et al., 2010; Zhu et al., 2008). The comparison group could be the control groups in randomized controlled trials, or the pre-intervention groups in quasi-experimental studies. The detailed study designs were listed in Table 1. Examples of comparison groups included the same group prior to the intervention (Amirkhanian et al., 2003; Zhang et al., 2010; Zhu et al., 2008), general health education (Young et al., 2013), default HIV risk reduction program (Zhang et al., 2009), and no intervention (Amirkhanian et al., 2005). Among the seven studies with comparison groups, three of these studies employed randomization (Amirkhanian et al., 2005; Lau et al., 2013; Young et al., 2013; Zhang et al., 2009). One study reported taking measures to prevent contamination between intervention and control groups (Zhang et al., 2009). The authors used difference in geographical locations to separate the intervention group and comparison group. Intervention. In six studies (Amirkhanian et al., 2003; 2005; Lau et al., 2013; Young et al., 2013; Zhang et al., 2009; Zhang et al., 2010), the seeds received multiple sessions of training. For example, the leaders received six and nine sessions of group intervention program in Amirkhanian et al. (2003) and Amirkhanian et al. (2005), respectively. Lau et al. (2003)’s training emphasized 10 different themes surrounding HIV risk reduction knowledge and behavioral skills. Young et al. (2013) employed a social media skill training program tailored to the unique characteristics of their intervention design. Lau et al. (2013) and Young et al. (2013) were superior in ensuring the methodological rigor and validity because the control groups in these two studies received general health knowledge information sessions similar to the intervention training in length and formats. In contrast, Zhang et al. (2009) reported that the control city, Yibin, received the default HIV prevention and intervention program that was in

place, without providing further information on the specific features of the existing intervention and how it was different from the social network intervention program. Theoretical Framework. Many studies did not clearly state the theoretical framework that guided the intervention design and choices of outcome measures (Liu et al., 2009; Amirkhanian et al., 2003; 2005; Young et al., 2013; Zhang et al., 2009; Zhang et al., 2010; Zhu et al., 2008). However, the selected intervention techniques, as well as outcome measures, often reflect the influence of social cognitive theory (Bandura, 1986) and Theory of Reasoned Action (Fishbein and Ajzen, 1975). For example, Liu et al. (2009) cited social norms as a possible mechanism for the effectiveness of social network interventions in HIV risk reduction behaviors. Similarly, Amirkhanian et al. (2003; 2005) proposed social learning and establishment of protective social norms as the intervening processes in their network intervention techniques. Lau et al. (2013) specified Theory of Planned Behavior (Ajzen, 1985) and The Health Belief Model (HBM) (Rosenstock, Stretcher, & Becker, 1994) as the theoretical foundations for their intervention design. The training of popular opinion leaders (POLs) in Zhang et al. (2009) is consistent with the theoretical basis in other variations of peer interventions employed in the past, namely, the diffusion of innovation model (Kegeles et al., 1996; Kelly et al., 1997; St Lawrence et al., 1994). For a full list of conceptual models and theories used in the nine selected studies in this review, please refer to Table 3. Outcome Measures. No studies had biological primary outcomes. Five studies (Amirkhanian et al., 2003; 2005; Zhang et al., 2009; Zhang et al., 2010; Zhu et al., 2008) reported condom use or/and unprotected intercourse (UI)/unprotected anal intercourse (UAI) before and after intervention between arms (RCTs and quasi-experimental studies), or only in intervention arm (pre- and post-intervention studies (Amirkhanian et al., 2003; Zhang et al.,

2010; Zhu et al., 2008). Three studies (Amirkhanian et al., 2005; Lau et al., 2013; Zhang et al., 2009) reported significant changes in psychological constructs (i.e., HIV/AIDS knowledge, behavioral intention, self-efficacy in condom use and in voluntary HIV testing and counseling) that predicted risk reduction behavioral changes in the intervention groups, compared to the comparison group. There are four studies that reported other behavioral changes, including HIV testing (Zhang et al., 2010; Zhu et al., 2008) and counseling uptake (Zhang et al., 2009), the rate of using free condom providing service (Zhang et al., 2009), reduction in number of casual sex partners and multiple partners (Zhu et al., 2008), and number of participants who requested an HIV testing kit (Young et al., 2013), in the intervention group, compared to the comparison groups. Although the majority of selected studies proposed that changes in social norms and perceived social support would be responsible for the changes in psychological and behavioral outcome measures, none reported significant differences in constructs related to these proposed mediators. Primary Outcome Analysis. Of the seven studies with comparison groups, six studies reported substantial improvement in behavioral outcomes (Amirkhanian et al., 2003; 2005; Zhang et al., 2009; Zhang et al., 2010; Zhu et al., 2008; Young et al., 2013). The exception was Lau et al. (2013), which showed significant improvement in prevalence of behavioral intention to use condoms consistently for anal sex with men in the next six months and self-efficacy on taking up HIV voluntary testing and counseling (VCT) in the intervention group, compared with a control group. However, the authors only included psychological variables, as delineated in the theory of planned behavior (TPB) and the Health Belief Model (HBM) as their outcome measures. Despite the lack of behavioral outcomes, the changes in theoretical constructs and the strong intervention designs are meritorious. Only two studies assessed condom use at multiple

follow-up periods (Amirkhanian et al., 2003; 2005). Amirkhanian et al. (2003) found a substantial difference between intervention and control at 3 months that disappeared at the 12month follow-up visit among Russian participants, but remained significant among those in Bulgaria. Interestingly, in their 2005 study, the authors reported that the effects of the intervention in reducing unprotected intercourses (UIs) and unprotected anal intercourses (UAIs) became attenuated at 12-month follow-up, but remained among participants who had multiple recent sexual partners (Amirkhanian et al., 2005). Only three studies (Amirkhanian et al., 2003; Zhang et al., 2010; Zhu et al., 2008) clearly stated that analyses were stratified based on casual partners and main partners. For example, Amirkhanian et al., (2003) reported that the intervention had significant effect with condom use behaviors only with casual sex partners but not main sex partners. Discussion The current study provides a systematical review of social network interventions in MSM communities. Apart from the two cross-sectional surveys, which offered insights into the relationship between network characteristics and HIV risk behaviors, the other seven studies all found substantial improvement in behavioral and psychological variables that are protective against HIV transmission among MSM. With a relatively exhaustive literature search and detailed analysis of study characteristics, this paper presents the most up-to-date and wellorganized synthesis of past research. In addition, it offers potential directions for future developments. The selected studies demonstrated that there is ample empirical evidence to suggest that social network-based interventions are efficacious in HIV prevention. Researchers observed significant amounts of improvement in safe sex practices (Amirkhanian et al., 2003; 2005;

Young et al., 2013; Zhang et al., 2009; Zhang et al., 2010; Zhu et al., 2008) and increases in individual-level HIV preventive measures such as requesting HIV testing and counseling services (Young et al., 2013; Zhang et al., 2009; Zhang et al., 2010; Zhu et al., 2008). However, there are substantial gaps in the literature that need to be addressed in future studies. The first problem emerged in the review is the lack of theoretical development in the field. Social network interventions share some similar features with other peer intervention programs (e.g., peer-referral, use of popular opinion leaders). Network-based interventions are unique in that they identify the specific network members that make up individuals’ social circles. The egocentric networks in the selected studies were the main source of social interactions of MSM in their communities and have substantial impact on daily interactions of their members. However, the overlap with other forms of peer-interventions did make it harder for researchers to clarify the theoretical frameworks for their designs. Two papers mentioned the diffusion of innovation model as their main theoretical guideline (Zhang et al., 2009; Zhu et al., 2008), but employed outcome variables in the social cognitive theory (Bandura, 1986) and Theory of Reasoned Action (Fishbein and Ajzen, 1975). The authors might have intended to propose diffusion of innovation model as the mediating mechanisms for social network interventions, but chose these outcome measures to fit in to the popular theories of behavioral changes. To clarify the mechanisms of social network interventions, future research could include measures proposed mediators in diffusion of innovation theories such as individualism, uncertainty avoidance. Other studies, on the other hand, did not specify their guiding theories (e.g., Amirkhanian et al., 2003; 2005; Choi et al., 2013; Young et al., 2013; Zhang et al., 2010). These studies cited social influence or social support as possible intervening mechanisms, but provided little theoretical backgrounds or rationales for such arguments. It is fair to say that the

field currently lacks proper theory application, integration and testing. The problem is not the lack of theoretical developments per se, but the lack of movements towards building an integrative and encompassing theoretical model. Among studies selected for the current review, most authors tacitly emphasized the role of social influence and network similarities. Indeed, people with similar social status, interests and activities tend to have tighter connections with each other, therefore, changing social norms within such networks tend to generate behavioral changes above individual level and enable researchers to reach otherwise inaccessible populations. This line of reasoning has received support from empirical studies (e.g., Kelly et al., 2010; Hao et al., 2013; Tobin et al., 2008), but no substantial efforts have been devoted to consolidating empirical findings into one or several coherent theoretical models. Secondly, in order to facilitate theory developments and accumulate high-quality data, researchers need to improve the methodological rigor of intervention studies by adopting more sophisticated designs. Among the nine selected studies, only three were randomized controlled trials (Amirkhanian et al., 2005; Lau et al., 2013: Young et al., 2013). Zhang et al. (2009) reported using a non-randomized trial with two cities with different geographical locations. However, the two cities were not comparable in some critical dimensions. For example, the authors reported that the control city, Yibin, did not have MSM venues, whereas the intervention city, Mianyang, had multiple nightclubs, bars and other MSM venues. Three other studies employed quasi-experimental design with pre- and post- measures (Amirkhanian et al., 2003; Zhang et al., 2010; Zhu et al., 2008). The conclusions one can draw from such designs are relatively limited, since they do not provide sufficient grounds for causal inference. As researchers have accumulated evidence in support of the effectiveness of social network interventions (e.g., the exploratory surveys and quasi-experimental trials), future research should

aim at producing more RCTs to investigate individual moderators and mediating mechanisms. Thirdly, future research should further address the conceptual and implementary facets of social network interventions. The current synthesis points out to the lack of theoretical and conceptual clarity in the current state of research in the field. Many studies did not report network size, or measures of network connectivity and density (e.g., Young et al., 2013; Zhang et al., 2009; Zhang et al., 2010; Zhu et al., 2008). Researchers need to develop more rigorous methods to establish and measure social ties among individuals within each network to accurately access social influence or social support. Furthermore, it would also be helpful to identify the different sources and types of social support provided by network members. In the egocentric networks investigated in the selected studies, network members might come from the seeds’ friendship group, family group, or sexual networks. Although the types and strength of ego-alter relationships were only explored in the cross-sectional surveys (Choi et al., 2013; Liu et al., 2009), future RCTs should include measures of ego-alter closeness and types of ego-alter relationships. Consideration and measurements of these critical network features will help researchers to gain greater insights into the relationship between network characteristics and patterns of HIV transmission. Lastly, researchers might want to consider expanding social network interventions to MSM communities with greater ethnic and cultural variety. For example, more studies should explore the feasibility of implementing social network intervention in African American MSM communities, since this group has the highest rates of new HIV infections in the US (CDC, 2011). Unlike models of individual behavioral changes such as social cognitive theory or theory of reasoned action, network interventions are ingrained in the local social environment where the intervention takes place. The current literature search only located one RCT done outside China

and Eastern Europe (i.e., Russia and Bulgaria) (Young et al., 2013). Although potential adjustments to local social environments might be necessary, the inspiring results gathered in this synthesis indicate that social network interventions hold huge potential for other communities as well, especially in regions where the most vulnerable members of society might not be reachable by conventional recruitment or intervention programs. Conclusion The review summarized past research in application of social network interventions in MSM communities. There is strong evidence that such interventions are efficacious in generating changes in behavioral and psychological constructs that might lead to potential risk reduction in HIV infection and transmission. The review also points to existing gaps in the literature, particularly, the lack of theoretical development, the insufficient numbers of methodologically rigorous studies, and the absence of clear conceptual distinctions and implementary recommendations. These issues pose momentous challenges for advancement of the field. Future research should address these issues and continue to generate more exciting findings and inform further application of network sciences in HIV prevention research.

Table 1. Study Characteristics
Sample Size n (seeds) 12 14 N (participants , total) 350 82

Authors Liu et al. (2009) Amirkhanian et al. (2003)

Location Shenzhen, China St. Petersburg, Russia and Sofia, Bulgaria St. Petersburg, Russia and Sofia, Bulgaria Hong Kong, China

Design Cross-sectional survey Pretest to 4-month posttest program evaluation Randomized controlled trial

Intervention 350 82

Comparison No No

Amirkhanian et al. (2005)

133

143

52

276

Lau et al. (2013)

Pilot randomized controlled trial

36

33

First wave: 8; Second wave: 6 40

First wave: 58; Second wave: 25 400

Zhang et al. (2009)

Sichuan, China

Non-randomized trial with the assessment points at baseline and 6month post-intervention follow-up Pretest to 3-month posttest program evaluation

200

200

Zhang et al. (2010)

Anhui, China

218

No

12

Zhu et al. (2008)

Anhui, China

Pretest to 3-month posttest program evaluation Cross-sectional survey

223

No

16

223

Choi et al. (2013)

Los Angeles, USA

1196

No

453

1196

Young et al. (2013)

Los Angeles, USA

Cluster randomized trial

57

55

16

112

Table 2. Network Recruitment and Characteristics
Participant Age Mean Network Size 29

Authors Liu et al. (2009)

Sampling and Recruitment Methods

Respondent-driven sampling (RDS) was used to recruit Age range: 18– MSM. A diverse group of 12 seeds, heterogenous in age, 45. MSM congregation venue, and engagement in commercial sex were selected and each was given 3 uniquely coupons to refer up to 3 peers.

Amirkhanian et al. Researchers first obtained ethnographic observations in Mean age: 24.4 (2003) nightclubs, pubs, and street venues in both cities. Two or years. three experienced ethnographers identified social circles and the "seeds" ("indexes"), who are community members observed to be popular with others. The seeds and their egocentric networks included their MSM friends. Each egocentric network included the index and his MSM friends. Amirkhanian et al. Two or three experienced ethnographers identified social Mean age: 22.5 (2005) circles and the "seeds" ("indexes"), who are community years. members observed to be popular with others. The seeds and their egocentric networks included their MSM friends. Each egocentric network included the index and his MSM friends. All participants received brief ADIS one-on-one ADIS riskreduction counseling before networks were randomized to either intervention or control. Lau et al. (2013) Eight non-randomly selected MSM were recruited by the Age range: 18– researchers via multiple sources to serve as the seed peer 25. educators (SPE) based on potential of having good communication skills and large social networks. Prospective seeds were assessed for confidence that they could recruit at least 5 network members. Seeds were randomized into either an intervention (training) or control (printed material distribution) condition. The seeds then recruited members of their networks into the study. No reported network structural evaluations.

6

5

6

Zhang et al. (2009) Mianyang was assigned as the intervention city and Yibin as Age range: 18– the control. Education materials and HIV counseling and 35+. testing were available in both cities. 40 popular opinion leaders (POLs) received four 90-minute training sessions over the course of 8 weeks.

10

Zhang et al. (2010) Recruitment of each network was initiated with a single MSM seed and continued as a peer referral chain. The seed was a community member perceived to be influential and respected by peers and recruited in a previous study. The networks included 3 to 5 waves of references. Zhu et al. (2008)

Mean age: 27 years.

18

The researchers used respondent driven sampling (RDS) for Mean age: 20.7 initial recruitment. Sixteen seeds were selected abased on years. their influence in the MSM community and received four training sessions on HIV prevention knowledge and skills over the course of one month. A chain-referral sample was recruited in Los Angeles Age range: 18– County, California. First recruited "seeds" through referrals 83. at MSM venues such as bars, dance clubs, and coffee shops. Each seed participant received $50 for compensation and was given 3 "recruitment coupons" to pass out to their MSM friends and acquaintances. Mean age: 31.5 years.

14

Choi et al. (2013)

3

Young et al. (2013) Participants were recruited from online venues, community venues frequented by African American and Latino MSM, and direct referrals from participants. Peer leaders were selected based on their personality qualities (e.g., friendly and well-respected), their willingness to use Facebook for HIV knowledge promotion, and their interest in educating others about health. Peer leaders received training on epidemiology of HIV or general health depending on their assigned conditions.

7

*Mean network size, when not originally reported, were calculated by the author (total N/seeds n), based on Amirkhanian (2014)’s method.

Table 3. Methods
Authors Liu et al. (2009) Theoretical Framework Social Cognitive Theory; Norming; subjective and descriptive norms. Social network as a source of sexual behavior norms Social influence; Social Cognitive Theory; Social learning, Social network as a source of sexual behavior norms Social Network Intervention No Comparison Group No

Amirkhanian et al. (2003)

Each network's sociometric leader attended 6session (5 main and 1 booster session) group intervention program. The objective was to train the leaders to effectively deliver risk-reduction advice and consultation to other network members to encourage them to reduce their sexual risk behavior levels by improving their AIDS knowledge, creating positive safer sex norms, attitudes, intentions, and self-efficacy through messages conveyed in regular daily communication. All intervention condition network sociometric leaders attended groups with one another and received a 9-session (5 main and 4 booster session) group intervention program. The objective was to train the leaders to effectively deliver risk-reduction advice and consultation to other network members to encourage them to reduce their sexual risk behavior levels by improving their AIDS knowledge, creating positive safer sex norms, attitudes, intentions, and self-efficacy through messages conveyed in regular daily communication.

No

Amirkhanian et al. (2005)

Social influence; Social Cognitive Theory; Social learning, Social network as a source of sexual behavior norms

Not specified.

Lau et al. (2013)

Theory of Planned Behavior; The Health Belief Model

All intervention condition seeds were trained in a workshop to become peer educators. The objective was to focus on 10 pre-identified themes such as AIDS knowledge, awareness about and selfefficacy in condom use, overcoming barriers to testing, skills of condom use with regular partners, and HIV susceptibility. Peer educators were asked to deliver risk-reduction advice and consultation to other network members to encourage them to reduce their sexual risk behavior levels using theme-based messages in regular daily communication. The peer education workshop was then offered to the first-wave network members so they could become educators to their own (second wave) MSM network members.

Members of the Control Group received some printed information about VCT and HIV/STD prevention. The research team developed the material. The logistics (recruitment, follow-up, and evaluation) of the Control Group were the same as those of the Intervention Group. Two waves of recruitment were also made in the Control Group. Training was not provided to members of the Control Group.

Zhang et al. (2009)

Social influence; Diffusion of Innovation Model

40 participants in Mianyang received training in skills and AIDS knowledge to become popular opinion leaders (POLs). The trained POLs actively distributed education materials, advocated for safe sex, promoted condom use, and encouraged HIV testing in their peer networks. Both seeds and their referral chain network members together attended a 4-session intervention group training. The objective was to train network members to encourage them to reduce sexual behavior risk levels by increasing their AIDS knowledge, evaluating behavioral risks, developing individual risk reduction plans, overcoming obstacles and barriers to practicing safe sex, and improving communication skills about sexual topics.

In the control city, Yibin, local CDC carried out regular HIV prevention and intervention procedure that were in place.

Zhang et al. (2010)

Social influence; Social Cognitive Theory

No

Zhu et al. (2008)

Diffusion of Innovation Model; Social influence; Social Cognitive Theory Risk-potential networks; Social support

No

No

Choi et al. (2013)

No

No

Young et al. (2013)

Social support; Social influence

Peer leaders assigned to the intervention group attended three 3-hour sessions at UCLA, and received lessons on the epidemiology of HIV and ways of using Facebook to discuss stigmatizing topics. Facebook was used to create closed groups (unable to be accessed or searched for by persons who were not group members).

Peer leaders assigned to the control group attended three 3-hour sessions at UCLA, and received lessons on general health and ways of using Facebook to discuss general health subjects. Facebook was used to crease closed groups (unable to be accessed or searched for by persons who were not group members).

Table 4. Outcomes
Authors Liu et al. (2009) Amirkhanian et al. (2003) Amirkhanian et al. (2005) Actual Behavioral Change No Outcome More egos' sexual-partner alters than their non-sexual partner alters encouraged them to use condoms (72% vs. 46%, p