|Home | About | Journals | Submit | Contact Us | Français|
Effective sexually transmitted infection (STIs)/HIV prevention programs are urgently needed, but translating evidence-based methods of STI/HIV prevention into sustainable programs has been difficult. Social influences are critical for establishing condom use norms. This study systematically reviewed social network-based interventions focused on condom promotion. Social networks were defined as groups who self-identified prior to the research study. Eleven eligible research studies were identified and included in this review. Only three studies measured biological endpoints and five studies used validated measures of condom use. Among the nine studies with control groups, eight showed significant improvements in at least one measure of condom use. There were large differences in how social network members were identified and involved in the interventions. This systematic review highlights the potential utility of social network-based condom promotion programs. More research is needed to show how these promising studies can be expanded.
Unsafe sex is a major contributor to mortality and morbidity in low-income settings . Each year, there are more than one billion newly acquired sexually transmitted infections . Despite the known effectiveness of condoms in preventing sexually transmitted infections (STIs) , effective condom promotion programs have been challenging to scale up and sustain .
Individual-based condom promotion interventions have many drawbacks. While individual behaviors do influence HIV risk, it is also important to consider the broader interpersonal and societal factors that contribute to HIV risk . For example, heterosexual condom promotion interventions that focus on one sex ignore the broader dimensions of male–female sexuality and the power dynamics that influence sexual risk and HIV/STI acquisition . Social situational factors tend to have stronger influences on behavior compared to individual level attributes . Furthermore, given the tendency of individual intervention effects to wane over time, exploring and accounting for social influences on condom use may be critical for sustaining safe sex behaviors .
Multifaceted condom promotion interventions that incorporate individual, small group, community, and structural level methods have shown promise . Social and structural influences are increasingly recognized as important in establishing an environment conducive to safe sex. Conceptual models of behaviour change have been developed that explicitly account for these social influences [9, 10].
Several studies have analyzed social networks and their effects on condom use. For the purposes of this paper, we define social network as a group of individuals who knew each other prior to the condom intervention. This social network definition is separate from peer education where peers may not know or interact with their peer group. Our more specific definition of social network was used because some studies are either unclear about recruitment methods or the networks are defined post-hoc by the researchers. Social networks influence group norms linked to health behaviors such as needle sharing and condom use, facilitate health-seeking behaviors, enhance our ability to reach vulnerable groups, and may promote condom use [11–14]. This study systematically reviews social network-based condom promotion interventions.
Pubmed, Embase, Psycinfo, and Cochrane Systematic Reviews were systematically searched for articles fulfilling search criteria that were indexed on or before July 13th, 2010. The search was updated on April 1, 2011 to find articles meeting the search criteria that were published after the initial search date. Search terms included “condom” and “social support or social network or peer group or peer network” (see supplemental data). No language or date restrictions were placed on the search.
A total of 1,244 citations were identified using our search algorithm. After removing duplicates, a total of 815 citations remained. Two researchers (KW and SY) independently evaluated article titles and only included titles mentioning social/peer networks and support with relation to condom usage. Articles that focused solely on manager influences or sexual partner networks were excluded. Studies that assessed condom use as a secondary measure were included. When there was doubt about an individual citation title, KW and SY examined the citation’s abstract. Discordant articles were resolved by a third researcher (JT) through abstract review. All articles had English abstracts. A total of 568 were excluded.
Of the 247 articles remaining, abstracts were read and excluded if they were not interventions. For exclusion categories, see Fig. 1. The methods section of 82 studies were examined and articles in which participants did not know and identify network members prior to the start of the intervention or did not explicitly describe the process of recruitment were excluded. Articles in which a peer educator identified network members post-hoc after the intervention began were excluded. Nine articles were eligible for qualitative synthesis, and one article was excluded as its dataset was part of another study that was already included in the synthesis. In this case, the more recently published study was included. One eligible study was identified in the second update search. Another two eligible studies, one which was not identified in either search and one which was published after updated the search date, were identified by experts. Qualitative synthesis was performed on eleven studies in total.
For each of the eleven studies, data on study date, location, recruitment, network member involvement, retention rate, condom use measures, and other outcomes indicative of condom use were extracted. Summary tables of key variables were created using Microsoft Excel version 2008 and the median follow-up time was calculated.
As seen in Fig. 1, 247 abstracts were examined and 82 research studies fulfilled basic inclusion criteria of having a social network component referring to condoms. Most of the 165 records excluded were descriptive studies and news articles. Among the 82 research studies, nine were comprised of self-identified social network interventions targeting condom promotion. Two studies used the same dataset and the most recently published study was chosen for inclusion. One study was identified in the update search and two studies were identified by experts. Eleven studies were identified in total for review.
Three studies were conducted in the 1990s [15–17] and eight were conducted in the 2000s [11, 18–23]. With the exception of two studies [11, 23], each study had between 250 and 1,000 individuals. Five studies were conducted in high income nations [15–17, 22, 23], five studies in middle income nations [11, 18, 19, 21, 24], and one study in a low-income nation  (high, middle and low income nations were defined according to the World Bank designations ). Nine studies were conducted in the United States or Europe [11, 15, 16, 19, 21–24] and no studies were conducted in Australia or South America. Five studies focused on drug users [15, 17–19, 22], one on female sex workers , two on men who have sex with men [11, 21], and one on African American youth . Only one study focused on commercial sexual interactions , while the rest focused on sexual interactions in casual and regular relationships.
There was a range in both the composition and the identification of social network members. Here index refers to individuals recruited initially by researchers and network member refers to those identified and recruited by indexes. In eight studies, network members did not participate in the intervention directly, but were recipients of diffusion of information from indexes and participated in baseline and/or follow up assessments [11, 15, 18, 19, 21–24]. A ninth study used similar methods, but network members did not complete baseline and follow-up surveys .
Two studies used indexes to recruit network members who then had equal participation in the study [16, 20]. In these two studies, individuals in the target population formed self-identified peer groups in which everyone in the group received the intervention or control.
Among the eleven interventions, two were designed without a control group [11, 20] and nine had some form of control group (Table 1). Examples of comparison groups included the same group prior to the intervention, standard of care informational or life skills curricula [18, 22, 23], and no intervention . Among the nine studies with control groups, six of these studies had randomization [16, 17, 21–24]. None of these studies reported any measures taken to prevent contamination between intervention and control groups.
Condom promotion interventions were informed by a number of conceptual frameworks (Table 2). Most studies (7/11) described multiple theory-based constructs and three studies did not fully specify theory used [15, 20, 22]. Conceptual models and theories used include persuasion and behavior change theories, behavioural science theory, social identity theory, protection motivation theory, and social cognitive theory.
Only three studies had primary biological outcomes [18, 23, 24] (Table 3). These studies evaluated social network interventions among drug users in Thailand , Roma men in Bulgaria , and high-risk women in Baltimore, MD, USA  and tested participants for several STIs including gonorrhea, chlamydia, syphilis, HSV-2, and HIV. In all reviewed studies, primary outcomes included condom use measures, but varied from ‘last episode of intercourse’ to ‘past 90 days’ or ‘past 5 partners.’ Other outcomes included drug use behaviors, other sexual behaviors, and psychosocial factors including attitudes and knowledge about sex and HIV transmission.
The final follow up visit ranged from 3 to 23 months following the intervention, with a median of 12 months. The retention at the last follow-up visit was generally quite high, ranging from 34 to 97.9% with a mean of 82.2%.
Of the nine studies with controls, eight showed substantial improvement in at least one condom measure comparing control and intervention conditions. The exception was Sherman et al. which showed equal increases in condom use for both control and intervention populations . This was one of the studies that measured biological outcomes and while the authors found a significant difference between incident gonorrhea in the total (index + network) experimental condition compared to control subjects (control subjects participated in the life-skills curriculum on methamphetamine use), they did not find significant differences in the incidence of other STIs. In the two studies without controls [11, 20], there was a significant increase in at least one condom outcome measure at post-intervention compared to baseline pre-intervention.
The reviewed studies classified index and network participants (both within the intervention arm) in several different ways. Two studies [18, 24] performed separate analyses of index members and total (index + network) members, reporting no significant difference between control and intervention populations in either case. Some studies only analyzed peer leaders  or indexes [17, 23], and did not report information on network members. Two studies performed separate analyses of peer educators (indexes) and network members compared to control [19, 22]. Both of these studies found that there were substantial reductions in sex without a condom for indexes versus control participants, but only one of these studies found that significant increases in condom use extended to network participants as well . The other five studies analyzed total (index + network) populations only.
Analysis related to condom use measures rarely separated the index and network members (2/11 studies). Comparing condom use outcomes was difficult because of differences in measures, follow up time, control conditions, and stratification based on casual/main partners.
Six studies were able to assess condom use at multiple follow-up periods [16, 18, 21–24]. Kelly et al. found fewer cases of reported unprotected intercourse with all sexual partners at 3 and 12 month follow-up. Amirkhanian et al. found a substantial difference between intervention and control at 3 months that disappeared at the 12 month follow up visit among Russian participants, but remained significant among those in Bulgaria . Another study found substantial differences between condom use at 6 and 18 months, but not at the 12 month follow up point . Tobin et al. similarly found significant increases in 100% condom use and decreases in sex risk overall, but that those findings were not statistically significant at the 12 and 18 month follow-ups among indexes .
Studies differed in their classification of types of sexual partners. Five studies [11, 17, 20, 23, 24] stratified results based on casual partners and main partners. Amirkhanian et al. 2003 found substantial differences between experimental and control condom use behaviors with casual sex partners among men who have sex with men, but no effect with main sex partners . Three other studies also found that changes in condom use did not extend to encounters with main sexual partners [17, 20, 23]. However, Kelly et al. found decreases in unprotected intercourse regardless of the type of partner .
This study systematically reviewed network-based condom promotion interventions. All of the eleven selected network intervention studies found substantial improvement in condom use or biological outcomes when comparing control and intervention conditions. This is the first systematic review focused on social-network based condom promotion interventions among pre-existing social networks. Although the reviewed interventions are not mutually exclusive from community popular opinion leader (CPOL) studies and other peer intervention programs, there are several distinctions that differentiate these social network based condom promotions. Most importantly, these social network interventions all used social networks that were defined by the individuals themselves. The CPOL model chooses leaders of the community, but may not identify the smaller social networks that impact one’s daily interactions. Most of these reviewed studies (10/11) had some measure of intervention effect within the identified network. The network members were either directly involved in the intervention, or indirectly when receiving information from the index participants. Focusing the review on pre-existing social networks allowed us to compare studies that were not post-hoc defined by researchers or others who may be unfamiliar with endogenous social connections. This review demonstrates the benefits and limitations of incorporating pre-existing social networks and relationships when designing condom promotion interventions.
Network-based condom promotion interventions have several advantages that were highlighted in these studies. Ten of the eleven reviewed studies showed substantial improvements in condom outcome measures. The one study that did not find increases in condom use, did find decreases in incident gonorrhea in the intervention group compared to the control. The median time of terminal follow-up was 12 months, substantially greater than 6.3 months found in a systematic review including non-network based condom promotion interventions . In addition, ten of the 11 studies had retention rates exceeding 80% at the last follow-up visit. The extended follow-up and high retention rate at terminal follow-up visit suggests that groups who are recruited through social network based methods may be easier to follow and retain compared to non-network recruitment methods. Social network based interventions may also be a valuable tool for recruiting marginalized and hard-to-reach populations.
This systematic review also highlighted several limitations both of network based condom promotions and in the assessment of their efficacy. First, social network interventions require the motivation of index participants to recruit network participants as well as disseminate information among their social networks. Additionally, these interventions are methodologically complex, and understanding the social networks before program implementation is crucial. Contamination is a somewhat inevitable research limitation in the studies that used controls, especially in those that studied densely connected social networks. However, while contamination does impact the evaluation of these types of interventions, it does not compromise the potential utility in implementing social network interventions and HIV prevention programs. The reviewed studies only assessed individual outcomes because there are not standardized measures for condom use norms or other group-level determinants of condom use. Several studies have analyzed group norms, but there have been no validated measures focused at this level of analysis [27, 28].
This systematic review has several limitations. First, different methods of demonstrating the effect of a social network based intervention were used. While nine studies had a control group, these studies used varying condom outcome measures. Other reviews of condom promotion interventions have noted heterogeneity in outcome measures . Differential effects of the intervention on casual versus regular partners may have underestimated the net effect of the intervention in the six studies that did not make this differentiation. Second, the reviewed studies had different recruitment methods and differed slightly in the way in which networks were defined and analyzed. Third, measurement of study quality was difficult as there currently is no gold standard for assessing study quality in social network based intervention studies. Finally, although we conducted an extensive search of the literature, some articles that did not focus on condom promotion as a means of HIV/STI prevention may not have been included.
This systematic review demonstrates some areas that are critical for future social network research focused on HIV prevention. More standardized mechanisms for evaluating contamination between intervention and control groups are essential for clearly showing the effect of social network based interventions. Although social relationships between peers and members of risk groups have been studied in greater detail, the local social worlds of brothels and other commercial sex establishments are less clear. Social networks bridging pimps, managers, and other men who control local sex industries may have a strong influence on condom use, either positive or negative. In addition, better understanding networks of men who purchase sex may provide a strong force for organizing condom promotion and related HIV prevention programs  in some regions where such networks exist.
The results of this review highlight the potential utility of social network based intervention programs. Previous network studies of high risk groups have focused on passive disease transmission, but this body of research reveals how high risk social networks can be used to actively promote STI/HIV prevention. The results of the reviewed interventions suggest that the use of network-based approaches lead to increases in condom use and decreases in other HIV-risk behaviors. Further research on how to recruit, retain, and change behaviors among social networks may hold great promise for behavioral STI/HIV prevention.
The authors would like to thank Professor Carl Latkin for helpful comments on an earlier version of this manuscript. The authors would also like to thank Bonnie Tucker for translating the abstract.
Kaidi Wang, Harvard Medical School, 25 Shattuck Street, Boston, MA 02115, USA.
Katherine Brown, Columbia University College of Physicians and Surgeons, 630 W. 168th St., New York, NY 10032, USA.
Song-Ying Shen, Guangzhou Birth Cohort Office, Guangzhou Women and Children’s Medical Center, 9 Jinsui Road, Guangzhou 510623, China.
Joseph Tucker, Division of Infectious Diseases, Massachusetts General Hospital, 55 Fruit Street, GRJ 504, Boston 02114, USA.