HIV prevention through social networks: evidence from a randomized controlled trial of a peer advocacy intervention in Kampala, Uganda Authors: Glenn J. Wagner, Laura M. Bogart, Joseph K. B. Matovu, Violet Gwokyalya, Harold D. Green, Susan Ninsiima, David J. Klein, Ryan K. McBain, Stephen Okoboi Published: 2025 ABSTRACT In a secondary analysis from a randomized controlled trial of a peer advocacy training intervention for people living with HIV (PLWH), we examined effects on advocacy targeting specific HIV protective behaviors among enrolled social network members (alters), and whether alter behaviors were predicted by advocacy receipt and tone of delivery. 210 PLWH and 599 alters enrolled and were followed over 18 months. Repeated measures logistic regressions showed that PLWH in the intervention group targeted more alters with advocacy for HIV testing, condom use, pre-exposure prophylaxis, and HIV care, compared to those in the control group. Alters targeted with HIV testing advocacy were more likely to report increased HIV testing at the subsequent assessment. Among alters with HIV-negative main partners, receipt of condom use advocacy predicted increased consistent condom use at the subsequent assessment. Advocacy training for PLWH can increase targeted HIV prevention advocacy, which then promotes HIV testing and condom use. Introduction Nearly 1.5 million people are living with HIV (PLWH) in Uganda, and the HIV prevalence has remained above 5% for over two decades (Uganda AIDS Commission, 2024). About one-quarter of those living with HIV do not know their HIV status, the rates of consistent condom use are low, and access to post-exposure prophylaxis (PrEP) remains poor (Vithalani & Herreros-Villaneuva, 2018; Nabunya et al., 2023). These challenges to HIV prevention and control are further exacerbated by the drastic reduction in funding for HIV care and prevention from the U.S. President’s Emergency Plan for AIDS Relief (PEPFAR) (UNAIDS, 2025). PLWH can contribute to HIV prevention within their communities by engaging in advocacy for HIV protective behaviors within their social networks. Our research suggests that as PLWH receive antiretroviral therapy (ART) and restore their health, they are motivated to protect their family and others in their social network, and therefore engage in prevention advocacy (i.e., encourage friends and family to seek HIV testing and care and reduce HIV risk) (Ssegujja et al., 2011; Tumwine et al., 2012). However, the success of this advocacy may be hampered by challenges related to message content, and style and tone of delivery (Tumwine et al., 2012). Peer advocacy interventions have been shown to have benefits for health behavior change (Kelly, 2004; Kegeles et al., 1999). Therefore, with effective advocacy training, mobilizing PLWH to be change agents within their social networks has the potential to be a “game changer” for HIV prevention, particularly in high-prevalence settings where virtually every family is touched by HIV. Ugandan PLWH have dense, interconnected networks (Green et al., 2011), and PLWH have often disclosed to many people in their network (Ssali et al., 2010), suggesting that transfer of HIV prevention messages and knowledge can be safe and efficient. Further, PLWH can be influential and credible in conveying prevention messages to family and friends, given their close and trusted relations, and their personal lived experience with HIV (Raja et al., 2007). To maximize the potential impact of peer advocacy for HIV prevention, we developed the Game Changer for HIV Prevention (GC-HIV) advocacy training intervention. GC-HIV is a multi-session group intervention that empowers PLWH to engage in advocacy with people in their social networks (Wagner et al., 2022). Game Changers draws on social diffusion theory (Valente, 1996; Centola, 2018) and principles of social influence (Broadhead et al., 1998), positing that behavior change can be initiated by a few and diffused to others through social norm change. Through advocacy training, the intervention seeks to harness the persuasion power of people in the network (i.e., the PLWH recipients of the intervention) whom others view as supportive and credible. We recently completed a randomized controlled trial of GC-HIV and published findings revealed that the intervention significantly increased engagement in HIV prevention advocacy with members of their social network, but no significant effects on HIV testing and condom use of social network members (Bogart et al., 2025). However, these findings only pertained to a general measure of advocacy, i.e. any discussion about HIV, and not measures of advocacy that targeted specific HIV prevention behaviors (e.g., HIV testing, condom use, PrEP use, HIV care engagement). It is possible that advocacy targeted to specific HIV protective behaviors may show stronger effects on the targeted behavior of the social network member. Furthermore, the tone in which advocacy is delivered may also influence the targeted behavior. Motivational enhancement principles suggest that communication that avoids judgmentalism and confrontation are more apt to promote behavior change (Moyers et al., 2005). To better understand how the GC-HIV intervention impacts various forms of HIV prevention advocacy, and how aspects of advocacy relate to the targeted HIV protective behaviors, we conducted a secondary analysis of data from the randomized controlled trial of GC-HIV. We examined the intervention’s effect on advocacy that targeted specific HIV protective behaviors (i.e., HIV testing, condom use, PrEP use, HIV care engagement, use of ART) among enrolled social network members (referred to as “alters”) who were followed prospectively over 18 months. Further, we examined whether alter use of specific HIV protective behaviors was predicted by their receipt of advocacy from the enrolled PLWH, as well as the tone in which the advocacy was delivered. Methods Study design The study was a parallel randomized controlled trial of the GC-HIV group peer advocacy intervention, conducted at The Infectious Diseases Institute (IDI) in Kampala, Uganda. PLWH were enrolled as index participants and randomized to receive either the intervention or a nonintervention usual care control group using a blocked 1:1 randomization design with stratification by sex. Each index participant was asked to recruit up to four social network members (alter participants) to whom they had disclosed their HIV status. After baseline, follow-up assessments were administered to both index and alter participants at months 6, 12 and 18. Participants received 30,000–70,000 Uganda shillings (~$8–20 USD), depending on the distance traveled, after completing each assessment, or attending each intervention session, to cover transportation costs. The study protocol was approved by ethical review boards at RAND (2021-N0155) and IDI (009–2021), and cleared by the Uganda National Council for Science and Technology. All study participants provided written informed consent. Further details of the study protocol are available in a prior publication (Wagner et al., 2022). The study has been registered with clinicaltrials.gov (NCT05098015, registered 2021-10-18). Participants Recruitment took place between January 2022 and February 2023. Eligibility criteria for index participants included: (1) age ≥18 years, (2) in HIV care for at least one year (at which point they are more likely to have adjusted to their HIV diagnosis, and have disclosed their HIV status to several people, and thus more ready to engage in advocacy), (3) speak fluent Luganda, and (4) does not have a partner/spouse or household member already enrolled in the study as an index participant. Index participants recruited members of their personal social network to participate in the study as “alter” participants. During the baseline interview of the index participant, we elicited the names of up to 20 alters in their social network and asked the participant to identify those who knew the index participant’s HIV status. At the end of the interview, the index participant was asked to select up to four alters from the list of those who knew their serostatus and were at least 18 years of age whom they would be willing to refer to the study. If they selected more than four alters, we randomly selected four to target for recruitment (to limit selection bias). The index participant was asked to call each selected alter and describe the study in the presence of the coordinator, who then scheduled a study visit for the alter. If an alter refused to participate or could not be reached, the coordinator randomly selected a remaining alter from the list of those referred by the index participant until four had agreed and enrolled. Midway through the study, we allowed index participants to recruit more than four alters to reach our enrollment goal. Intervention The intervention, administered to index participants (in groups of 15) by two trained peer facilitators with experience in provision of HIV education, consisted of eight weekly group sessions, each of which lasted about 2 hours. The intervention provided training on HIV prevention advocacy and index participants were encouraged to engage in advocacy with members of their social network (including those enrolled as alter participants). The session content focused on use of self-compassion and peer support to overcome internalized stigma, HIV disclosure decision making, HIV facts and myths, healthy positive living with HIV, and learning and practicing skills for HIV prevention advocacy (how to start and sustain conversations about HIV). The sessions included sharing of experiences to build support; role playing to build skills and self-efficacy; setting personal goals regarding disclosure and advocacy; and take-home activities to reinforce new skills and generate personal experiences to be processed in the sessions. The sessions were conducted in Luganda by two trained peer-facilitators using a structured manual. The facilitators were trained by the senior investigators over three days. The supervisor of the facilitators observed the implementation of each session and provided feedback and further training as needed during weekly supervision. Measures Survey assessments were interviewer-administered using Network Canvas software20 and conducted in Luganda or English, depending on the preference of the participant. All measures were translated from English to Luganda using a standard translation/backtranslation methodology in a pilot study (Bogart et al., 2020). Index participant engagement in HIV prevention advocacy For each of their alters enrolled in the study, index participants were asked in separate yes/no questions if they had talked with the alter in the past three months about (1) HIV testing (if alter was not living with HIV), (2) condom use, (3) use of PrEP (if alter was not living with HIV), and if alter was living with HIV—(4) engagement in HIV care, and (5) use of ART. For each enrolled alter, dichotomous variables for each of the five HIV protective behaviors was derived for analysis to represent whether the index participant had engaged in advocacy for that behavior with the alter. Alter participants were also asked with yes/no questions if they had talked with the index participant who referred them to the study about the above five HIV protective behaviors (in separate questions for each behavior) during the three months prior to their assessment. For each of these five areas of advocacy, if the alter reported that such a conversation with their index participant had taken place, three follow-up questions were asked to assess the tone of advocacy delivery; the alter participant was asked if the index participant (1) showed caring, (2) was confrontational, and (3) was judgmental, during the conversation, in separate items. For each of these single-item delivery tone measures, the response options were 1 ‘not at all’, 2 ‘a little’, 3 ‘somewhat’, and 4 ‘very much’. Alter use of HIV protective behaviors Condom use in the past six months with main and casual sex partners—in separate questions—was assessed. Participants who reported having a main sex partner and having sexual intercourse with that partner in the past six months were asked how frequently they used condoms during sexual intercourse with their main partner; response options ranged from 1 ‘never’ to 5 ‘always.’ This condom use item was also asked regarding sexual intercourse with casual sex partners in the past six months, if present. For condom use with all present partners (main and/or casual), a binary variable was created to represent consistent condom use, defined as always using condoms during sexual intercourse with all reported partners (main and/or casual) in the past six months. To assess HIV testing behavior, the participant was first asked if they had ever tested for HIV; those who had tested were then asked to report when their last HIV test was done and the result of that test. A binary variable was created to represent recent HIV testing, defined as the most recent test being within the past six months. Participants who reported living with HIV were asked if they were receiving HIV care, and if they were using ART, in separate (yes/no) questions. Participants who did not report living with HIV were asked about use of PrEP (ever and currently). Background characteristics For both index and alter participants, these included age (in years), sex (male, female), highest level of education, and relationship characteristics (presence of a spouse, romantic partner, or main sex partner; HIV status of that partner). Index participants were asked the nature of relationship (e.g., family, friend, other such coworker) with the enrolled alter participant. HIV medical characteristics of the index participants included time since HIV diagnosis, time in HIV care, and ART status, all of which were abstracted from participants’ medical charts. Data analysis Bivariate statistics (independent 2-tailed t-tests; chi square tests) were used to examine background characteristics of the sample between those in the intervention vs. controls groups, with separate analyses for index and alter participants. To test intervention effects on index participant engagement in measures of HIV prevention efficacy, we compared advocacy measures between the intervention and control arms at all follow-up time points using logistic regression analysis. All follow-up assessments were pooled in these regression models to increase statistical power to detect change from baseline, and the primary predictor was treatment arm. For models where the outcomes were alter reported HIV protective behaviors, ‘lagged’ regression models were run where the primary predictor was alter report of receipt of HIV prevention advocacy or alter perceived tone of advocacy delivery measured in the previous survey assessment; these models were not examining intervention effects, but we controlled for treatment condition by including study arm as a covariate. These repeated measures logistic regression models included fixed effects for the intervention condition, the baseline value of the outcome, and categorical time. To account for potential correlation of responses within an individual over time, models also included random effects for the index participant and for alter nested within index participant to account for potential correlation of responses from the same alters over time, as well as multiple alters from the same referring index participant. The regressions included all individuals with values for the outcome at baseline and at least one follow-up assessment, with each participant contributing up to three follow-up responses (from the 6-, 12-, and 18-month assessments). Results Sample characteristics Two hundred and ten index participants (105 in each of the intervention and control arms) enrolled in the study, from whom 599 alter participants enrolled [a median of 3 alters (maximum was 7) were enrolled per index participant]. Table 1 shows characteristics of enrolled index and alter participants overall and within the control and intervention arms; there were no statistically significant differences between arms on any characteristic. Among both index and alter participants, the mean age was near 40 years, roughly a third were male, and most had at least some secondary education and were in a committed relationship. Over half (56.9%) of the enrolled alters were family members of the index participant, while others were friends (42.4%) and/or romantic/sex partners (10.0%). Retention through the month 18 follow-up assessment was very high among both index (n = 208; 99%) and alter (n = 596; 99%) participants. PLWH engagement in HIV prevention advocacy with enrolled alters At baseline, engagement in HIV prevention advocacy regarding specific HIV protective behaviors by index participants with their enrolled alters ranged from 34.4% − 71.3%. These levels of advocacy did not differ between index participants in the control or intervention arms. Controlling for baseline level of advocacy and time, index participants in the intervention arm reported increased advocacy with a significantly higher percentage of enrolled alters for each of the HIV protective behaviors except for ART use, compared to those in the control group (see Table 2). Index participants in the intervention arm reported higher levels of advocacy for HIV testing and PrEP use among alters who were not living with HIV, higher advocacy for HIV care with enrolled alters living with HIV, and higher condom use advocacy among all enrolled alters who reported any sexual intercourse in the six months prior to the follow-up assessment. Alter HIV protective behaviors At baseline, nearly all alters had tested for HIV in their lifetime (n = 578; 96.5%), of whom just over a third (n = 207; 35.8%) reported living with HIV. Of the 207 alters who were PLWH, 204 (98.6%) reported being in HIV care, and 203 (98.1%) reported being on ART. Among those who reported not living with HIV, only 11 reported ever using PrEP, including 6 who were current users at baseline. Due to the very high rates of HIV care and ART use among PLWH alters and the very low PrEP use among the other alters, we concentrated our subsequent analysis on recent (past 6 month) HIV testing and condom use. HIV testing At baseline, HIV testing in the past six months was relevant for 392 alter participants who did not report living with HIV. Of these 392, 178 (45.4%) reported being tested for HIV in the six months prior to baseline. The lagged logistic regression analysis showed that alters reported HIV testing at the next assessment in 79.4% of the index-alter dyad measurements where HIV testing advocacy occurred in the prior three months, compared to 69.9% of alters in measurements where HIV testing advocacy did not occur [OR (95% CI) = 1.54 (1.07, 2.21); p =.02; d = .24]. In analyses involving alters who reported that HIV testing advocacy took place (n = 293), alter perception of how judgmental the index was in those discussions was marginally predictive of the alter being less likely to report HIV testing at the subsequent assessment [OR (95% CI) = 0.73 (0.53, 1.03); p = .07; d = .17]. Neither alter perceived level of care [OR (95% CI) = 1.37 (0.67, 2.81); p = .39] from the index, nor how confrontational the index was in the discussions [OR (95% CI) = 0.97 (0.65, 1.44); p = .87], were associated with HIV testing at the subsequent assessment. Consistent condom use At baseline, 480 alters reported having a main sex partner (n = 459) and/or casual partner (n = 116) in the past six months, of whom 430 reported having sexual intercourse with either type of partner in the past six months. Of these 430, 60 (14.0%) reported always using condoms during sexual intercourse, regardless of type of sex partner. The lagged logistic regression analysis showed that in 16.4% of the index-alter dyad measurements where condom use advocacy occurred in the prior three months, the alter reported always using condoms at the subsequent assessment, compared to 11.0% of measurements in the dyads where condom advocacy did not occur [OR (95% CI) = 1.57 (0.96, 2.57); p =.07; d = .25). In analyses involving alters who reported that condom use advocacy took place (n = 221), there was no association between alter perception of how judgmental [OR (95% CI) = 0.73 (0.42, 1.27); p = .27], caring [OR (95% CI) = 1.92 (0.62, 5.90); p = .26], or confrontational [OR (95% CI) = 0.97 (0.67, 1.40); p = .88] the index was in the discussion, with alter report of always using condoms at the subsequent assessment. We replicated these analyses among alters with a main sex partner (stratified by main partner HIV status). Among alters whose main partner was not living with HIV, receipt of condom advocacy was associated with an increased likelihood of always using condoms at the subsequent assessment, compared to those who did not receive advocacy [13.2% vs. 6.5%; OR (95% CI) = 2.79 (1.47, 5.30), p =.002; d= .57]. Among alters whose main sex partner was living with HIV, receipt of condom advocacy did not change whether condoms were always using [OR (95% CI) = 0.47 (0.17, 1.32), p =.15]. Discussion This secondary analysis showed that the GC-HIV advocacy training intervention served to increase advocacy among social network members (alters) with respect to several specific HIV protective behaviors. Furthermore, the alters who were targeted with advocacy to promote HIV testing were more likely to report increased HIV testing, and alters with HIV-negative partners who received condom use advocacy reported increased consistent condom use. These findings highlight the impact of the intervention for increasing HIV prevention advocacy, and how advocacy that targets specific HIV protective behaviors can change behavior and promote HIV prevention within social networks. Prior research has shown that nearly all PLWH can be expected to engage in some level of HIV prevention advocacy with people they care about, particularly PLWH whose health is stable with the help of HIV care (Ssegujja et al., 2011; Tumwine et al., 2012). Indeed, our data revealed that PLWH in the control group had engaged in advocacy with many of their enrolled alters, even without receiving advocacy training. However, those in the intervention arm reported increased advocacy targeting specific HIV protective behaviors with a significantly higher percentage of their enrolled alters at follow-up. This was true for advocacy that targeted HIV testing, condom use, PrEP use, and HIV care engagement, demonstrating a consistent effect of the intervention across a wide array of content-focused advocacy performed by PLWH. Our data provides empirical support for intervention effects on the ultimate goal of HIV prevention advocacy, which is to promote the use of HIV protective behaviors. Relative to those who did not receive advocacy, alters who received HIV testing advocacy reported increased rates of HIV testing, and alters with HIV-negative partners who received condom use advocacy reported increased consistent use of condoms. Receipt of condom use advocacy was not associated with change in condom use among alters with partners who were living HIV, suggesting that a risk for HIV transmission to the partner may be key to motivating consistent use of condoms, particularly in the context of PrEP being available for some and increasing awareness of the protection against transmission provided by undetectable HIV viral load among those living with HIV (Irungu et al., 2021). Our data did not reveal any insights into how the tone of advocacy influences behavior change. Consistent with principles of motivational enhancement (Moyers et al., 2005), perceived judgmentalism was negatively correlated with the likelihood of the recipient of advocacy getting tested for HIV, but this association was only marginally significant. Our measures of advocacy tone of delivery are vulnerable to recall bias and the use of only single items to measure the construct. Future research should improve upon the measurement of these more nuanced aspects of advocacy. Other limitations of this study include the subjective nature of the self-reported outcome measures. Social desirability could bias self-reported data; this bias is more likely to impact self-reported advocacy engagement among index participants in the intervention arm, but this bias is likely to be balanced across the intervention and control groups with respect to alter reported HIV testing and condom use. The generalizability of our findings is limited by the selection bias related to index participants needing to be in HIV care at an HIV research center of excellence for at least the past year, and the inherent motivation of such participants to participate in advocacy training and research—both of which may differ from people living with HIV more generally. Lastly, future research should assess the frequency of advocacy conversations, as advocacy that takes place over time may differ in its influence on behavior beyond the mere presence of any advocacy. In conclusion, along with our prior results (Bogart et al., 2025; Bogart et al., 2020), these findings demonstrate the effectiveness of the GC-HIV peer advocacy training program for empowering PLWH to engage in more HIV prevention advocacy within their social networks across a wide array of topics with relevance for social network members who are not living with HIV (HIV testing, PrEP, condom use) as well as those living with HIV (engagement in HIV care, ART use, condom use). Further, social network members who received advocacy from PLWH related to HIV testing and condom use were more likely to report increased use of these HIV protective behaviors, suggesting that PLWH can be effective change agents for HIV prevention. In high HIV prevalence settings such as Uganda, promotion of HIV prevention advocacy among PLWH has the potential for wide dissemination of HIV prevention messaging and information, positioning PLWH to play a key role in combating HIV.