The Co-evolution of Online Network Structure and HIV Prevention Behaviors among Peer Change Agents in a Community-based Network Intervention
Lindsay Young, Schneider JohnApplied to public health, peer change agent (PCA) interventions position selected members of a target population in the roles of health educators who disseminate information about a health innovation in their personal networks. The impacts of PCA interventions are typically measured on the basis of observed changes in the health behaviors of the peers with whom PCAs engage. That said, an underexplored secondary impact of these interventions is their temporal effects on the network and behavioral dynamics among PCAs themselves.
In this vein, we evaluate the secondary impact of a community-based network intervention developed to engage young Black men who have sex with men (YBMSM) living in southside Chicago around the HIV prevention pill PrEP. From March 2016-March 2017, YBMSM study participants (N=423) were recruited via peer referral chains and randomized to one of two conditions: (1) an intervention condition (n=209) where participants were trained and motivated to be PrEP PCAs, or (2) an attention control (n=214) where participants were not activated as PCAs but instead participated in a sexual risk workshop. During their PCA training, intervention participants were encouraged and provided opportunities to interact and connect with each other online and offline throughout the course of their 12-month enrollment, whereas control participants received no such guidance.
Data were collected from study participants at baseline and 12-months. A computer-assisted self-administered survey assessed PrEP attitudes, sex and health behaviors, and demographics. Facebook friendship data were also collected from each participant, which enabled us to generate a baseline and 12-month social network among study participants.
With these data, we investigate the intervention’s impact on the co-evolution of Facebook friendships among study participants and two prevention behaviors: (1) their personal adoption of PrEP and (2) the number of non-biomedical risk reduction behaviors they engage in (i.e., serosorting, seropositioning, and condom use). To this end, we will employ stochastic actor-based modeling using RSiena. Our model will consist of two parts: Facebook friendship network change (i.e., selection effects) and PrEP and risk reduction change (i.e., influence effects). The friendship network portion of the model will include ego, alter, and similarity effects for the following covariates on Facebook friendship formation: (1) intervention assignment, (2) PrEP adoption, and (3) engagement in risk reduction behaviors, while also adjusting for structural effects. The behavior portion of the model will evaluate: (1) the influence effects of each behavioral dependent behavior, and (2) the effect of being assigned to the intervention condition on the adoption of PrEP and the adoption of risk reduction behaviors. This portion of the model will also adjust for linear trends as well as participant demographics, sexual orientation (bisexual), and HIV status.
Although PCAs are members of a health intervention’s targeted community, their network and behavioral dynamics are all too often neglected in formal evaluations of the intervention’s impact. We anticipate that results of this analysis will cast attention on this critical yet understudied aspect of community-based social diffusion and will help identify where improvements to PCA training and engagement are needed.