Model-Informed Targeted Network Interventions on Social Networks Among Men Who Have Sex With Men in Zhuhai, China

Abstract

Men who have sex with men (MSM) are at disproportionally high risk for human immunodeficiency virus (HIV) infection in China. The increasing HIV prevalence among MSM highlights the urgent need for effective prevention interventions among MSM. Interventions targeted at individuals who are highly vulnerable to HIV infection have been proven effective in reducing incidence rates. However, existing targeted interventions are limited to small-scale programs. To investigate the effectiveness of large-scale targeted network interventions in real-world settings, we build a stochastic agent-based network model informed by the comprehensive online social networking and dating behavior data and epidemiological data among MSM in Zhuhai, China. With the proposed model, we simulate HIV transmissions and compare the efficacy of different targeted intervention programs. We propose a new method, namely, RiskRank, to prioritize nodes for targeted interventions by incorporating: 1) their topological features on the online social network; 2) the underlying epidemic dynamics; and 3) the position of identified HIV-infected individuals on the sexual network. Results show that the targeted interventions are more effective than random interventions in large-scale HIV epidemic control. The proposed RiskRank method consistently outperforms state-of-the-art baselines in various intervention scenarios.