The Prevalence and Associated Factors of HIV Self-Testing Among High-Risk Men Who Have Sex with Men in China: A Study Based on Bayesian Network Model.

Abstract

Promoting HIV self-testing (HIVST) among men who have sex with men (MSM) is an effective strategy for preventing and controlling HIV infection. We aimed to explore the prevalence, associated factors, and probabilistic dependencies of HIVST uptake among HIV-negative MSM who engaged in condomless sex during the past 6 months in China using logistic regression and Bayesian networks. We performed analysis of baseline data from a large longitudinal randomized controlled trial. In total, 1800 MSM were included and 37.5% (n = 675) reported HIVST in the past 6 months. Multivariate logistic regression analysis revealed that HIVST uptake was positively associated with the following factors: higher monthly income, migrant from other cities, sexual partners engaged in HIVST, sex under the influence of substances, and higher frequency of browsing/using WeChat to explore HIV prevention or treatment information and services. HIVST uptake showed negative associations with older age, inconsistent condom use with casual sex partners, seeking sexual partners through gay venues, and having sex under the influence of alcohol. The Bayesian network model revealed direct associations between HIVST uptake and four factors: age, condom use with casual sexual partners, sex partner's HIVST status, and frequency of browsing/using WeChat to explore HIV prevention or treatment information and services. HIVST uptake peaked at 85.7% under the optimal configuration of: age 18-24 years, consistent condom use with casual partners, unknown partner HIVST status, and at least once a month browse/use WeChat to explore HIV prevention or treatment information and services. Future HIVST promotion strategies should prioritize youth-focused digital education via WeChat, normalized condom use reinforcement, and partner-initiated HIVST encouragement to operationalize the synergistic determinants identified by the Bayesian network, maximizing uptake potential under optimized variable configurations.