Artificial intelligence applications in HIV Care domains among men who have sex with men (MSM): a systematic review
Аннотация
Background Men who have sex with men (MSM) remain disproportionately affected by human immunodeficiency virus (HIV), with persistent gaps in prevention, testing, and treatment across selected stages of the HIV care continuum. Artificial intelligence (AI) offers emerging opportunities to support HIV-related services; however, evidence of its application in MSM populations remains limited and fragmented. This study aims to systematically examine emerging evidence on AI applications across selected stages of the HIV care continuum. Methods Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 guidelines, this systematic review was registered in the International Prospective Register of Systematic Reviews (CRD420261292658) on 25 January 2026. It included studies published between 2015 and 2025 that examined MSM and evaluated AI-enabled tools across the HIV care continuum. PubMed and Scopus were searched using structured terms. Two reviewers independently screened studies, extracted data, and assessed risk of bias (using the Joanna Briggs Institute and Mixed Methods Appraisal Checklist). Evidence was synthesized narratively. Results Ten studies from diverse global settings were included. AI applications included machine learning, deep learning, network-based models, and chatbot systems. Predictive models showed promising performance for specific tasks, including HIV risk prediction, Pre exposure prophylaxis adherence prediction, sexual-risk classification, and clinic-attendance forecasting. Evidence for AI-enabled interventions was more limited. Chatbot-based interventions primarily demonstrated feasibility and acceptability, while influencer-identification approaches were evaluated using behavioural or testing-related outcomes, with mixed findings. Overall, evidence demonstrating sustained improvements in clinical or population-level HIV outcomes was limited. Conclusion AI-enabled tools demonstrated promising predictive performance and potential utility for supporting selected HIV prevention and care activities among MSM. However, evidence of real-world effectiveness in improving behavioural, clinical, or service-delivery outcomes remains limited. Further prospective studies, external validation, and real-world implementation evaluations across diverse populations and settings are needed before conclusions regarding clinical or public health effectiveness can be established.