Application of Artificial Intelligence and Machine Learning for Cardiovascular Risk Identification and Stratification Among People Living with HIV: A Scoping Review

Cardiovascular disease (CVD) and stroke are the leading causes of non-AIDS-related mortality among people living with HIV (PLWH) in the antiretroviral therapy (ART) era. Despite a well-documented 1.5 to 2-fold elevation in CVD risk compared to the general population, driven by chronic immune activation, ART-related metabolic effects, and endothelial dysfunction, conventional risk prediction tools, including the Framingham Risk Score (FRS) and ACC/AHA Pooled Cohort Equations (PCE), systematically underestimate absolute cardiovascular risk in PLWH. Artificial intelligence (AI) and machine learning (ML) offer the potential to overcome these limitations by integrating HIV-specific immunological, virological, and pharmacological variables alongside traditional risk factors. However, the scope, methodology, and implementation readiness of AI/ML applications for CVD risk identification and stratification in PLWH have not been systematically mapped. This scoping review was conducted following the five-stage methodological framework of Arksey and O'Malley, as refined by Levac et al., and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA-ScR) guidelines. Six databases were searched from inception through April 30, 2025: PubMed/MEDLINE, Embase, CINAHL, Web of Science, Google Scholar, and medRxiv. Hand-searching of reference lists and ClinicalTrials.gov was also performed. All records were screened in Rayyan by two independent reviewers. Six primary studies meeting all inclusion criteria were identified, representing 16,163 PLWH across China, Malaysia/Hong Kong, Italy, and South Africa. This review maps the extent, range, and nature of evidence on AI/ML applications for CVD and stroke risk identification and stratification in PLWH. Specific objectives are to: (1) characterize the AI/ML algorithms applied, input variables used, and outcomes modeled; (2) assess model performance relative to conventional risk tools; (3) evaluate the integration of HIV-specific variables including CD4 count, viral load, nadir CD4, and ART class; (4) identify methodological, equity, and implementation gaps; and (5) provide a prioritized research agenda for next-generation AI-based risk stratification tools for this population, with particular emphasis on stroke as a distinct and under addressed cardiovascular outcome. Expected outcomes include a comprehensive evidence map of existing AI/ML CVD risk models in PLWH; identification of algorithm categories, performance benchmarks, and HIV-specific variable contributions; characterization of geographic, equity, and outcome gaps in the current literature; and a structured research agenda to guide the development and prospective validation of AI-based CVD and stroke risk stratification tools for diverse PLWH populations, with emphasis on U.S. cohorts, integration of longitudinal HIV biomarkers, inflammatory markers, renal function, and social determinants of health.