Multi-Agent AI Systems in Healthcare: Systematic Evidence Synthesis Via PRISMA of Clinical Decision Support Systems, Robotic Interventions, and Critical Care

Abstract: Multi-agent artificial intelligence systems (MAS) is transforming the healthcare sector by enabling intelligent decision-making, robotic interventions, and critical care management. Despite growing interest, a comprehensive synthesis of their real-world applications, performance, and limitations is lacking. This systematic review aims to evaluate the deployment of MAS in healthcare across three key domains: Clinical Decision Support Systems (CDSS), Robotic Interventions, and Critical Care Monitoring. It also addresses ethical and legal considerations, offers a comparative framework, and identifies research gaps. A systematic search of PubMed, IEEE Xplore, Scopus, and Web of Science was conducted following PRISMA guidelines. A total of 150 records were retrieved. After duplicate removal and screening, 32 studies met inclusion criteria. The Joanna Briggs Institute (JBI) Critical Appraisal Checklist guided quality assessment. Data were extracted using a structured form and organized by topic. Among the 32 studies included, 12 focused on MAS-based CDSS, 11 on robotic interventions, and 9 on MAS in critical care. Most MAS demonstrated improved diagnostic accuracy, aided real-time decision support, and enhanced patient monitoring. However, over 60% involved practical models and lacked clinical validation. Ethical concerns, particularly around autonomy, data privacy, transparency, and legal accountability were sometimes overlooked. Only 7 studies discussed ethical or legal implications in depth. A comparative framework was developed, comparing MAS tools based on scalability, clinical validation, autonomy, human-AI interaction, and data transparency. No doubt, MAS hold transformative potential across healthcare domains. However, real-world deployment remains unclear. Greater emphasis on clinical validation, ethical design, and regulatory frameworks is needed. This review provides suggestions for future research and responsible adoption of MAS in healthcare.