Medical AI Agents for Clinical Decision Support: Viewpoint Using the Planning, Action, Reflection, and Memory (PARM) Analytical Lens

Agentic AI in medicine: architectures, applications, evaluation, and challenges for clinical translation

A scoping review with systematic evidence mapping across five electronic sources, screened 1,649 exportable records, and provisionally included 557 unique studies that met predefined criteria for goal-directed task execution, tool use, interaction with external resources, feedback-based refinement, or multi-agent collaboration.

Ask-Audit-Apply: A Framework for Clinical Reasoning with Artificial Intelligence.

Cognitive fog in AI-assisted care: Preserving clinical reasoning, patient education, and shared decision-making.

Effective mitigation must address both individual barriers, including time pressure, variable AI literacy, and reluctance to challenge automated output, and systemic barriers, including weak governance, opaque tools, misaligned incentives, and inadequate monitoring.

AARD: A Risk-Aware Agentic AI Framework for Sequential Clinical Decision Support Using Large Language Models

: Modern healthcare systems require decision-support tools that can operate effectively in complex, uncertain, and data-rich environments. This paper proposes the Adaptive Agentic Risk-aware Decision (AARD) framework, a policy-based approach for modelling clinical decision-making as a sequential and adaptive process. The framework integrates structured clinical data, such as vital signs and laboratory measurements, with un-structured clinical text, where Large Language Models (LLMs) are used to extract contextual representations. AARD employs an agentic architecture in which decisions are generated through a learned policy and refined using a risk-aware mechanism. This allows the system to adapt actions over time based on evolving state representations while incorporating safety considerations into the decision process. A recursive learning mechanism enables continuous policy updates using observed state transitions, supporting consistent decision behaviour across sequential steps. The framework is evaluated using open-source healthcare datasets, with performance assessed through both predictive metrics and decision-oriented measures. The results indicate that integrating multimodal representations with risk-aware policy learning provides a structured approach to sequential decision support. Overall, the proposed framework offers a scalable and interpretable approach for combining agentic decision processes, LLM-based feature extraction, and risk-aware optimisation in health-care settings.

From assistance to autonomy: AI agent systems in cardiovascular medicine—a review of paradigms, architectures, and clinical translation

Background Cardiovascular medicine faces persistent implementation gaps driven by workforce shortages, fragmented data systems, and the cognitive burden of complex clinical decision-making—structural constraints that limit the delivery of guideline-directed care. Artificial intelligence (AI) is transitioning from isolated predictive models toward autonomous agent systems capable of perceiving, reasoning, and acting in clinical environments, offering a potential pathway to address these challenges. Objective This review synthesizes current evidence on AI agent systems in cardiovascular medicine across four interconnected dimensions: technical paradigms enabling agentic functionality (multi-agent systems, digital twins, multimodal integration), emerging clinical applications across the cardiovascular continuum, and governance frameworks essential for responsible translation. Methods We conducted a narrative review of peer-reviewed literature published between January 2020 and March 2026, drawing from PubMed, Web of Science, IEEE Xplore, and Scopus databases. Search terms included combinations of “artificial intelligence”, “AI agents”, “autonomous agents”, “multi-agent systems”, “large language models”, “cardiovascular diseases”, “heart failure”, “digital twins”, and “clinical decision support”. Emphasis was placed on high-quality original research, systematic reviews, and position papers from major cardiovascular societies, with particular attention to developments from 2024 to 2026. Results Agentic AI systems should function as augmented intelligence—enhancing rather than replacing clinical judgment—to close implementation gaps in cardiovascular care. Multi-agent architectures, digital cardiovascular twins, and multimodal integration are emerging as core technical paradigms. The ClinNoteAgents system demonstrates high extraction fidelity (conditional accuracy ≥90%) for clinical variables while achieving 60%–90% text reduction. Digital twin applications span therapy planning, risk prediction, and monitoring, with 69% relying on mechanistic models and 76% utilizing imaging data for personalization. Heart failure has emerged as a paradigmatic use case, driven by structural workforce gaps and the ARPA-H ADVOCATE initiative launched in January 2026. The C.A.R.D.I.O. framework (Clinical validation, Auditability, Risk stratification, Data privacy, Integration, Ongoing vigilance) and the CURACO framework (Clinical safety, Understanding, Research-informed care, Authentic patient-centred approaches, Conscientious ethics, Optimised technology) provide governance structures for responsible deployment. A rapid systematic review of 13 studies including 22,641 participants found that 85% of AI interventions improved cardiovascular outcomes, with mortality reductions of 0.8%–12% and major adverse cardiovascular event reductions of 4%–12%. Conclusion Cardiovascular medicine stands at an inflection point. The transition from assistance to autonomy requires rigorous fit-for-purpose evaluation, transparent interpretability mechanisms, and robust governance frameworks. Agentic AI systems should function as augmented intelligence—enhancing rather than replacing clinical judgment—to close implementation gaps in cardiovascular care.

Clinical AgentOps: Runtime Governance for Autonomous AI Healthcare Agents

ClinicalAgentOps is proposed, a framework that relocates governance from the artifact into the agent’s execution path and contributes an explicit argument from the premises of design-time assurance to the necessity of inpath control.