A PATIENT-CENTRIC ADAPTIVE AI AGENT FOR REAL-TIME CLINICAL DECISION SUPPORT

Abstract

Artificial Intelligence (AI) is increasingly integrated into healthcare systems to support clinical workflows and medical decision-making. However, most existing agent-based healthcare solutions remain system-centric and governance-focused, offering limited adaptability to real-time patient conditions. This study proposes a patient-centric adaptive AI agent designed to provide real-time clinical decision support while maintaining safety, ethical compliance, and clinician oversight. The proposed framework continuously monitors patient physiological signals, analyzes short-term trends, and dynamically adapts care recommendations within predefined safety boundaries. Unlike governance-heavy architectures, the approach prioritizes direct patient benefit through bounded autonomy and proactive assistance. A simulation-based evaluation using emergency care scenarios was conducted to assess system responsiveness, reliability, and clinician workload. Experimental results demonstrate faster reaction times to critical physiological changes, improved early warning capability, and reduced clinician cognitive burden compared to static rule-based systems. These findings indicate that adaptive patient-centric AI agents can enhance clinical decision support without replacing human judgment. The study contributes a practical and ethically grounded framework aligned with real-world clinical requirements and suitable for future intelligent healthcare deployments.