Presentation (CTI)

OBJECTIVES: AI-assisted healthcare delivery has been proposed as a scalable approach to improving healthcare access and affordability, particularly in resource-constrained healthcare systems. However, inappropriate reliance on AI recommendations may reduce workflow efficiency, increase unnecessary healthcare resource utilization, and limit the real-world benefits of clinical AI deployment. This study evaluated whether adaptive AI-assisted decision support strategies can improve the efficiency and quality of healthcare delivery in simulated clinical decision-making settings.
METHODS: A behavioral decision experiment was conducted using simulated clinical cases across multiple diagnostic domains. Physicians, residents, and medical students completed diagnostic decision-making tasks under two conditions including static AI assistance and adaptive AI-assisted support. The adaptive condition incorporated context-sensitive interventions, including uncertainty cues, evidence-based explanations, and risk-sensitive prompts. Outcome measures included diagnostic accuracy, inappropriate reliance on incorrect AI recommendations, decision time, unnecessary escalation of care, and simulated healthcare resource utilization.
RESULTS: Compared with static AI assistance, adaptive AI-assisted support reduced inappropriate reliance on incorrect AI recommendations and improved diagnostic accuracy. Participants in the adaptive condition demonstrated lower rates of unnecessary escalation and reduced simulated resource utilization, suggesting improved efficiency in healthcare delivery. Adaptive support also improved decision consistency and reduced decision time in moderate-complexity cases, indicating enhanced workflow performance during AI-assisted clinical decision-making.cc
CONCLUSIONS: Adaptive AI-assisted healthcare delivery strategies may improve the efficiency, scalability, and reliability of AI-supported clinical care. By reducing inefficient resource utilization and supporting more appropriate clinical decision-making, adaptive AI support may contribute to more accessible and affordable healthcare systems. These findings highlight the importance of integrating behaviorally adaptive mechanisms into the implementation of clinical AI technologies.