2025 年 12 巻 1 号 p. 96
Objectives: This study examines the potential of generative AI, specifically ChatGPT®, in enhancing disaster training planning and evaluation. It aims to address the rise in emerging infectious diseases, extreme weather events, and other untested disaster scenarios by assessing how AI can simplify the creation and evaluation of training scenarios, thereby increasing efficiency and effectiveness.
Methods: A qualitative case study was conducted with participants in a tabletop exercise at a health center. ChatGPT® was used to generate materials for disaster drills, including assumptions about health issues, extracting “current situation analysis and issues” from a hypothetical timeline, and developing evaluation criteria. Ethical guidelines were followed to avoid including personal or facility-specific information.
Findings: AI significantly streamlined disaster drill planning and preparation, saving time and effort. It enabled the creation of scenarios tailored to specific needs without requiring extensive specialized knowledge. The evaluation process was enhanced by quantifying content, allowing for detailed feedback to participants.
Implications: This study highlights both the benefits and challenges of integrating generative AI into disaster medicine. While AI can streamline planning and improve scenario quality, issues such as data insufficiency, AI inaccuracies, and the need for human oversight and ethical considerations are critical. The findings emphasize the importance of careful integration and regulation of AI technologies in disaster preparedness.
