人工智能与围手术期护士在手术室转运中的错误检测：基于模拟的比较

BackgroundThe safe and efficient transfer of patients to the operating room is a critical component of surgical care. The growing integration of artificial intelligence (AI) in healthcare introduces new possibilities for enhancing clinical decision-making.ObjectiveThis study compared the performance of AI and perioperative nurses with varying experience levels in identifying errors during the preoperative patient transfer process.MethodsA controlled simulation study was conducted involving three nurses (novice, intermediate, and expert) and a ChatGPT-4o AI model. The AI component was implemented as a prompt-guided assessment using ChatGPT-4o. It was instructed through structured prompts and reference examples to detect errors across five domains: general overview, invasiveness, makeup, jewelry, and site marking. Using the standardized "Patient Admission Criteria Form for the Operating Room," 30 participants were assessed on essential safety factors, including identification and surgical site marking. Data were collected between September and October 2024 and analyzed using SPSS 26.ResultsWithin this controlled simulation, the most experienced nurse achieved higher scores than the AI model in selected context-sensitive domains, particularly jewelry and site marking. While AI demonstrated high accuracy in structured tasks such as assessing invasiveness, it exhibited greater variability in intricate scenarios. Assessment duration was positively correlated with overall performance (r = 0.441, p < 0.001).ConclusionsThe hypothesis that guided AI would outperform nurse evaluators was not supported. Because inputs differed and one nurse represented each experience level, observed differences cannot be attributed solely to evaluator ability. Further prospective equivalent-input studies with multiple nurses and AI systems are needed before human-AI task allocation.