中国感染控制杂志2026，Vol.25Issue(5)：645-651,7.DOI:10.12138/j.issn.1671-9638.20263074
大语言模型在多重耐药菌感染防控领域的应用与评估
Application and evaluation of a large language model in the prevention and control of multidrug-resistant organism infection
摘要
Abstract
Objective To evaluate the knowledge capability of the large language model(LLM)ChatGPT-4o in the field of multidrug-resistant organism infection prevention and control,and explore its potential applications in healthcare-associated infection(HAI)control.Methods A professional knowledge question-and-answer dataset covering 5 categories(basic concept,transmission risk identification,infection control practice,antimicrobial stewardship,surveillance evaluation)was constructed based on guidelines and expert consensus.The dataset was input into ChatGPT-4o to generate responses.Four experts in the HAI management field evaluated the model's output results using a 5-point Likert scale.The accuracy,completeness,and usability of LLM in the field of multidrug-resistant organism infection prevention and control were compared and analyzed.Results Among the 50 questions,ChatGPT-4o achieved an overall accuracy rate of 58.0%,a partially accuracy rate of 40.0%,and an inaccuracy rate of 2.0%.ChatGPT-4o performed best on"basic concept"questions,with an accuracy rate of 86.0%,and worst on"antimicrobial stewardship"questions,with an accuracy rate of 40.0%.The average scores for accuracy,complete-ness,and usability were 4.63,4.70,and 4.57 points,respectively.Conclusion ChatGPT-4o demonstrates an excellent response capability for basic knowledge in the field of multidrug-resistant organism infection prevention and control.Although it has limitations such as content generalization and insufficient decision support ability in complex infection control situations,LLM in the field of HAI control has broad application prospects in the future.关键词
多重耐药菌感染/医院感染控制/大语言模型/人工智能Key words
multidrug-resistant organism infection/healthcare-associated infection control/large language model/artificial intelligence分类
医药卫生引用本文复制引用
于涵,王丽华,杨颖,郑怡,何帅,刁一卓..大语言模型在多重耐药菌感染防控领域的应用与评估[J].中国感染控制杂志,2026,25(5):645-651,7.基金项目
2024年辽宁省教育厅基本科研项目(LJ112410159078) （LJ112410159078）
