贝叶斯网络模型在体检结果分析中的应用  

The Application of Bayesian Network Model in Analysis of Physical Examination Results

作　　者：王思聪 石超珺 滕斌[1] 乔鲁燕[3] 李赞华 王镜涵 曾庆嘉 秦亚星 冯珊 Wang Sicong;Shi Chaojun;Teng Bin(Institute for Financial Studies,Shandong University(250100),Jinan)

机构地区：[1]山东大学金融研究院,250100 [2]Department of Statistical Science,Duke University [3]山东大学第二医院疼痛科 [4]Institute of Health Informatics,University College London

出　　处：《中国卫生统计》2020年第6期832-834,839,共4页Chinese Journal of Health Statistics

基　　金：国家重点研发计划“变革性技术关键科学问题”重点专项(2018YFA0703900);国家自然科学基金(11871309)。

摘　　要：目的探讨贝叶斯网络在医务人员健康状况分析中的应用,为医务人员健康管理提供方向和思路。方法通过半朴素贝叶斯(TAN)构建年龄、性别、部门(临床/非临床)等基本信息之间的连接,以此为基础建立贝叶斯网络表示各体检指标间的关联关系。结果在2014-2017年某三甲医院医务人员体检数据上,贝叶斯网络以年龄、性别、肝脏为3个中心结点,建立起与其他体检指标的关联。以中心结点肝脏为条件的分组异常检出率统计及贝叶斯网络推断结果同时显示:该院医务人员肝脏与甲状腺、胆囊、肾脏、体重指标之间的关联性差异有统计学意义。结论贝叶斯网络对于建立医务人员体检管理体系具有可参考价值。Objective To better manage and care for the health of medical staff by examining medical staff′s physical examination results with models using Bayesian network.Methods The connection between age,gender and department(clinical or non-clinical)is constructed by semi-naive Bayes,and by which we establish Bayesian Networks to express the association among physical examination indicators.Results Bayesian network is constructed by three links based on age,gender and liver with other physical examination(PE)indicators with a portion of medical examination data of a tertiary general hospital during 2014-2017.Both results of abnormality rates by group with the PE indicator liver as the central node and results of Bayesian Network inference show that there is a significant correlation of abnormality rates between liver and other PE indicators such as thyroid,gallbladder,kidney and weight.Conclusion Bayesian network is advantageous in establishing a medical examination system for medical personnel.

关 键 词：贝叶斯网络 半朴素贝叶斯 关联性分析 医务人员 体检结果 

分 类 号：R195.1[医药卫生—卫生统计学]