连续贝叶斯网络在尿酸与慢性代谢性疾病相关性中的应用  

Application of continuous Bayesian networks in the association study between uric acid and chronic metabolic diseases

作　　者：崔宇 宋伟梅 赵瑞青 任浩 王旭春 乔宇超 赵执扬 任家辉 刘静 李一汀 仇丽霞 CUI Yu;SONG Weimei;ZHAO Ruiqing;REN Hao;WANG Xuchun;QIAO Yuchao;ZHAO Zhiyang;REN Jiahui;LIU Jing;LI Yiting;QIU Lixia(Department of Health Statistics,School of Public Health,Shanxi Medical University,Taiyuan 030001,China)

机构地区：[1]山西医科大学公共卫生学院卫生统计学教研室,太原030001

出　　处：《中华疾病控制杂志》2023年第9期1078-1083,共6页Chinese Journal of Disease Control & Prevention

基　　金：国家自然科学基金(81973155)。

摘　　要：目的基于改进的偏相关(improved partial-correlation-based,IPCB)算法建立连续贝叶斯网络模型,探寻尿酸(uric acid,UA)的代谢性影响因素,并通过与传统的多重线性回归模型分析比较,分析连续贝叶斯网络模型对疾病影响因素的效果和优势。方法以2015年山西省慢性病监测的4846例监测人群数据为研究对象,分别用多重线性回归模型和连续贝叶斯网络模型分析UA与其余9个代谢性疾病的特征指标的相关性,比较两种方法结果的优劣。结果经多重线性回归模型逐步分析,三酰甘油(triglyceride,TG)、SBP、DBP、低密度脂蛋白(low density lipoprotein,LDL)、高密度脂蛋白(high density lipoprotein,HDL)共5个变量与UA水平直接相关,解释了UA 9.5%的变异。连续贝叶斯网络模型共包含24条有向边,年龄、TG、LDL、HDL、SBP、DBP与UA水平直接相关,随着年龄、TG、LDL的增加和HDL的降低均会导致UA水平升高,而UA水平升高又导致SBP、DBP升高;总胆固醇(total cholesterol,TC)与UA间接相关。结论连续贝叶斯网络模型能发现更多UA的直接影响因素,还可以找到UA的间接影响因素,整体解释度更好。Objective The study aimed to construct a continuous Bayesian networks using improved partial-correlation-based(IPCB)algorithm,investigate the metabolic influencing factors of blood uric acid(UA).Meanwhile,traditional multiple linear regression model was compared to discuss the effect and advantages of Bayesian network in disease risk factor detection.Methods Chronic disease surveillance data involving 4846 cases in Shanxi Province in 2015 was taken for study.Both continuous Bayesian networks and multiple linear regression models were utilized for correlation analysis between UA and the indexes of nine metabolic diseases.Afterwards,the advantages and disadvantages of the two models were compared.Results Multiple linear regression stepwise analysis showed that triglyceride(TG),systolic blood pressure(SBP),diastolic blood pressure(DBP),low density lipoprotein(LDL)and high density lipoprotein(HDL)were directly correlated with UA,explaining 9.5%of the variation.The continuous Bayesian networks comprised 24 directed edges,and age,TG,LDL,HDL,SBP,and DBP were directly related to UA.As age,TG,LDL increase and HDL decreases,UA levels rise,which in turn leads to an increase in SBP and DBP.Besides,total cholesterol(TC)was indirectly related to UA.Conclusions Continuous Bayesian networks allows for more direct and indirect influencing factors for UA,which features a better overall explanation.

关 键 词：连续贝叶斯网络 改进的偏相关算法 多重线性回归 尿酸 

分 类 号：R195.1[医药卫生—卫生统计学]