中华预防医学杂志
中華預防醫學雜誌
중화예방의학잡지
CHINESE JOURNAL OF
2014年
4期
296-300
,共5页
张娜%王国永%朱晓艳%杨兴光%康殿民
張娜%王國永%硃曉豔%楊興光%康殿民
장나%왕국영%주효염%양흥광%강전민
HIV%贝叶斯定理%模型,结构%获得性免疫缺陷综合征
HIV%貝葉斯定理%模型,結構%穫得性免疫缺陷綜閤徵
HIV%패협사정리%모형,결구%획득성면역결함종합정
HIV%Bayes theorem%Models,structural%Acquired immunodeficiency syndrome
目的 构建AIDS发病及影响因素的贝叶斯网络模型,阐述AIDS发病与影响因素间的关系.方法 以山东省1992-2011年确证报告的2 431例HIV抗体阳性者的随访资料为信息源,依据必要路径条件运算(NPC)构建网络拓扑结构,期望最大化运算法则(EM)训练样本计算网络节点条件概率,构建AIDS发病与传播途径、样本来源、随访干预、抗病毒治疗、机会性感染治疗及HIV抗体确证阳性时的CD4+T淋巴细胞计数等影响因素的贝叶斯网络模型,并用该网络进行贝叶斯推理.结果 2 431例研究对象中有49.77％(1 210/2 431)发病成为AIDS患者.通过对2 431个样本资料学习,构建了一个含有7个节点、11条有向边的AIDS发病与影响因素的贝叶斯网络模型,受试者工作特征(ROC)曲线下面积为0.75.样本来源、传播途径、确证阳性时CD4+T淋巴细胞计数、接受抗病毒治疗、机会性感染治疗和随访干预情况均与AIDS发病有直接的因果联系,其中,接受抗病毒治疗且随访干预依从性好者的发病概率为42.83％,治疗后未接受随访干预病例者发病概率为62.03％.经同性传播病例接受随访干预概率为68.96％,其由医疗机构、检测咨询、监管场所和专题调查检测报告的概率分别为34.00％、42.24％、1.06％和22.70％.结论 通过网络推理揭示了AIDS发病多因素间、多层次的相互关系及影响强度,接受抗病毒治疗且随访干预依从性好的病例发病概率较低.
目的 構建AIDS髮病及影響因素的貝葉斯網絡模型,闡述AIDS髮病與影響因素間的關繫.方法 以山東省1992-2011年確證報告的2 431例HIV抗體暘性者的隨訪資料為信息源,依據必要路徑條件運算(NPC)構建網絡拓撲結構,期望最大化運算法則(EM)訓練樣本計算網絡節點條件概率,構建AIDS髮病與傳播途徑、樣本來源、隨訪榦預、抗病毒治療、機會性感染治療及HIV抗體確證暘性時的CD4+T淋巴細胞計數等影響因素的貝葉斯網絡模型,併用該網絡進行貝葉斯推理.結果 2 431例研究對象中有49.77％(1 210/2 431)髮病成為AIDS患者.通過對2 431箇樣本資料學習,構建瞭一箇含有7箇節點、11條有嚮邊的AIDS髮病與影響因素的貝葉斯網絡模型,受試者工作特徵(ROC)麯線下麵積為0.75.樣本來源、傳播途徑、確證暘性時CD4+T淋巴細胞計數、接受抗病毒治療、機會性感染治療和隨訪榦預情況均與AIDS髮病有直接的因果聯繫,其中,接受抗病毒治療且隨訪榦預依從性好者的髮病概率為42.83％,治療後未接受隨訪榦預病例者髮病概率為62.03％.經同性傳播病例接受隨訪榦預概率為68.96％,其由醫療機構、檢測咨詢、鑑管場所和專題調查檢測報告的概率分彆為34.00％、42.24％、1.06％和22.70％.結論 通過網絡推理揭示瞭AIDS髮病多因素間、多層次的相互關繫及影響彊度,接受抗病毒治療且隨訪榦預依從性好的病例髮病概率較低.
목적 구건AIDS발병급영향인소적패협사망락모형,천술AIDS발병여영향인소간적관계.방법 이산동성1992-2011년학증보고적2 431례HIV항체양성자적수방자료위신식원,의거필요로경조건운산(NPC)구건망락탁복결구,기망최대화운산법칙(EM)훈련양본계산망락절점조건개솔,구건AIDS발병여전파도경、양본래원、수방간예、항병독치료、궤회성감염치료급HIV항체학증양성시적CD4+T림파세포계수등영향인소적패협사망락모형,병용해망락진행패협사추리.결과 2 431례연구대상중유49.77％(1 210/2 431)발병성위AIDS환자.통과대2 431개양본자료학습,구건료일개함유7개절점、11조유향변적AIDS발병여영향인소적패협사망락모형,수시자공작특정(ROC)곡선하면적위0.75.양본래원、전파도경、학증양성시CD4+T림파세포계수、접수항병독치료、궤회성감염치료화수방간예정황균여AIDS발병유직접적인과련계,기중,접수항병독치료차수방간예의종성호자적발병개솔위42.83％,치료후미접수수방간예병례자발병개솔위62.03％.경동성전파병례접수수방간예개솔위68.96％,기유의료궤구、검측자순、감관장소화전제조사검측보고적개솔분별위34.00％、42.24％、1.06％화22.70％.결론 통과망락추리게시료AIDS발병다인소간、다층차적상호관계급영향강도,접수항병독치료차수방간예의종성호적병례발병개솔교저.
Objective To explore the influencing factors of AIDS pathogenesis using the Bayesian network.Methods Based on follow-up data of 2 431 cases of HIV/AIDS from 1992-2011 in Shandong province,this study constructed the network structure by NPC algorithm,and used the EM algorithm for parameter learning to construct the Bayesian network of influencing factors and AIDS pathogenesis,then did inference by the Bayesian network.Results A total of 49.77％ (1 210/2 431)were AIDS.Get a Bayesian network with 7 nodes and 11 directed arcs and the related parameters by studying the follow-up data of 2 431 cases.The area under receiver operating curve (ROC) was 0.75.There was a direct causal association among sample resource,transmission route,CD4 + T lymphocyte count of HIV-antibody confirmed positive,antiviral therapy,opportunistic infection therapy,follow-up intervention and AIDS pathogenesis.The incidence probability was 42.83％ for those who received antiviral therapy and follow-up intervention,and it was 68.96％ for those who received antiviral therapy without follow-up intervention.The probability to receive follow-up intervention was 68.96％ for cases transmitted by homosexual behaviors,and it was 34.00％,42.24％,1.06％ and 22.70％ respectively to be reported by medical institutions,testing and counselling,supervision institutions and special surveys.Conclusion The Bayesian network revealed the mutual relation and effect intension among multi-factors and multi-stages by network inference.It showed that the rate of AIDS pathogenesis was lower for those who received antiviral therapy and follow-up intervention.
