摘要
目的采用随机森林与Logistic回归相结合的方法分析慢性心力衰竭合并心房颤动患者发生心源性死亡的中西医预后因素。方法运用队列研究方法,选择山东省6家中医院(山东中医药大学附属医院、潍坊市中医院、日照市中医院、淄博市中医院、青岛市中医院、济南市中医院)符合研究标准的慢性心力衰竭合并心房颤动患者,收集其入院时基线资料、合并疾病、中医证素、心脏彩超、实验室检查及中西医药物治疗情况的72项原始数据,出院后每6个月随访1次,获得患者心源性死亡的发生情况。按照是否发生心源性死亡将纳入病例分为终点组及非终点组,运用单因素分析筛选出组间差异有统计学意义的变量并依次纳入随机森林模型,将袋外估算误差率最小的变量集作为患者心源性死亡的预后因素,根据平均基尼指数减少量对预后因素进行重要性排序,并进一步运用多因素Logistic回归分析预后因素的作用性质。结果共纳入病例723例,其中发生心源性死亡99例。单因素分析共筛选出25项变量,逐步随机森林模型结果显示纳入变量数为11时模型的袋外估算误差率最小,最终将平均基尼指数减少量排名前11的变量作为心源性死亡的预后因素,依次为心功能分级、主动脉瓣狭窄、抗凝类药物、益气类中药注射剂、心率、胆红素、舒张压、阳虚证、周围血管疾病、降血脂类药物和血浆氨基末端脑利钠肽前体(NT-ProBNP),多因素Logistic回归分析结果显示心功能分级增高(OR=2.204)、合并主动脉瓣狭窄(OR=12.852)、心率增快(OR=1.008)、胆红素增高(OR=1.025)、舒张压增高(OR=1.029)、合并周围血管疾病(OR=8.751)、NT-ProBNP升高(OR=1.922)是患者心源性死亡的西医危险因素,使用抗凝类药物(OR=0.437)、使用降血脂类药物(OR=0.366)是患者心源性死亡的西医保护因素;阳虚证(OR=1.127)是患者心源性死亡的中医危险因�
Objective The random forest and Logistic regression method were used to analyze the prognostic factors of cardiogenic death in patients with chronic heart failure complicated by atrial fibrillation.Methods Using the method of the cohort study,we selected patients with chronic heart failure complicated by atrial fibrillation in six traditional Chinese medicine(TCM)hospitals in Shandong Province(Affiliated Hospital of Shandong University of TCM,Weifang Hospital of TCM,Rizhao Hospital of TCM,Zibo Hospital of TCM,Qingdao Hospital of TCM,and Jinan Hospital of TCM)as the research object.We collected a total of 72 original data samples during hospitalization,including baseline data,complications,TCM pattern factors,cardiac color Doppler ultrasound,laboratory examination,and Chinese and Western medicinal treatment.After discharge,we followed up with the patients every six months to obtain the incidence of cardiac death.According to the occurrence of cardiac death,the included cases were divided into the endpoint group and the non-endpoint group.The statistically significant variables between the groups were selected by single factor analysis and successively included in the random forest model.The variable set with the smallest estimation error rate outside the bag was taken as the pre-treatment factor of cardiac death,and the importance of prognostic factors was ranked according to the reduction in the average Gini index.Multivariate logistic regression was further used to analyze the nature of prognostic factors.Results A total of 723 cases were included in the study,including 99 cases of cardiogenic death;25 of 72 variables were screened out by single-factor analysis.Results of the stepwise random forest model showed that when the number of included variables was 11,estimation error rate outside the bag of the model was the smallest.Finally,the top 11 variables of the mean Gini index reduction were selected as the prognostic factors of cardiac death.Cardiac function grading,aortic valve stenosis,anticoagulant drugs,Yi
作者
孙聪
戴国华
高武霖
管慧
任丽丽
李润民
刘鑫
李珏
赵泽鹏
付浩然
SUN Cong;DAI Guohua;GAO Wulin;GUAN Hui;REN Lili;LI Runmin;LIU Xin;LI Jue;ZHAO Zepeng;FU Haoran(The Second Affiliated Hospital of Shandong Universitry of Traditional Chinese Medicine,Jinan 20002,China;Affiliated Hospital of Shandong University of Traditional Chinese Medicine,Jinan 250014,China;Shandung Univesity of Traditional Chinese Medicine,Jinan 250014,China)
出处
《北京中医药大学学报》
CAS
CSCD
北大核心
2022年第6期621-629,共9页
Journal of Beijing University of Traditional Chinese Medicine
基金
国家自然科学基金项目(No.81774047)。
关键词
随机森林
LOGISTIC回归
慢性心力衰竭
心房颤动
预后因素
阳虚证
益气
random forest
Logistic regression analysis
chronic heart failure
atrial fibrillation
prognostic factors
yang deficiency pattern
tonifying qi
