2021 年第 9 期 第 16 卷

心脏重症监护病房急性心力衰竭患者心源性死亡的预测模型构建

Construction of predictive model of cardiac death in patients with acute heart failure in cardiac intensive care unit

作者：成力1张霞2

英文作者：Cheng Li1 Zhang Xia2

单位：1陆军军医大学第二附属医院心内科，重庆400037；2贵州省铜仁市德江县人民医院心血管内科565200

英文单位：1Department of Cardiology the Second Affiliated Hospital of Army Medical University Chongqing 400037 China; 2Department of Cardiovascular Medicine Dejiang County People′s Hospital Guizhou Province Tongren 565200 China

关键词：急性心力衰竭；心脏重症监护病房；心源性死亡；预测模型；危险因素

英文关键词：Acuteheartfailure;Cardiacintensivecareunit;Cardiacdeath;Predictivemodel;Riskfactors

• 摘要：

• 目的 分析影响心脏重症监护病房（CICU）急性心力衰竭（AHF）患者心源性死亡的因素并构建预测模型。方法 选择2018年4月至2020年4月陆军军医大学第二附属医院CICU收治的132例AHF患者作为研究对象，根据CICU内是否死亡将患者分为死亡组（23例）和存活组（109例）。多因素Logistic回归分析影响AHF患者CICU内发生心源性死亡的因素；构建XGboost模型并评价其区分度。结果 年龄＞65岁、Killip分级Ⅳ级、心率增加、白细胞计数增多、红细胞计数减少、白蛋白降低、血红蛋白降低、血细胞比容降低、血钠降低、血尿素氮升高、心肌肌钙蛋白I升高、N末端B型脑钠肽前体升高、空腹血糖升高均是影响AHF患者CICU内发生心源性死亡的独立危险因素（均P＜0.05）。上述危险因素在XGboost模型中的重要度评分分别为10、13、22、29、15、20、17、25、19、26、32、35、11分；XGboost模型预测AHF患者CICU内发生心源性死亡的区分度的受试者工作特征曲线下面积为0.867（95%置信区间：0.792～0.942，P＜0.001），区分度良好。结论 XGboost模型预测AHF患者CICU内发生心源性死亡的区分度较高，可为临床上CICU内AHF患者的治疗、护理、改善预后提供参考。

• Objective To analyze the factors affecting the cardiac death of patients with acute heart failure (AHF) in cardiac intensive care unit (CICU) and to construct a predictive model. Methods From April 2018 to April 2020, 132 AHF patients admitted to CICU, the Second Affiliated Hospital of Army Medical University were selected. According to the death in CICU, the patients were divided into the death group (23 cases) and the survival group (109 cases). Multivariate Logistic regression analysis was used to analyze factors affecting cardiac death of AHF patients in CICU; XGboost model was established and its differentiation was evaluated. Results Age＞65 years old, Killip grade Ⅳ, heart rate increase, white blood cell count increase, red blood cell count decrease, albumin decrease, hemoglobin decrease, hematocrit decrease, blood sodium decrease, blood urea nitrogen increase, cardiac troponin I increase, N-terminal pro-brain natriuretic peptide increase and fasting blood glucose increase were independent risk factors for cardiac death of AHF patients in CICU (all P＜0.05). The differentiation scores of the above risk factors in XGboost model were 10, 13, 22, 29, 15, 20, 17, 25, 19, 26, 32, 35 and 11; the area under the receicer operating characteristic curve of the differentiation of XGboost model for predicting cardiac death of AHF patients in CICU was 0.867 (95% confidence interval: 0.792-0.942, P＜0.001), and the differentiation was good. Conclusion The differentiation of XGboost model for predicting cardiac death of AHF patients in CICU is high, which can provides reference for clinical treatment, care and improving prognosis of AHF patients in CICU.