天津医药

天津医药 ›› 2023, Vol. 51 ›› Issue (11): 1221-1226.doi: 10.11958/20230055

• 临床研究 • 上一篇    下一篇

急性心肌梗死并发心力衰竭的风险预测模型构建及验证

马萌雪1(), 马萍2, 徐清斌2,△(), 张世昌2   

  1. 1.宁夏医科大学总医院心脑血管病医院老年与特需医学科（邮编750000）
    2.宁夏医科大学总医院心脑血管病医院心血管内科（邮编750000）
  • 收稿日期:2023-01-18 修回日期:2023-06-27 出版日期:2023-11-15 发布日期:2023-11-07
  • 通讯作者: △E-mail：xu-qb@163.com
  • 作者简介:马萌雪（1995），女，硕士，住院医师，主要从事冠心病、心力衰竭的临床研究。E-mail：mmx18309610283@163.com

Construction and verification of risk prediction model for acute myocardial infarction with heart failure

MA Mengxue1(), MA Ping2, XU Qingbin2,△(), ZHANG Shichang2   

  1. 1. Department of Geriatrics and Special Needs, Cardiovascular and Cerebrovascular Disease, the General Hospital of Ningxia Medical University, Yinchuan 750000, China
    2. Department of Cardiovascular, Cardiovascular and Cerebrovascular Disease, the General Hospital of Ningxia Medical University, Yinchuan 750000, China
  • Received:2023-01-18 Revised:2023-06-27 Published:2023-11-15 Online:2023-11-07
  • Contact: △E-mail：xu-qb@163.com

RichHTML

4

PDF (PC)

530

摘要：

目的 分析急性心肌梗死（AMI）患者发生心力衰竭（HF）的影响因素，并构建风险预测模型。方法 纳入1 061例AMI患者，分为模型构建的训练集（786例）和模型验证的测试集（275例）。利用Lasso回归和多因素Logistic回归构建AMI患者发生HF的预测模型，并绘制列线图。采用受试者工作特征（ROC）曲线和校准曲线评价模型的区分度和校准度。结果 利用Lasso回归和多因素Logistic回归筛选出年龄、心率（HR）、ST段偏移、N端脑利钠肽前体（NT-proBNP）、同型半胱氨酸（Hcy）、纤维蛋白原（Fib）、左心室射血分数（LVEF）共7个变量建立模型。多因素Logistic回归构建预测模型的回归方程为Logit（P）=0.718×ST段偏移+0.042×年龄+0.037×HR+0.000 294×NT-proBNP+0.040×Hcy+0.220×Fib-5.617×LVEF-5.781。预测模型训练集的ROC曲线下面积（AUC）为0.846（95%CI：0.817~0.875），敏感度为78.50%，特异度为76.60%。校准曲线显示训练集患者HF的发生率与实际发生率基本相符。利用测试集对模型进行外部验证，AUC为0.848（95%CI：0.801~0.896），敏感度76.40%，特异度78.00%。结论 AMI患者发生HF与ST段偏移、年龄、入院HR、NT-proBNP、Hcy、Fib、LVEF有关，利用以上变量构建的预测模型具有较高的预测效能，有助于早期识别此类患者。

关键词: 心肌梗死, 心力衰竭, 危险因素, 预测, 列线图, 敏感性与特异性

Abstract:

Objective To analyze the factors affecting of heart failure (HF) in patients with acute myocardial infarction (AMI), and use the selected indicators to construct a risk prediction model. Methods A total of 1 061 AMI patients were included, and they were divided into the model-constructed training set (786 cases) and the test set (275 cases). Lasso regression and multiple Logistic regression were used to build a predictive model for HF occurrence in AMI patients, and a Nomogram diagram was drawn. Receiver operating characteristic (ROC) curve and calibration curve were used to evaluate the discrimination and calibration of the model. Results Lasso regression and multiple Logistic regression were used to select 7 variables to establish the model, including age, heart rate (HR), ST segment deviation, N-telencephalic natriuretic peptide precursor (NT-proBNP), homocysteine (Hcy), fibrinogen (Fib) and left ventricular ejection fraction (LVEF). The regression equation for constructing predictive model by multivariate Logistic regression was Logit (P) =0.718×ST segment deviation+0.042×age+0.037×HR+0.000 294×NT-proBNP+0.040×Hcy+0.220×Fib-5.617×LVEF-5.781. The area under ROC curve of the training set was 0.846 (95%CI: 0.817-0.875), the sensitivity was 78.50% and the specificity was 76.60%. The calibration curve showed that the incidence of HF in the training set was basically consistent with the actual incidence. The test set was used to verify the model externally. The area under ROC curve was 0.848 (95%CI: 0.801-0.896), the sensitivity was 76.40% and the specificity was 78.00%. Conclusion The occurrence of HF in AMI patients is related to ST segment deviation, age, HR, NT-proBNP, Hcy, Fib and LVEF. The predictive model based on the above variables has high predictive efficacy and is helpful for early identification of such patients.

Key words: myocardial infarction, heart failure, risk factors, predict, nomograms, sensitivity and specificity

中图分类号: