急性心肌梗死患者预后不良的影响因素分析及其风险预测列线图模型构建

Analysis of influential factors for poor prognosis in patients with acute myocardial infarction and construction of a risk prediction nomograph model

目的探究急性心肌梗死(AMI)患者预后不良的影响因素,并构建其风险预测列线图模型。方法研究对象为湖州市第一人民医院2018年6月至2021年6月接诊的AMI患者173例,根据发生心肌梗死6个月后随访结果将其分为预后良好组(n=130)和预后不良组(n=43)。采用回顾性分析研究方法,比较两组患者临床资料的差异;通过LASSO回归分析初步筛选潜在影响因素;通过logistic回归分析方法探究AMI患者预后不良的影响因素;列线图模型运用R 4.2.6语言“rms”包构建,并通过绘制受试者工作特征曲线(ROC曲线)、校准曲线、决策曲线评价模型的区分度、校准度及有效性,模型验证采用Bootstrap法进行内部验证(重复抽样1000次)。结果两组患者罪犯血管、Killip分级、血管开通时间、肌钙蛋白(cTnI)、高血压史、N末端B型脑钠肽前体(NT-proBNP)、糖尿病史、肌酐、高脂血症史、左室射血分数(LVEF)、吸烟史、肌酸激酶同工酶(CK-MB)比较,差异均有统计学意义(均P<0.05)。采用LASSO回归模型筛选出7个潜在的影响因素,分别为糖尿病史、梗死血管-前降支、KillipⅣ级、血管开通时间、cTnI、NT-proBNP、LVEF。logistic回归分析显示,血管开通时间(OR=0.171,95%CI:0.053~0.548,P=0.003)、cTnI(OR=0.201,95%CI:0.079~0.510,P=0.001)、LVEF(OR=1.469,95%CI:1.167~1.847,P=0.001)、NT-proBNP(OR=0.996,95%CI:0.993~1.00,P=0.025)是AMI患者预后不良的独立影响因素(均P<0.05),线性回归分析提示模型无明显共线性(VIF<10)。基于logistic回归分析提示的4个影响因素构建AMI患者预后不良风险预测的列线图模型,ROC曲线显示该模型的曲线下面积为0.979[95%CI(0.959,0.999)],一致性指数为0.934;模型的校准曲线与理想曲线接近;决策曲线分析提示,当该模型预测的概率阈值为0.61~0.99时,模型预测价值较优越。结论AMI患者预后不良的影响因素有血管开通时间、cTnI、NT-proBNP、LVEF等,构建的列线图模型对预测AMI患者预后不良具有较好的效能,可以为临床医生及早发现识别预后不良的AMI患者提供一定的参考。

Objective To investigate the influential factors for poor prognosis in patients with acute myocardial infarction and construct a risk prediction nomograph model.Methods A total of 173 patients with acute myocardial infarction who received treatment in The First People's Hospital of Huzhou from June 2018 to June 2021 were included in this study.They were divided into a good prognosis group(n=130)and a poor prognosis group(n=43)according to the follow-up results at 6 months after developing acute myocardial infarction.The clinical data of the two groups were compared using retrospective analysis methods.The potential influential factors were preliminarily screened using LASSO regression analysis.The influential factors of poor prognosis for acute myocardial infarction were investigated using logistic regression analysis.The risk prediction nomograph model was constructed using the"rms"package of R 4.2.6 language.The discriminability,calibration,and effectiveness of the model were evaluated by drawing the receiver operating characteristic curve,calibration curve,and decision curve.Model validation was conducted internally using the Bootstrap method(repeated sampling 1000 times).Results There were significant differences in the culprit vessel,Killip classification,vessel opening time,cardiac troponin I(cTnI),hypertension history,N-terminal pro-brain natriuretic peptide(NT-proBNP),diabetes history,creatinine,hyperlipidemia history,left ventricular ejection fraction,smoking history and creatine kinase isoenzymes-MB between the two groups(all P<0.05).Seven potential influential factors were screened using LASSO regression model,including diabetes history,infarcted vessel anterior descending branch,Killip IV,vascular opening time,cTnI,NT-proBNP,and left ventricular ejection fraction.Logistic regression analysis showed that vascular opening time(OR=0.171,95%CI:0.053-0.548,P=0.003),cTnI(OR=0.201,95%CI:0.079-0.510,P=0.001),left ventricular ejection fraction(OR=1.469,95%CI:1.167-1.847,P=0.001),NT-proBNP(OR=0.996,95%CI:0.993-1.00,P=0.025)were independent influential factors of poor prognosis in patients with acute myocardial infarction(all P<0.05).Linear regression analysis results indicate that the regression model did not exhibit significant multicollinearity(variance inflation factor<10).Based on the four influential factors identified by logistic regression analysis,a nomogram model for predicting the poor prognosis of patients with acute myocardial infarction was developed.The area under the receiver operating characteristic curve was 0.979[95%CI(0.959,0.999)],and the consistency index was 0.934.The calibration curve of the model was close to the ideal curve.Decision curve analysis revealed that when the probability threshold predicted by the model ranged from 0.61 to 0.99,the predictive value of the model was superior.Conclusion Factors influencing the poor prognosis of acute myocardial infarction include the time of vessel opening,cTnI,NT-proBNP,and left ventricular ejection fraction.The constructed nomogram model demonstrates good efficacy in predicting the poor prognosis of patients with acute myocardial infarction and can provide some reference for clinical doctors and nurses to identify patients with poor prognosis as soon as possible.