自发性脑出血患者术后合并肺栓塞的预测模型的构建与验证

Construction and verification of a nomogram model for postoperative pulmonary embolism in patients with spontaneous cerebral hemorrhage

目的探讨自发性脑出血患者术后合并肺栓塞的危险因素,构建并验证列线图模型。方法本研究是回顾性队列研究,回顾性选取2015年1月至2023年1月于重庆医科大学附属第一医院及重庆大学附属三峡医院住院治疗的393例患者为训练组,165例患者为验证组,采用单因素和多因素逐步Logistic回归分析,筛选出与自发性脑出血术后合并肺栓塞相关的危险因素,构建并验证基于这些因素的列线图模型。结果基于年龄、出血量、格拉斯哥昏迷评分(Glasgow coma scale,GCS)、手术方式、纤维蛋白(原)降解产物(fibrinogen degradation product,FDP)、D-二聚体、血红蛋白和血浆、渗透压、有无深静脉血栓等危险因素绘制列线图模型,受试者工作特征曲线(receiver operating characteristic curve,ROC)显示模型对是否存在肺栓塞区分度良好,曲线下面积(area under curve,AUC)=0.908,验证集的Hosmer-Lemeshow拟合优度检验表明此模型对于验证集的拟合度良好(χ^(2)=14.805,df=8,P=0.063),校正曲线与理想曲线较为接近,模型的预测概率与实际发生概率接近,说明该模型准确性较高,决策曲线分析显示在较大范围的阈值概率下,建立的列线图模型均可获得收益。结论本研究建立的自发性脑出血患者术后合并肺栓塞的预测模型具有良好的预测性能,可在临床工作中准确、及时、快速识别肺栓塞的发生。

Objective To investigate the risk factors for postoperative pulmonary embolism in patients with spontaneous cerebral hemorrhage,and construct and verify the nomogram model.Methods A retrospective cohort study was conducted on 558 patients admitted in the First Affiliated Hospital of Chongqing Medical University and the Three Gorges Hospital Affiliated to Chongqing University.And 393 of them who hospitalized from January 2015 to January 2021 were assigned into a modeling group,and the other 165 patients from February 2021 to January 2023 into a validation group.Univariate and multivariate stepwise logistic regression analyses were used to screen out the risk factors associated with pulmonary embolism after spontaneous cerebral hemorrhage surgery.Then a nomogram model was build based on these factors and verified.Results Based on age,blood loss,Glasgow coma scale(GCS)score,surgical treatments,levels of fibrin degradation products,D-dimer and hemoglobin,plasma osmolality,and deep vein thrombosis,a risk model of pulmonary embolism was built.Receiver operating characteristic(ROC)curve analysis showed the model had good discriminability for the presence of pulmonary embolism,and the area under the curve(AUC)value was 0.908.Hosmer-Lemeshow goodness-fit test indicated that the model had a good fit to the verification set(Chi-square=14.805,df=8,P=0.063),the correction curve was close to the ideal curve,and the prediction probability of the model was close to the actual occurrence probability,suggesting the model having good accuracy.Decision curve analysis revealed that the established nomogram model can get benefits under a large range of threshold probabilities.Conclusion We develop a prediction model for postoperative pulmonary embolism in patients with spontaneous cerebral hemorrhage after surgical treatment,which shows good prediction performance in both the training and validation groups,and can be used for accurate,prompt and quick identification for the occurrence of pulmonary embolism in clinical practice.