急性胰腺炎患者院内病死风险预测模型的建立与评价  

Establishment and evaluation of a predictive model for the risk of in-hospital mortality in patients with acute pancreatitis

作　　者：陈春露 杨晓寰[2] 吴晓佳 陈翠红 蔡敏华 许学强 蔡先彬[1] CHEN Chunlu;YANG Xiaohuan;WU Xiaojia;CHEN Cuihong;CAI Minhua;XU Xueqiang;CAI Xianbin(Department of Gastroenterology,The First Affiliated Hospital of Shantou University Medical College,Shantou 515041,China;Department of Traditional Chinese Medicine,The First Affiliated Hospital of Shantou University Medical College,Shantou 515041,China)

机构地区：[1]汕头大学医学院第一附属医院消化内科,广东汕头515041 [2]汕头大学医学院第一附属医院中医科,广东汕头515041

出　　处：《汕头大学医学院学报》2024年第3期129-135,共7页Journal of Shantou University Medical College

基　　金：汕头大学医学院第一附属医院英才计划基金(YCTJ-2023-07)。

摘　　要：目的:探讨重症监护室急性胰腺炎患者的院内病死的相关因素,构建列线图死亡风险预测模型。方法:从美国重症监护医学信息数据库IV中提取520例重症监护室的急性胰腺炎患者,按7∶3的比例随机分成训练集(n=364)和验证集(n=156)。训练集中男性210例(57.7%),女性154例(42.3%),年龄M (P25,P75)为57.9 (45.9,72.4)岁;验证集中男性87例(55.8%),女性69例(44.2%),年龄为54.6 (42.0,70.3)岁。利用logistic回归分析筛选急性胰腺炎患者院内病死的相关因素,构建列线图预测模型。通过受试者工作特征曲线下面积(area under curve,AUC)评估模型的区分度。采用Hosmer-Lemeshow检验及校准曲线评估模型的校准度。采用决策曲线分析法评估预测模型的临床适用性。结果:共筛选出7个预测因素,包括年龄(OR=1.057,95%CI:1.024~1.091)、心肌梗死(OR=4.152,95%CI:1.576~10.937)、严重肝病(OR=11.73,95%CI:3.291~42.190)、脓毒血症(OR=14.354,95%CI:4.612~44.675)、白细胞计数(OR=1.056,95%CI:1.006~1.108)、红细胞体积分布宽度(OR=1.258,95%CI:1.002~1.579)、活化部分凝血活酶时间(OR=1.031,95%CI:1.013~1.050)为急性胰腺炎患者院内病死的相关因素。基于以上7个预测因子构建列线图模型。受试者工作特征曲线显示该列线图模型预测训练集和验证集院内病死的AUC值分别为0.916和0.891,均大于简明急性生理学评分Ⅱ、急性生理学评分Ⅲ及牛津急性疾病严重程度评分的AUC值(分别为0.838和0.857,0.808和0.831,0.776和0.759)。表明该列线图模型的区分度较好,预测性能优于简明急性生理学评分Ⅱ、急性生理学评分Ⅲ和牛津急性疾病严重程度评分。校准曲线显示该列线图模型预测院内病死在训练集和验证集中都与实际结果具有较好的一致性。决策曲线表明,当阈值概率为20%~78%时,使用该列线图模型预测急性胰腺炎患者院内病死比简明急性生理学评分Ⅱ等能提供更大的净获益。结论:�Objective:To investigate the factors associated with in-hospital mortality in patients with acute pancreatitis in the intensive care unit(ICU),and to construct a nomogram prediction model for mortality risk.Methods:520 patients with acute pancreatitis admitted to the ICU were extracted from the U.S.Medical Information Mart for Intensive Care-IV database and randomized in a 7:3 ratio into a training set(n=364)and a validation set(n=156).There were 210(57.7%)males and 154(42.3%)females in the training set with an age of 57.9(45.9,72.4)years,and 87(55.8%)males and 69(44.2%)females in the validation set with an age of 54.6(42.0,70.3)years.Logistic regression analysis was utilized to screen for factors associated with in-hospital mortality in patients with acute pancreatitis,and a nomogram prediction model was constructed.The discrimination of the model was assessed by the area under curve(AUC)of receiver operating characteristic(ROC).The calibration of the model was assessed using the Hosmer-Lemeshow test and calibration curve.Decision curve analysis was used to assess the clinical applicability of the predictive model.Results:A total of 7 predictors were screened,including age(OR=1.057,95%CI:1.024-1.091),myocardial infarction(OR=4.152,95%CI:1.576-10.937),severe liver disease(OR=11.73,95%CI:3.291-42.190),sepsis(OR=14.354,95%CI:4.612-44.675),white blood cell count(OR=1.056,95%CI:1.006-1.108),red cell volume distribution width(OR=1.258,95%CI:1.002-1.579),and activated partial thromboplastin time(OR=1.031,95%CI:1.013-1.050)were factors associated with in-hospital mortality in patients with acute pancreatitis.A nomogram prediction model was constructed based on the above 7 predictors.The ROC curves showed that the AUC values of this model for predicting death in the training and validation sets were 0.916 and 0.891,respectively,which were greater than those of the simplified acute physiology scoreⅡ(SAPSⅡ),acute physiology scoreⅢ(APSⅢ),and the Oxford acute severity of illness score(OASIS)(0.838 and 0.857,0.808 and

关 键 词：急性胰腺炎 院内病死 美国重症监护医学信息数据库IV 预测模型 

分 类 号：R576[医药卫生—消化系统]