急性循环衰竭患者院内死亡风险预测模型的建立及评价  

Establishment and evaluation of in-hospital death risk prediction model for patients with acute circulatory failure

作　　者：阿米娜·阿不列孜 佧米莱·赛伊提江 张慧芳 里提甫江·买买提艾力 热依汉古丽·艾则孜 穆叶赛·尼加提[1] Amina Abuliezi;Kamilai Saiyitijiang;Zhang Huifang;Litifujiang Maimaitiaili;Reyihanguli Aizezi;Muyesai Nijiati(Xinjiang Uygur Autonomous Region People’s Hospital,Wulumuqi 830001,China)

机构地区：[1]新疆维吾尔自治区人民医院新疆急救中心,乌鲁木齐830001

出　　处：《中华急诊医学杂志》2024年第7期1019-1025,共7页Chinese Journal of Emergency Medicine

基　　金：国家自然科学基金地区科学基金项目(82060076)。

摘　　要：目的探讨急性循环衰竭(acute circulatory shock,ACF)患者院内死亡的危险因素,并进一步构建预后预测模型。方法回顾性分析2014年1月至2023年1月在新疆维吾尔自治区人民医院收治的224例符合急性循环衰竭中国急诊临床实践专家共识中符合休克诊断的患者临床资料。包括年龄、性别、入院诊断等基本信息,以及入院24 h内完成的血小板、乳酸、淋巴细胞计数、NK细胞计数、CD4、CD8等指标。并按照出院时的情况分为存活组及死亡组。将单因素分析中P<0.1的变量纳入最小绝对收缩选择算子(LASSO)回归模型,筛选出ACF患者院内死亡最重要的预测因子,通过Logistic回归构建预测模型。模型的区分度用受试者工作特征(receiver operator characteristic,ROC)曲线、曲线下面积(AUC)进行评价;采用Hosmer-Lemeshow检验评价预测模型校准度;最后用临床决策曲线分析(DCA)来检测模型的临床获益和应用价值。结果在224例ACF患者中,存活113例、死亡111例。单因素分析结果显示,年龄、神志、休克类型、呼吸频率、APACHE评分、淋巴细胞计数、乳酸、CD4、CD8、qsofa等指标在两组之间差异有统计学意义(P<0.05)。根据LASSO回归法筛选出的4个预测变量和结局变量构建Logistic回归预测模型,其中神志嗜睡、昏迷、呼吸频率和APACHE评分的增高为危险因素,CD4增加为保护因素。将上述指标用于构建ACF患者院内死亡预测列线图模型,预测院内死亡的概率截断值为0.4404,相对应的列线图总分约为136分。此模型AUC为0.830(0.764~0.895),敏感度为81.25%,特异度为68.83%。建模集Hosmer-Lemshow检验结果显示χ2=712和P=0.624,提示模型预测有较好的准确性。DCA分析评估表明该模型的净收益显著高于两个极端状况,具有较好的临床适用性。结论神志情况、呼吸频率、APACHE评分是急性循环衰竭患者院内死亡的危险因素,CD4是保护因素。据此构建的预测Objective To explore the risk factors of in hospital death in patients with acute circulatory failure,and to further construct the prediction model.Methods This study retrospectively analysed clinical data of 224 shock patients admitted to Xinjiang Uygur Autonomous Region People’s Hospital from January 2014 to January 2023,and patients were eligible for shock diagnosis according to the expert consensus of emergency clinical practice in China for acute circulatory failure.Including age,gender,admission diagnosis and other basic information,as well as platelet,lactic acid,lymphocyte count,NK cell count,CD4,CD8 and other indicators completed within 24 hours of admission.They were divided into survival group and death group according to the condition at discharge.Variables with P<0.1 in the univariate analysis were included in the LASSO regression model to screen out the most important predictors of hospital death in ACF patients,and the prediction model was constructed by Logistic regression.The model differentiation was evaluated by receiver operator characteristic(ROC)curve and area under the curve(AUC).Hosmer-Lemeshow test was used to evaluate the calibration degree of the prediction model.Finally,clinical decision curve analysis(DCA)was used to test the clinical benefit and application value of the model.Results A total of 224 ACF patients,113 survived and 111 died.The results of the univariate analysis showed statistically significant differences between the two groups in age,mental status,type of shock,respiratory rate,APACHE score,lymphocyte count,lactate,CD4,CD8 and qsofa(P<0.05).The Logistic regression prediction model was constructed according to the 4 predictors and outcome variables selected by LASSO method,in which increased delirium,coma,respiratory rate and APACHE score were risk factors and increased CD4 was a protective factor.The above indicators were used to construct a line graph model for predicting in-hospital death in ACF patients,with a probability cut-off value of 0.4404 for predicting in-

关 键 词：急性循环衰竭 休克 预后 淋巴细胞亚群 预测模型 

分 类 号：R459.7[医药卫生—急诊医学]