开发MIMIC-Ⅲ脓毒性休克患者死亡风险预测模型的队列研究

A cohort study of developing a death risk prediction model in patients with septic shock based on MIMIC-Ⅲdatabase

目的应用LASSO-Logistic回归法建立脓毒性休克死亡风险预测模型,并进行验证。方法采用回顾性队列研究方法,基于开源的美国重症监护医学信息数据库Ⅲv1.4(MIMIC-Ⅲv1.4),纳入符合脓毒症3.0标准的脓毒性休克患者,提取人群特征、主要体征和实验室指标、住院情况及结局指标等数据。采用LASSO回归法筛选预测变量,使用Logistic回归法构建脓毒性休克死亡风险预测模型。采用Hosmer-Lemeshow检验评价预测模型校准度,采用受试者工作特征曲线(ROC曲线)评价模型区分度。结果共纳入693例脓毒性休克患者,30 d存活445例,死亡248例,30 d病死率为35.8%。根据LASSO回归法筛选出的9个预测变量和结局变量构建Logistic回归模型,结果显示,年龄、Elixhauser共病指数、血乳酸(Lac)、K^(+)升高和使用机械通气与30 d病死率增加相关〔优势比(OR)及95%可信区间(95%CI)分别为1.023(1.010~1.037)、1.047(1.022~1.074)、1.213(1.133~1.305)、2.241(1.664~3.057)、2.165(1.433~3.301),均P<0.01〕,收缩压(SBP)、舒张压(DBP)、体温、脉搏血氧饱和度(SpO2)下降也与30 d病死率增加相关〔OR(95%CI)分别为0.974(0.957~0.990)、0.972(0.950~0.994)、0.693(0.556~0.857)、0.971(0.949~0.992),均P<0.05〕。校准曲线显示,脓毒性休克死亡风险预测模型的预测风险与实际情况有较好的一致性;ROC曲线分析显示,预测模型的ROC曲线下面积(AUC)为0.839(95%CI为0.803~0.876),具有可较好地区分死亡与非死亡风险患者的能力。结论脓毒性休克死亡风险预测模型对脓毒性休克患者30 d死亡风险有较好的识别能力,包含9个医院容易获得的变量(年龄、Elixhauser共病指数、机械通气、Lac、K^(+)、SBP、DBP、体温、SpO2),可被临床医生用来计算脓毒性休克患者个体死亡风险。

Objective To develop and validate a model for predicting death risk in septic shock patients using LASSO-Logistic methods.Methods A retrospective cohort study was conducted.Based on the open-source database Medical Information Mart for Intensive Care-Ⅲv1.4(MIMIC-Ⅲv1.4),the septic shock patients meeting the Sepsis-3 criteria were included,and the data on demographic characteristics,major signs,laboratory examinations,hospitalization,and outcomes were extracted.Predictive variables were selected by LASSO regression and predictive models were derived using Logistic regression.The calibration of the model was evaluated using the Hosmer-Lemeshow test and discrimination was evaluated using the receiver operator characteristic curve(ROC curve).Results A total of 693 patients with septic shock were enrolled,in which 445 patients survived and 248 patients dead within 30 days and the mortality was 35.8%.Logistic regression model was constructed according to nine predictive variables and outcome variables screened by LASSO regression method,which showed that advanced age,Elixhauser index,blood lactic acid(Lac),K^(+)level and mechanical ventilation were associated with increased 30-day mortality[odds ratio(OR)and 95%confidence interval(95%CI)was 1.023(1.010-1.037),1.047(1.022-1.074),1.213(1.133-1.305),2.241(1.664-3.057),2.165(1.433-3.301),respectively,all P<0.01],and reduced systolic blood pressure(SBP),diastolic blood pressure(DBP),body temperature,and pulse oxygen saturation(SpO2)were also associated with increased 30-day mortality[OR(95%CI)was 0.974(0.957-0.990),0.972(0.950-0.994),0.693(0.556-0.857),0.971(0.949-0.992),respectively,all P<0.05].The calibration curve showed that the predicted risk of septic shock death risk prediction model had good agreement with the real situation.ROC curve analysis showed that the area under the ROC curve(AUC)of the prediction model was 0.839(95%CI was 0.803-0.876),which could distinguish patients at risk of death from those at risk of survival.Conclusions The septic shock death risk prediction model has a good ability to identify the 30-day mortality risk of septic shock patients,including nine hospital readily variables(age,Elixhauser index,mechanical ventilation,Lac,K^(+),SBP,DBP,body temperature and SpO2).The model could be used by clinicians to calculate the risk of death in septic shock individuals.