基于脑功能监测指标构建预测心搏骤停行心肺复苏后自主循环恢复患者神经功能预后不良的多因素Logistic回归模型

Construction of Prediction Model of Multivariate Logistic Regression for Poor Prognosis of Neurological Function in Patients with Return of Spontaneous Circulation after Cardiopulmonary Resuscitation in Cardiac Arrest Based on Brain Function Monitoring Indicators

背景心肺复苏(CPR)是抢救心搏骤停(CA)患者的重要方法,但多数患者抢救后可能出现神经功能预后不良,加强此类患者脑功能监测并寻找合适的神经功能预后判断方法对患者的康复具有重要意义。目的基于脑功能监测指标构建预测CA行CPR后自主循环恢复(ROSC)患者神经功能预后不良的多因素Logistic回归模型。方法选取惠州市第三人民医院2017年11月至2020年3月收治的CA行CPR后ROSC患者93例为研究对象。收集患者基线资料、脑血流参数、颈内静脉球血氧饱和度(SjvO2)及脑动脉-静脉血氧含量差(a-vDO2)、脑氧代谢率(CMRO2)。根据患者转出ICU时格拉斯哥-匹兹堡脑功能表现计分系统评分,将其分为预后良好组(1~2分,n=36)和预后不良组(3~4分,n=57)。采用多因素Logistic回归分析探讨CA行CPR后ROSC患者神经功能预后不良的影响因素,并构建多因素Logistic回归模型;绘制CBF、a-vDO2、CMRO2以及多因素Logistic回归模型预测CA行CPR后ROSC患者神经功能预后不良的受试者工作特征(ROC)曲线。结果预后不良组ROSC时间长于预后良好组,急性生理学与慢性健康状况评分系统Ⅱ(APACHEⅡ)评分高于预后良好组,ICU入住时间短于预后良好组,出ICU时格拉斯哥昏迷量表(GCS)评分低于预后良好组(P<0.05)。预后不良组左、右侧收缩期血流速度(Vs)、舒张期血流速度(Vd)、平均血流速度(Vm)、脑血流量(CBF)低于预后良好组,搏动指数(PI)、阻力指数(RI)高于预后良好组(P<0.05)。预后不良组SjvO_(2)高于预后良好组,a-vDO_(2)、CMRO_(2)低于预后良好组(P<0.05)。多因素Logistic回归分析结果显示,左侧CBF、a-vDO_(2)、CMRO_(2)是CA行CPR后ROSC患者神经功能预后不良的影响因素(P<0.05)。将左侧CBF、a-vDO_(2)、CMRO_(2)分别作为协变量X_(1)、X_(2)、X_(3),构建多因素Logistic回归模型,其表达式为:P=1/{1+Exp〔-(-8.735+0.553X1+0.062X2+0.117X3)〕}。CBF、a-vDO2、CMRO2以及多因素Logistic回归模型预测CA行CPR后ROSC患者神经功能预后不良的曲线下面积分别为0.664〔95%CI(0.448,0.887),P=0.035〕、0.603〔95%CI(0.395,0.818),P=0.047〕、0.712〔95%CI(0.513,0.918),P=0.013〕、0.856〔95%CI(0.713,0.985),P=0.002〕,最佳临界值分别为5.5 ml/min、28.1 ml/L、155.8μmol·100 g-1·min-1、0.267,灵敏度分别为73.68%、64.91%、78.95%、85.96%,特异度分别为72.22%、63.89%、83.33%、91.67%,正确率分别为73.12%、64.52%、80.65%、88.17%。结论本研究基于脑功能监测指标构建的多因素Logistic回归模型对CA行CPR后ROSC患者神经功能预后不良具有较高的预测价值,值得临床推广使用。

Background Cardiopulmonary resuscitation(CPR)is an important method to rescue patients with cardiac arrest(CA),but most patients may have poor neurological prognosis after rescue.Strengthening the monitoring of brain function of such patients and finding suitable methods for judging the prognosis of neurological function are of great significance to the rehabilitation of patients.Objective To construct a prediction model of multivariate Logistic regression for poor prognosis of neurological function in patients with return of spontaneous circulation(ROSC)after CPR in CA.Methods A total of 93 patients with ROSC after CPR in CA treated in Huizhou Third People's Hospital from November 2017 to March 2020 were selected as the research objects.The baseline data,cerebral blood flow parameters,jugular venous bulb oxygen saturation(SjvO2),arterio-venous oxygen content difference(a-vDO2)and cerebral oxygen metabolism rate(CMRO2)were collected.According to the score of Glasgow Pittsburgh brain function performance scoring system when patients were transferred out of ICU,they were divided into good prognosis group(1-2 points,n=36)and poor prognosis group(3-4 points,n=57).Multivariate Logistic regression analysis was used to explore the influencing factors of poor prognosis of neurological function in ROSC patients after CPR in CA,and a prediction model of multivariate Logistic regression was constructed.The receiver operating characteristic(ROC)curve of CBF,a-vDO2,CMRO2 and prediction model of multivariate Logistic regression predicting the poor prognosis of neurological function in ROSC patients after CPR in CA was drawn.Results The ROSC time in the poor prognosis group was longer than that in the good prognosis group,the score of acute physiology and chronic health evaluation scoring systemⅡ(APACHEⅡ)was higher than that in the good prognosis group,the ICU stay time was shorter than that in the good prognosis group,and the Glasgow Coma Scale(GCS)score when leaving ICU was lower than that in the good prognosis group(P<0.05).The left and right systolic velocity(Vs),diastolic velocity(Vd),mean velocity(Vm)and cerebral blood flow(CBF)in the poor prognosis group were lower than those in the good prognosis group,and the pulsitility index(PI)and resistence index(RI)were higher than those in the good prognosis group(P<0.05).SjvO2 in poor prognosis group was higher than that in good prognosis group,and a-vDO2 and CMRO2 were lower than those in good prognosis group(P<0.05).Multivariate Logistic regression analysis showed that left CBF,a-vDO2 and CMRO2 were the influencing factors of poor prognosis of neurological function in ROSC patients after CPR in CA(P<0.05).Left CBF,a-vDO2 and CMRO2 were taken as covariates X1,X2 and X3,respectively,and constructed the prediction model of multivariate Logistic regression.It's expression was P=1/{1+Exp[-(-18.735+0.553X1+0.062X2+0.117X3)]}.The areas under the curve of CBF,a-vDO2,CMRO2 and prediction model of multivariate Logistic regression for predicting the poor prognosis of neurological function in patients with ROSC after CPR in CA were 0.664[95%CI(0.448,0.887),P=0.035],0.603[95%CI(0.395,0.818),P=0.047],0.712[95%CI(0.513,0.918),P=0.013]and 0.856[95%CI(0.713,0.985),P=0.002],respectively.The optimum critical values were 5.5 ml/min,28.1 ml/L and 155.8μmol•100 g-1•min-1 and 0.267,respectively.The sensitivity was 73.68%,64.91%,78.95%and 85.96%,the specificity was 72.22%,63.89%,83.33%and 91.67%,and the accuracy was 73.12%,64.52%,80.65%and 88.17%,respectively.Conclusion In this study,the prediction model of multivariate Logistic regression constructed based on brain function monitoring indicators has a high predictive value for the poor prognosis of neurological function in ROSC patients after CPR in CA,which is worthy of clinical promotion.