神经重症患者术后中枢神经系统感染预测模型的构建及验证

Construction and verification of prediction model for postoperative central nervous system infection in patients with severe neurological diseases

目的构建列线图模型,以预测神经重症患者术后发生中枢神经系统感染(CNSI)的风险.方法回顾性分析2020年1至2022年1月武汉大学人民医院重症医学科收治的神经外科术后患者的临床资料,共1264例.将首义院区的987例患者作为训练集,将光谷院区的277例患者作为验证集.以术后30 d出现CNSI为研究终点.采用先单因素后多因素logistic回归分析法确定患者术后出现CNSI的危险因素,基于危险因素构建列线图模型.采用受试者工作特征(ROC)曲线、C指数及校准曲线评估列线图模型的预测准确性及判断能力,采用临床决策曲线分析(DCA)法评估模型的临床应用价值.在验证集对模型进行内部验证.结果1264例患者中,146例(11.6%)发生CNSI,其中训练集有102例(10.3%,102/987),验证集有44例(15.9%,44/277).多因素logistic回归分析显示,糖尿病史、急性生理与慢性健康状况(APACHE)Ⅱ评分≥10分、急诊手术、手术时间≥4 h、术中出血量≥400 ml、合并休克、术后行腰大池引流术和脑室外引流术、术后血清白蛋白≤30g/L及入住重症监护室时间≥3 d是神经重症患者术后发生CNSI的独立危险因素(均P<0.05).构建的列线图模型预测训练集及验证集患者术后发生CNSI的曲线下面积分别为0.79(95%CI:0.75~0.84)、0.72(95%CI:0.65~0.80).训练集及验证集校准曲线经Hosmer-Lemeshow拟合优度检验显示P值分别为0.267、0.179.临床DCA显示,训练集及验证集的阈值概率均在0~0.8,净获益率均>0,提示列线图预测模型的临床应用价值高.结论基于神经重症患者术后CNSI危险因素构建的列线图模型具有良好的预测能力,有助于识别高CNSI风险的患者,并进行早期干预,降低神经重症患者术后发生CNSI的风险.

Objective To establish a nomogram model for prediction of the risk of postoperative central nervous system infection(CNSI)in patients with severe neurological diseases.Methods The clinical medical records of I 264 patients who underwent neurosurgical treatment in the Department of Critical Medicine,Renmin Hospital of Wuhan University from January 2020 to January 2022 were retrospectively collected.In this series,987 patients admitted to the Department of Intensive Care at the Shouyi campus were used as the training set;277 patients admitted to the Department of Critical Care Medicine at the Guanggu campus were used as the validation set.Univariate and multivariate logistic regression analyses were used to identify the risk factors of postoperative CNSI in patients with severe neurological diseases and build a nomogram model.The prediction accuracy and judgment ability of the nomogram model were evaluated by the receiver operating characteristic curve,C index and calibration curve.The clinical application value of the nomogram model was evaluated by the decision curve analysis.The model was internally verified in the validation set.Results A total of 1 264 severe patients after neurosurgery were included in this study,of which 146 cases(11.6%)developed CNSI.A total of 1264 severe patients after neurosurgery were included in this study,of which 146 cases(11.6%)developed CNSI,including 102(10.3%,102/987)in the training set and 44(15.9%,44/277)in the validation set.Multivariate logistic regression analysis showed that diabetes history,APACHE Ⅱ score≥10,emergency operation,operation time≥4 h,postoperative lumbar cistern drainage,postoperative extraventricular drainage,intraoperative bleeding≥400 ml,combination with shock,postoperative albumin≤30 g/L and intensive care unit(ICU)stay≥3 days were independent risk factors for CNSI in patients with severe neurological diseases after operation.The area under the curve(AUC)of the training set was0.79(95%CI:0.75-0.84)and that of the validation set was 0.72(95%CI:0.65-0.80).The calibration curves between training set and verification set showed P values of 0.267 and 0.179 respectively by Hosmer-Lemeshow test.The analysis of clinical decision curve showed that the threshold probabilities of the training set and the validation set were 0-0.8,and the net benefit rate was>0,suggesting that the nomogram prediction model has high clinical application value.Conclusion The nomogram model constructed based on the risk factors of postoperative CNSI in neurocritically ill patients has good predictive ability,which is helpful to identify patients with high risk of CNSI,administer early intervention,and reduce the risk of postoperative CNSI in neurocritically ill patients.