构建与验证神经外科术后中枢神经系统感染的预后模型  

Development and validation of prognostic nomogram for post-operative central nervous system infections after neurosurgery

作　　者：程利 王莉[2] 蔡强[3] Cheng Li;Wang Li;Cai Qiang(Department of Critical Care Medicine,Eastern Campus,Renmin Hospital of Wuhan University,Wuhan 430200,China;Jianghan University Affiliated Hospital,Wuhan 432200,China;Department of Neurosurgery,Eastern Campus,Renmin Hospital of Wuhan University,Wuhan 430200,China)

机构地区：[1]武汉大学人民医院东院重症医学科,武汉430200 [2]江汉大学附属医院,武汉432200 [3]武汉大学人民医院东院神经外科,武汉430200

出　　处：《中华实验外科杂志》2024年第11期2585-2589,共5页Chinese Journal of Experimental Surgery

基　　金：国家自然科学基金资助项目(81971158、81671306);武汉市卫生健康委科研重点项目(WX20A12)。

摘　　要：目的:构建与验证针对神经外科术后中枢神经系统感染(PCNSIs)患者影响预后的风险列线图模型。方法:收集2020年1月至2022年9月武汉大学人民医院和湘潭市中心医院收治的320例PCNSIs确诊患者的临床资料,其中男152例,女168例,年龄18~85岁(62.4±17.7)岁。Cox回归确定危险因素,内部验证采用自举法;外部验证经受试者工作特征(ROC)曲线、C指数及校准曲线评价模型的预测性能;决策曲线分析(DCA)评估模型的实用价值。计量资料的组间比较采用Mann-Whitney U检验,计数资料的组间比较采用χ^(2)检验。结果:共纳入PCNSIs确诊患者320例,43.4%(139/320)预后良好,56.6%(181/320)预后不良。Cox回归显示,≥3种基础疾病[风险比(HR)=1.507,95%置信区间(CI):1.018~2.232,P<0.05]、检出多重耐药菌(HR=1.521,95%CI:1.029~2.248,P<0.05)、合并脓毒症(HR=1.539,95%CI:1.046~2.263,P<0.05)、急性生理与慢性健康状况评分(APACHEⅡ)(HR=1.096,95%CI:1.043~1.152,P<0.05)及降钙素原(HR=1.010,95%CI:1.001~1.018,P<0.05)是影响预后的独立危险因素;格拉斯哥昏迷评分(GCS)(HR=0.873,95%CI:0.834~0.915,P<0.05)是唯一保护性因素。训练队列30 d和60 d模型的曲线下面积(AUC)分别为0.786(95%CI:0.656~0.931)和0.872(95%CI:0.723~0.968);验证队列30 d和60 d模型的AUC分别为0.709(95%CI:0.612~0.847)和0.776(95%CI:0.707~0.903)。Hosmer-Lemeshow检验结果显示,训练队列30 d(χ^(2)=5.062,P>0.05)低于60 d模型(χ^(2)=4.749,P>0.05);验证队列30 d(χ^(2)=8.117,P>0.05)低于60 d模型(χ^(2)=7.658,P>0.05)。DCA分析显示,两队列60 d模型的净获益明显高于30 d模型(χ^(2)=2.132,P<0.05)。结论:本研究构建的针对PCNSIs患者的预后模型经内部及外部验证后,该模型的预测准确性较高。Objective To construct and validate a risk nomogram for predicting the prognosis of patients with post-operative central nervous system infections(PCNSIs).Methods The clinical data of 320 patients with PCNSIs in Renmin Hospital of Wuhan University and Xiangtan Central Hospital from January 2020 to September 2022 were collected,including 152 males and 168 females aged 18 to 85 years(62.4±17.7)years.Cox regression to identify risk factors.Internal validation was conducted using the bootstrap method,while external validation evaluated the predictive performance of the model through receiver operating characteristic(ROC)curve,C-index,and calibration curve.Decision curve analysis(DCA)was performed to assess the utility of the model.Mann Whitney U test was used for inter group comparison of quantitative data;the comparison of count data between groups was conducted using chi square test.ResultsA total of 320 patients with confirmed PCNSIs were included,among which 43.4%(139/320)had good prognosis and 56.6%(181/320)had poor prognosis.The results of Cox analysis showed that≥3 basic disease histories[risk ratio(HR)=1.507,95%confidence interval(CI):1.018-2.232,P<0.05],detection of multidrug-resistant bacteria(HR=1.521,95%CI:1.029-2.248,P<0.05),sepsis(HR=1.539,95%CI:1.046-2.263,P<0.05),acute physiology and chronic health evaluationⅡ(APACHEⅡ)score(HR=1.096,95%CI:1.043-1.152,P<0.05)and procalcitonin(HR=1.010,95%CI:1.00-1.018,P<0.05)were independent risk factors affecting prognosis,while the glasgow coma score(GCS)(HR=0.873,95%CI:0.834-0.915,P<0.05)was the only protective factor.The area under the curve(AUC)of the 30 d and 60 d models of the training cohort were 0.786(95%CI:0.656-0.931)and 0.872(95%CI:0.723-0.968),while the AUC of the 30 d and 60 d models of the validation cohort were 0.709(95%CI:0.612-0.847)and 0.776(95%CI:0.707-0.903)respectively.The Hosmer-Lemeshow test results showed that the training cohort for 30 d model(χ^(2)=5.062,P>0.05)was lower than the 60 d model(χ^(2)=4.749,P>0.05),as well as for the 30 d

关 键 词：神经外科术后中枢神经系统感染 列线图 预后 预测 

分 类 号：R651[医药卫生—外科学]