作者：王宇梅，徐格曼，丁俭，王猛猛，盛世英，刘猛

单位：

苏州大学附属第三医院神经内科

单位（英文）：

DepartmentofNeurology,theThirdAffiliatedHospitalofSoochowUniversity,Changzhou213003,China

关键词：

大脑梗死; 大脑半球大面积梗死; 预后; 影响因素分析; 列线图模型;

关键词（英文）：

Cerebralinfarction;Largehemisphereinfarction;Prognosis;Rootcauseanalysis;Nomogrammodel

中图分类号：

R743.33

DOI：

10.12114/j.issn.1008-5971.2023.00.039

基金项目：

常州市科技计划资助项目( CJ20190085?，CJ20220201?)

摘要：

目的 构建大脑半球大面积梗死（LHI）患者短期预后不良风险预测列线图模型并验证其预测效能。方法 回顾性收集2019年1月至2022年8月苏州大学附属第三医院神经内科收治的LHI患者264例为研究对象，根据发病30 d后改良Rankin量表（mRS）评分将患者分为预后良好组（mRS评分≤3分）和预后不良组（mRS评分4～6分）。收集患者的临床资料、实验室检查指标及影像学检查指标，采用多因素Logistic回归分析探讨LHI患者短期预后不良的影响因素，采用R 4.2.1软件包及rms程序包建立LHI患者短期预后不良风险预测列线图模型；采用Hosmer-Lemeshow拟合优度检验评价该列线图模型的拟合程度；采用Bootstrap法重复抽样1 000次，计算一致性指数；绘制校准曲线以评估该列线图模型预测LHI患者短期预后不良的效能；采用ROC曲线分析该列线图模型对LHI患者短期预后不良的预测价值。结果 264例LHI患者中预后良好47例（17.8%），预后不良217例（82.2%）。两组年龄、有心房颤动史者占比、有心功能不全史者占比，入院时美国国立卫生研究院卒中量表（NIHSS）评分、急性生理学与慢性健康状况评分系统Ⅱ（APACHEⅡ）评分、格拉斯哥昏迷量表（GCS）评分、TOAST分型、凝视发生率，住院期间意识障碍发生率、肺部感染发生率、需要机械通气者占比，白细胞计数、中性粒细胞计数、淋巴细胞计数、中性粒细胞与淋巴细胞比值（NLR）、C反应蛋白（CRP）、红细胞分布宽度（RDW）、D-二聚体、空腹血糖、肌酸激酶同工酶（CK-MB）、梗死体积、中线移位，侧脑室受压、基底池受压、其他区域受累发生率比较，差异有统计学意义（P<0.05）。多因素Logistic回归分析结果显示，入院时NIHSS评分、入院时APACHEⅡ评分、NLR、侧脑室受压是LHI患者短期预后不良的独立影响因素（P<0.05）。基于上述影响因素构建LHI患者短期预后不良风险预测列线图模型。Hosmer-Lemeshow拟合优度检验结果显示，该列线图模型拟合较好（χ2=4.867,P=0.772）；该列线图模型的一致性指数为0.910；校准曲线分析结果显示，该列线图模型预测LHI患者短期预后不良的发生率与实际发生率基本吻合；ROC曲线分析结果显示，该列线图模型预测LHI患者短期预后不良的AUC为0.915[95%CI(0.880,0.950)]，灵敏度为70.5%，特异度为97.9%，约登指数为0.684。结论 入院时NIHSS评分升高、入院时APACHEⅡ评分升高、NLR升高、侧脑室受压是LHI患者短期预后不良的独立危险因素，基于上述因素构建的列线图模型对LHI患者短期预后不良具有较高的预测价值。

英文摘要：

ObjectiveToconstructanomogrammodelforpredictingtheriskofpoorshort-termprognosisinlarge hemisphereinfarction(LHI)patientsandverifyitspredictiveefficiency.MethodsAtotalof264LHIpatientshospitalizedinDepartmentofNeurologyintheThirdAffiliatedHospitalofSoochowUniversityfromJanuary2019toAugust2022wereretrospectivelyselected.Thepatientsweredividedintogoodprognosisgroup[modifiedRankinScale(mRS)score≤ 3]andpoorprognosisgroup(mRSscore4-6)accordingtothemRSscoreat30daysafterstroke.Theclinicaldata,laboratoryindexesandimagingexaminationindexeswerecollected.ThemultivariateLogisticregressionanalysiswasusedtoanalyzetheinfluencingfactorsofpoorshort-termprognosisinLHIpatients.Thenomogrammodelforpredictingtheriskofpoorshort-termprognosisinLHIpatientswasconstructedbyR4.2.1softwarepackageandrmspackage.Hosmer-Lemeshowgoodnessoffittestwasusedtoevaluatethefittingdegreeofthenomogrammodel.Bootstrapmethodwasusedtorepeatedlysample1000times,andtheconsistencyindex(CI)wascalculated.Calibrationcurvewasdrawntoevaluatetheeffectivenessofthenomogrammodelfor predictingtheriskofpoorshort-termprognosisinLHIpatients,andtheROCcurvewasusedtoanalyzethepredictivevalueofthenomogrammodelforpoorshort-termprognosisinLHIpatients.ResultsAmong264LHIpatients,47(17.8%)hadgoodprognosisand217(82.2%)hadpoorprognosis.Thereweresignificantdifferencesinage,proportionofpatientswithatrialfibrillation,proportionofpatientswithcardiacinsufficiency,NationalInstitutesofHealthStrokeScale(NIHSS)score,AssessmentofAcutePhysiologyandChronicHealthStatusⅡ(APACHEⅡ)scoreatadmission,GlasgowComaScale(GCS)scoreatadmission,TOASTclassificationatadmission,incidenceofgazeatadmission,incidenceofconsciousnessdisorderandpulmonaryinfectionduringhospitalization,proportionofpatientsrequiringmechanicalventilationduringhospitalization,whitebloodcellcount,neutrophilcount,lymphocytecount,neutrophiltolymphocyteratio(NLR),C-reactiveprotein(CRP),redbloodcelldistributionwidth(RDW),D-dimer,fastingbloodglucose,creatinekinaseisoenzyme(CK-MB),infarctvolume,midlineshift,incidenceoflateralventricularcompression,incidenceofbasalcisterncompressionandincidenceofotherregionscompressionbetweenthetwogroups(P<0.05).MultivariateLogisticregressionanalysisshowedthatNIHSSscoreatadmission,APACHEⅡscoreatadmission,NLR,andlateralventricularcompressionweretheinfluencingfactorsofpoorshort-termprognosisinLHIpatients(P<0.05).Thenomogrammodelforpredictingpoorshort-termprognosisinLHIpatientswasconstructedbasedontheaboveinfluencingfactors.TheresultsofHosmer-Lemeshowgoodnessoffittestshowedthatthenomogrammodelfitswell(χ =42.867,P=0.772).TheCIofthenomogrammodelwas0.910.Theresultsofcalibrationcurveanalysisshowedthattheincidenceofpoorshort-termprognosisinLHIpatientspredictedbythenomogrammodelwasbasicallyconsistentwiththeactualincidenceofpoorshort-termprognosisinLHIpatients.TheresultsofROCcurveanalysisshowedthattheAUCofthenomogrammodelforpredictingthepoorshort-termprognosisinLHIpatientswas0.915[95%CI(0.880,0.950)],thespecificitywas70.5%,thesensitivitywas97.9%,andYoudenindexwas0.684.ConclusionTheincreasedNIHSSscoreatadmission,APACHEⅡscoreatadmissionandNLR,andlateralventricularcompressionaretheriskfactorsofpoorshort-termprognosisinLHIpatients.Thenomogrammodelconstructedbasedontheabovefactorshasahighpredictivevaluefortheriskofpoorshort-termprognosisinLHIpatients.

参考文献：