基于序贯器官功能衰竭评分的LASSO-Logistic诊断模型在脓毒症中的效能分析  

Analysis of the efficacy of LASSO-Logistic diagnostic model based on sequential organ failure assessment score in sepsis

作　　者：张立琳 章金鹏 金律 王雷[2] 蔡榕松 杨亚东 ZHANG Lilin;ZHANG J inpeng;JIN Lü;WANG Lei;CAI Rongsong;YANG Y adong(Health Science Center,Yangtze University,Jingzhou 434023,Hubei,China;Huanggang Central Hospital Affiliated to Yangtze University,Huanggang 438000,Hubei,China)

机构地区：[1]长江大学医学部,湖北荆州434023 [2]长江大学附属黄冈市中心医院,湖北黄冈438000

出　　处：《武汉大学学报（医学版）》2024年第10期1207-1212,共6页Medical Journal of Wuhan University

基　　金：湖北省自然科学基金资助项目(编号:2021CFB530)。

摘　　要：目的:使用LASSO-Logistic分析影响脓毒症发生的相关因素,构建基于序贯器官功能衰竭评分(SOFA)的脓毒症诊断模型,并进行深度验证。方法:本研究回顾性收集了长江大学附属黄冈市中心医院重症医学科在2020年7月至2022年10月间收治的危重症患者在入科24 h内包括实验室检测指标在内的临床数据,采用SOFA评分评估患者器官功能障碍。依据患者入住ICU后的诊断分为脓毒症组和非脓毒症组,采用最小绝对收缩和选择算子(LASSO)回归模型初步筛选变量,使用多因素Logistic回归建立脓毒症诊断模型,计算曲线下面积(AUC),应用K折交叉验证深度评估模型效能。结果:共纳入患者235例,其中130例诊断为脓毒症,非脓毒症患者105例。LASSO-Logistic分析显示,抗凝血酶Ⅲ活性物质(ATⅢ)、纤维蛋白原(FIB)、超敏C反应蛋白(hs-CRP)和SOFA评分被评定为独立危险因子,将其构建联合模型,模型的AUC值为0.928(95%CI:0.896~0.960),优于各单项指标(P<0.05),100次10折交叉验证提示模型具有较好的泛化能力。结论:基于SOFA评分构建LASSO-Logistic诊断模型,对脓毒症诊断表现出较好预测效能和泛化能力,可提高对脓毒症患者的早期诊断。Objective:To analyze the relevant factors affecting the occurrence of sepsis using LASSO-Logistics,to construct the efficacy of a sequential organ failure assessment(SOFA)score-based diagnostic model for sepsis,and to conduct in-depth validation.Methods:This study retrospectively collected clinical data of critically ill patients,including laboratory test indicators,ad-mitted to the Department of Critical Care Medicine(CCM)of Huanggang Central Hospital affiliated to Yangtze University between July 2020 and October 2022 within 24 hours of admission,and used the SOFA score to assess patients'organ dysfunction.Patients were divided into sepsis and non-sepsis groups based on their diagnosis on admission to the intensive care unit,and variables were initially screened using a LASSO regression model,and a sepsis diagnostic model was developed using multi-factorial logistic regression to calculate the area under the curve(AUC).Results:A total of 235 pa-tients were included,and 130 were diagnosed with sepsis and 105 without.LASSO-logistic analysis showed that antithrombinⅢactive substance(ATⅢ),fibrinogen(FIB),hypersensitive C-reactive protein(hs-CRP),and SOFA were assessed as independent risk factors,and a joint diagnostic model was constructed.The joint model with an AUC value of 0.928(95%CI:0.896-0.960)was better than each indicator(P<0.05),with better generalization ability by the ten-fold cross-validated sugges-tive model.Conclusion:The LASSO-logistic diagnostic model was constructed based on the SOFA score,which showed better predictive efficacy and generalization ability for sepsis diagnosis and could improve the early diagnosis of septic patients.

关 键 词：脓毒症 SOFA评分 LASSO回归 预后分析 K折交叉验证 深度验证 

分 类 号：R459.7[医药卫生—急诊医学]