基于机器学习的脓毒症谵妄患者死亡预测模型的构建与评估

doi:10.20223/j.cnki.1000-8535.2025.11.005

广州医药 ›› 2025, Vol. 56 ›› Issue (11): 1501-1510.DOI: 10.20223/j.cnki.1000-8535.2025.11.005

基于机器学习的脓毒症谵妄患者死亡预测模型的构建与评估

黄毅文¹, 胡北²

1 汕头大学医学院（广东汕头 515041）
2 南方医科大学附属广东省人民医院（广东省医学科学院）急诊科（广东广州 510080）

收稿日期:2025-01-10 出版日期:2025-11-20 发布日期:2025-12-25
通讯作者: 胡北,E-mail：wymanhuang@qq.com

Machine learning prediction model for sepsis-associated delirium mortality

HUANG Yiwen¹, HU Bei²

1 Medical College,Shantou University,Shantou 515041,China
2 Emergency Department,Guangdong Provincial People’s Hospital（Guangdong Academy of Medical Sciences）,Southern Medical University,Guangzhou 510080,China

Received:2025-01-10 Online:2025-11-20 Published:2025-12-25

摘要/Abstract

摘要： 目的通过机器学习方法构建脓毒症谵妄患者30 d死亡的预测模型,并识别关键预测因子。方法采用基于医疗信息集成重症监护数据库（Medical Information Mart for Intensive Care IV）的回顾性队列研究方法,boruta筛选重要特征,并通过决策树,K近邻,LightGBM,随机森林,支持向量机,XGBoost构建模型进行分析,通过ROC曲线下面积进行评估,利用F1分数、召回率、精确率、特异度、灵敏度和阳性预测值比较模型表现。结果 XGBoost模型在训练集和验证集中的ROC曲线下面积分别为0.906和0.762,表明该模型具有良好的预测能力,入院年龄、红细胞分布宽度和白细胞计数是最重要的预测因子。结论基于机器学习的脓毒症谵妄患者预后预测模型展现出良好的预测效能,为临床早期干预提供了重要参考依据。

关键词: 脓毒症, 谵妄, 机器学习, 死亡

Abstract: Objective To construct a 30-day mortality prediction model for patients with sepsis-associated delirium using machine learning methods and identify key predictive factors．Methods A retrospective cohort study was conducted based on the Medical Information Mart for Intensive Care IV database．Important features were selected using the Boruta algorithm,and models including Decision Tree,K-Nearest Neighbors,LightGBM,Random Forest,Support Vector Machine,and XGBoost were constructed and analyzed．Model performance was evaluated using the area under the reciver operater characteristic（ROC） curve（AUC）,along with F1 score,recall,precision,specificity,sensitivity,and positive predictive value．Results The XGBoost model demonstrated strong predictive performance,with AUC values of 0.906 in the training set and 0.762 in the test set．Key predictors identified included admission age,red blood cell distribution width,and white blood cell count．Conclusions The machine learning-based prediction model for sepsis-associated delirium prognosis exhibits robust predictive efficacy,providing a valuable tool for early clinical intervention．

Key words: sepsis, delirium, machine learning, mortality

黄毅文, 胡北. 基于机器学习的脓毒症谵妄患者死亡预测模型的构建与评估[J]. 广州医药, 2025, 56(11): 1501-1510.

HUANG Yiwen, HU Bei. Machine learning prediction model for sepsis-associated delirium mortality[J]. Guangzhou Medical Journal, 2025, 56(11): 1501-1510.

参考文献

[1] SINGER M,DEUTSCHMAN C S,SEYMOUR C W,et al.The third international consensus definitions for sepsis and septic shock(Sepsis-3)[J].JAMA,2016,315(8):801-810.
[2] 杨思敏,卞兴航,邓开基,等.脓毒症患者谵妄发生风险预测模型构建[J].中华急诊医学杂志,2024,33(6):784-791.
[3] GIRARD T D,JACKSON J C,PANDHARIPANDE P P,et al.Delirium as a predictor of long-term cognitive impairment in survivors of critical illness[J].Crit Care Med,2010,38(7):1513-1520.
[4] 柏晓奇,顾琼,许俊,等.脓毒症相关谵妄患者预后预测模型构建及验证[J].中华急诊医学杂志,2024,33(5):651-657.
[5] JOHNSON A E W,BULGARELLI L,POLLARD T J,et al.MIMIC-IV,a freely accessible electronic health record dataset[J].Sci Data,2021,8(1):1-18.
[6] 胡畅,胡波,李志峰,等.四种评分系统对脓毒症患者ICU死亡风险的预测价值比较[J].南方医科大学学报,2020,40(4):513-518.
[7] 梁霞. 缺失数据的多重插补及其改进[D].长沙:中南大学,2007.
[8] GORDON A D,BREIMAN L,FRIEDMAN J H,et al.Classification and regression trees[J].Biometrics,1984,40(3):874.
[9] COVER T,HART P.Nearest neighbor pattern classification[J].IEEE Trans Inf Theory,1967,13(1):21-27.
[10] ZHANG J,MUCS D,NORINDER U,et al.LightGBM:An effective and scalable algorithm for prediction of chemical toxicity—Application to the Tox21 and Mutagenicity data sets[J].J Chem Inf Model,2019,59(10):4150-4158.
[11] BREIMAN L.Random forests[J].Mach Learn,2024(45):5-32.
[12] CORTES C,VAPNIK V.Support-vector networks[J].Mach Learn,1995,20(3):273-297.
[13] LI B,EISENBERG N,BEATON D,et al.Using machine learning(XGBoost)to predict outcomes after infrainguinal bypass for peripheral artery disease[J].Ann Surg,2024,279(4):705-713.
[14] KURSA M B,RUDNICKI W R.Feature selection with the Boruta package[J].J Stat Soft,2010,36(11):1-13.
[15] HASTIE T,TIBSHIRANI R,FRIEDMAN J.The elements of statistical learning:Data mining,inference,and prediction[M].2nd ed.New York:Springer,2009.
[16] FAWCETT T.An introduction to ROC analysis[J].Pattern Recognit Lett,2006,27(8):861-874.
[17] LUNDBERG S M,ERION G,CHEN H,et al.From local explanations to global understanding with explainable AI for trees[J].Nat Mach Intell,2020,2(1):56-67.
[18] WICKHAM H,GROLEMUND G.R for Data Science:Import,Tidy,Transform,Visualize,and Model Data[M].Sebastopol:O’Reilly Media,2017.
[19] HOTCHKISS R S,KARL I E.The pathophysiology and treatment of sepsis[J].N Engl J Med,2003,348(2):138-150.
[20] MIKKELSEN M E,MILTIADES A N,GAIESKI D F,et al.Serum lactate is associated with mortality in severe sepsis independent of organ failure and shock[J].Crit Care Med,37(5):1670-1677.
[21] MARTIN G S,MANNINO D M,MOSS M.The effect of age on the development and outcome of adult sepsis[J].Crit Care Med,2006,34(1):15-21.
[22] ZHANG L,LI X,HUANG J,et al.Predictive model of risk factors for 28-day mortality in patients with sepsis or sepsis-associated delirium based on the MIMIC-IV database[J].Sci Rep,2024,14(1):18751.
[23] STEYERBERG E W,VICKERS A J,COOK N R,et al.Assessing the performance of prediction models:A framework for traditional and novel measures[J].Epidemiology,2010,21(1):128-138.
[24] COLLINS F S,VARMUS H.A new initiative on precision medicine[J].N Engl J Med,2015,372(9):793-795.

基于机器学习的脓毒症谵妄患者死亡预测模型的构建与评估

Machine learning prediction model for sepsis-associated delirium mortality

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	陈志江, 曾成, 赵纬, 韦秋菊, 单文琪, 王惠丽. 抗增殖蛋白2对脓毒症心肌损伤线粒体功能稳态的作用机制研究[J]. 广州医药, 2025, 56(9): 1201-1207.
[2]	张茅平, 陈国东. 构建基于MIMIC-IV数据库的主动脉夹层B型患者急性期死亡风险列线图预测模型：一项回顾性分析[J]. 广州医药, 2025, 56(8): 1134-1144.
[3]	唐启群, 于杰, 张涵飞, 胡惠菊, 王宇昕. 分析护理专业学生死亡态度的潜在类别及其影响因素[J]. 广州医药, 2025, 56(6): 836-842.
[4]	张千, 王菲, 孟纯雪, 张生富, 赵海龙, 唐旭红, 陈紫晔, 郭斌. 超声引导结合机器学习技术的智能针灸精准诊疗系统研究[J]. 广州医药, 2025, 56(5): 599-604.
[5]	陈冠君, 李建厂, 孙维梅, 周明轩, 尼坤洋, 王浩. 重楼皂苷Ⅰ通过p53信号通路诱导慢性粒细胞白血病细胞铁死亡[J]. 广州医药, 2025, 56(5): 605-615.
[6]	卢晓霞, 王佩, 王静喆. 免疫及靶向药物联合肝动脉灌注化疗治疗晚期肝癌的临床分析[J]. 广州医药, 2025, 56(5): 662-668.
[7]	杨佳林, 谢晓英. 双硫死亡与其他细胞死亡在卵巢癌治疗中的研究进展[J]. 广州医药, 2025, 56(4): 452-456.
[8]	黄静, 谢琳, 管婷, 魏莎, 李永滨. 老年营养风险指数与慢性阻塞性肺疾病患者急性加重期预后的相关性分析[J]. 广州医药, 2025, 56(2): 192-196.
[9]	董基毅, 陆金帅. 脓毒症患者发生心肌损伤的高危因素分析[J]. 广州医药, 2025, 56(10): 1338-1345.
[10]	江竹月, 孙洲华, 张庆庆, 张成欢, 陈雯. 机器学习算法在术后谵妄风险评估中的应用进展[J]. 广州医药, 2025, 56(1): 42-47.
[11]	郭秀林, 林亚芬, 张琴华. 俯卧位通气在脓毒症合并急性肺损伤患儿的应用效果[J]. 广州医药, 2024, 55(7): 796-800.
[12]	芮苗. 高龄髋部骨折患者术后谵妄与术前痛阈水平的相关性及处理措施[J]. 广州医药, 2024, 55(6): 658-662.
[13]	郑桂花, 周志聪, 林桂永, 张晓红, 银孟卓. 养老机构老年人死亡态度及其影响因素研究[J]. 广州医药, 2024, 55(3): 274-278.
[14]	刘育伟, 刘文刚, 高嘉骅. 穴位刺激预防与治疗老年人骨折术后谵妄的临床研究进展[J]. 广州医药, 2023, 54(7): 110-116.
[15]	唐苏予, 陈光辉, 王真珍. S-ChE联合T细胞表面PD-1水平对脓毒症患者病情预后状况评估的价值[J]. 广州医药, 2023, 54(2): 83-87.