基于智能决策系统的脓毒症预测模型构建及效能研究  

Study on construction and effectiveness of sepsis prediction model based on intelligent decision system

作　　者：聂炳荣 黄河清 胡慧婷 贺艳 Nie Bingrong;Huang Heqing;Hu Huiting;He Yan(Department of Emergency Medicine,Zhuhai People's Hospital(Zhuhai Clinical Medical College Affiliated to Jinan University,Zhuhai Hospital Affiliated to Jinan University),Zhuhai,Guangdong 519000,China.)

机构地区：[1]珠海市人民医院(暨南大学珠海临床医学院,暨南大学附属珠海医院)急诊医学部,广东珠海519000

出　　处：《齐齐哈尔医学院学报》2024年第24期2333-2338,共6页Journal of Qiqihar Medical University

基　　金：2023年度珠海市医学科研项目(2220009000270)。

摘　　要：目的探讨基于智能决策系统的脓毒症预测模型构建及其临床应用效能。方法选择2022年9月—2023年11月本院急诊科接诊的脓毒症高危患者100例为研究对象,所有患者入院后均接受血常规检查,随访掌握脓毒症发生情况。将研究对象按约7︰3划为建模组和验证组,对临床相关资料行单/多因素分析,并以差异性参数构建脓毒症预警模型,并通过R软件筛选出脓毒症发生的重要性变量,构建基于AI的预测模型,计算OR值并构建诺模图,通过绘制ROC曲线及计算ROC下面积(AUC)对预测模型效能予以验证。结果本研究100例患者中,临床诊治中出现脓毒症的14例,发生率14%。通过单因素、多因素Logistic逐步回归分析表明,入院时合并感染、入院第一次血常规指标WBC、PCT、hs-CRP、TNF-α是脓毒症发生的独立危险因素,通过AI算法建立预测模型。通过分析,AUC为0.778,敏感度和特异性分别为0.708(95%CI:0.651~0.760)和0.864(95%CI:0.823~0.899)。结论SOFA、入院时合并感染是脓毒症发生的独立危险因素,据此建立基于AI决策系统的预测模型,有助于临床医师早期发现和识别潜在的脓毒症患者。Objective To explore the construction of a sepsis prediction model based on an intelligent decision system and its clinical application effectiveness.Methods High-risk patients with sepsis admitted to the emergency department of our hospital from September 2022 to November 2023 were selected as the study subjects.All patients underwent blood routine tests after admission and were followed up to track the occurrence of sepsis.They were divided into a modeling group and a validation group in a ratio of approximately 7︰3.Single/multiple factor analyses were conducted on clinically relevant data,and a sepsis warning model was constructed using differential parameters.R software was used to identify important variables for sepsis occurrence,and an AI-based prediction model was built,calculating OR values and constructing a nomogram.The effectiveness of the prediction model was verified by plotting ROC curves and calculating the area under the ROC(AUC).Results Among the 100 patients in this study,sepsis occurred in 14 patients during clinical diagnosis and treatment,with an incidence rate of 14%.Univariate and multivariate logistic stepwise regression analysis indicated that infection at admission and initial blood routine indicators such as WBC,PCT,hs-CRP,and TNF-αwere independent risk factors for sepsis occurrence,and a prediction model was established through an AI algorithm.The AUC was 0.778,and the sensitivity and specificity were 0.708(95%CI:0.651~0.760)and 0.864(95%CI:0.823~0.899),respectively.Conclusions SOFA scores and infection at admission are independent risk factors for sepsis.The established prediction model based on an AI decision system aids clinical physicians in early detection and identification of potential sepsis patients.

关 键 词：智能决策系统 脓毒症 相关危险因素 预测模型 效能 

分 类 号：R459.7[医药卫生—急诊医学]