中国实用内科杂志2024，Vol.44Issue(12)：1026-1031,1038,7.DOI:10.19538/j.nk2024120111
基于机器学习构建成人脓毒性休克患者早期死亡风险模型NSOFA的研究
Study on new sequential organ failure score model of early death risk in adult patients with septic shock based on machine learning
摘要
Abstract
Objective To investigate the risk factors of early death in adult patients with septic shock based on machine learning,build a risk prediction model and verify it.Methods The clinical data of adult patients with septic shock admitted to the critical care unit of Zhongshan Hospital(Xiamen)of Fudan University from September 1,2020 to August 31,2023 were collected.The patients were divided into survival group and death group according to their 28-day clinical outcomes.The cases of rom September 1,2020 to August 31,2022 were used to construct a training set,and the cases from September 1,2022 to August 31,2023 were used to construct a validation set.Multivariate binary logistic regression was used to analyze the independent risk factors affecting the prognosis of patients,and the prediction model of death risk was constructed.Receiver operating characteristic(ROC)curve analysis was used to verify and evaluate the predictive ability of the model.Results A total of 101 adult patients with septic shock were included in the study(75 patients in the survival group and 26 patients in the death group).Multivariate binary logistic analysis showed that Sequential Organ Failure Score(SOFA)and Neutrophil-to-Lymphocyte Ratio(NLR)were independent risk factors for 28-day death in adult patients with septic shock(P＜0.05),and a new sequential organ failure score(NSOFA)was constructed.Internal and external verification of NSOFA scores showed that the area under ROC curve(AUC)was 0.94 and 0.90,and the overall model quality was 0.88 and 0.79,respectively,indicating good model quality(P＜0.01).The receiver operating characteristic curves of NSOFA,NLR,SOFA score and APACHE Ⅱ score were drawn,and the pair-to-sample regional differences were analyzed and compared.The AUC of NSOFA was the largest,which was 0.93,the Jordon index was 0.77,the cut-off value was 16.21,the sensitivity was 84.6％,and the specificity was 92.0％,better than other scoring systems(P＜0.05).Conclusion The SOFA score and NLR are independent risk factors for 28-day death in adult patients with septic shock,and the combined NSOFA score has higher predictive efficiency and has certain clinical application value.关键词
序贯器官衰竭评分/中性粒细胞与淋巴细胞比值/急性生理与慢性健康评分Ⅱ/脓毒性休克/死亡Key words
sequential organ failure score/neutrophil-to-lymphocyte ratio/acute physiology and chronic health evaluation Ⅱ/septic shock/death分类
医药卫生引用本文复制引用
林小明,马杰飞,洪翔宇,黄挺,张连芳..基于机器学习构建成人脓毒性休克患者早期死亡风险模型NSOFA的研究[J].中国实用内科杂志,2024,44(12):1026-1031,1038,7.基金项目
国家自然科学基金(82072131) （82072131）
福建省自然科学基金(2023J011691) （2023J011691）
厦门市医疗卫生指导性项目(3502Z20214ZD1082) （3502Z20214ZD1082）
