摘要
目的:构建老年上消化道穿孔患者预后的新型评估系统及应用机器学习分类器,有助于入院快速预测老年上消化道穿孔患者的预后。方法:回顾性收集在2017年6月—2023年7月收住福建医科大学附属协和医院急诊外科诊断为老年上消化道穿孔的95例患者。收集患者的临床资料和实验室检查资料,并对患者的术后严重并发症进行分级。采用二阶聚类分组(TWO-STEP cluster grouping,TSC)对其预后进行自动分组,分为预后良好组(GP组,70例)和预后不良组(PP组,25例)。利用机器学习分类器综合入院多因素对老年上消化道穿孔患者的预后进行预测,并采用受试者工作特征(ROC)曲线分析其预测效果。结果:PP组患者术后胃肠道恢复时间、重症监护时间及住院费用均明显高于GP组;对比TSC评估系统和术后严重并发症(severe adverse events,SAE)分级对住院日的区分度,发现TSC评估系统相较于SAE分级有更好的区分度(TSC:P<0.001,SAE:P=0.01)。进一步对比两组患者的入院情况,发现PP组年龄明显高于GP组[77.00(71.50~82.50)岁vs 72.00(67.00~78.00)岁,P=0.043],其术前外周血白蛋白的水平明显低于GP组。对比不同机器学习分类器对TSC评估系统的预测效能,发现自适应增强分类器的预测效能最优,其曲线下面积(AUC)为0.97(95%CI:0.52~1.00,精确度为0.86)。结论:TSC评估系统能有效针对老年上消化道穿孔患者的预后进行评估。超高龄和低白蛋白血症是上消化道穿孔患者预后不良的独立危险因素。自适应增强分类器有助于入院快速预测老年上消化道穿孔患者的预后,协助临床诊疗。
Objective:To construct a novel assessment system for the prognosis of elderly patients with upper gastrointestinal perforation and the application of machine learning classifiers,for admission to quickly predict the prognosis of elderly patients with upper gastrointestinal perforation.Methods:A total of 95 patients admitted to the Department of Emergency Surgery of Union Hospital of Fujian Medical University from June 2017 to July 2023 with a diagnosis of upper gastrointestinal perforation in the elderly were retrospectively collected.The clinical data and laboratory examination data of the patients were collected,and the postoperative serious complications of the patients were graded.Their prognosis was automatically grouped into good prognosis group(GP group,70 cases) and poor prognosis group(PP group,25 cases) using TWO-STEP cluster grouping(TSC).The prognosis of elderly patients with upper gastrointestinal perforation was predicted by integrating admission multifactors using machine learning classifiers,and the predictive effect was analysed by using the subject's work characteristics(ROC) curve.Results:The postoperative gastrointestinal recovery time,intensive care time,and hospitalisation cost of the patients in the PP group were significantly higher than those in the GP group.Comparing the differentiation of hospitalisation days between the TSC assessment system and the severe adverse events(SAE) classification,it was found that:the TSC assessment system had a better differentiation compared with the SAE classification(TSC:P<0.001,SAE:P=0.01).Further comparing the admission status of the two groups,it was found that:the PP group was significantly older than the GP group(77.00[71.50-82.50] vs 72.00[67.00-78.00],P=0.043),and its preoperative peripheral blood albumin level was significantly lower than that of the GP group.Comparing the predictive efficacy of different machine learning classifiers for the TSC assessment system,it was found that:the adaptive boosting classifier(AB) had the best predictive efficacy,with an area under the curve(AUC) of 0.97(95%CI:0.52-1.00,precision 0.86).Conclusion:The TSC assessment system is effective in targeting the prognosis of elderly patients with upper gastrointestinal perforation.Advanced age and hypoalbuminaemia were independent risk factors for poor prognosis in patients with upper gastrointestinal perforation.The AB helps to rapidly predict the prognosis of elderly patients with upper gastrointestinal perforation on admission and assists in clinical management.
作者
陈帅
黄藏典
涂鹏声
张俊榕
陈先强
CHEN Shuai;HUANG Cangdian;TU Pengsheng;ZHANG Junrong;CHEN Xianqiang(Department of Emergency Surgery,Union Hospital of Fujian Medical University,Fuzhou,350001,China)
出处
《临床急诊杂志》
CAS
2024年第5期239-245,共7页
Journal of Clinical Emergency
基金
福建省医疗“创双高”建设经费资助[No:闽卫医政[2021]76号]。
关键词
老年上消化道穿孔
二阶聚类
预后
机器学习
自适应增强分类器
elderly upper gastrointestinal perforation
TWO-STEP cluster
prognosis
machine learning
adaptiveboosting classifier
