摘要
目的基于机器学习构建腹腔镜胆囊切除术(LC)患者发生全身麻醉诱导后低血压(PIH)的预测模型。方法回顾性分析2019年5月至2023年9月行LC手术的患者资料,按照7∶3的比例分配训练集和验证集。使用Logistic回归、随机森林(RF)、支持向量机(SVM)方法构建PIH的预测模型,模型的区分准确度使用受试者工作特征(ROC)曲线下面积进行评估,模型的校准度采用校准曲线与霍斯默(H-L)检验进行评估,采用列线图对模型可视化。结果共纳入患者260例,其中有82例(31.5%)发生PIH,训练集182例,验证集78例,其中训练集中有58例(31.9%)发生PIH,验证集中有24例(30.8%)发生PIH。年龄、BMI、使用血管紧张素转化酶抑制剂(ACEIs)/血管紧张素受体阻滞剂(ARBs)、基础HR和MAP、入室HR和MAP、HR变化值以及MAP变化值是预测PIH的影响因素。基于机器学习构建的Logistic回归模型的预测性能最佳,在验证集中曲线下面积(AUC)、准确度、召回率分别为0.93(95%CI 0.90~0.99)、0.81、0.81。结论基于机器学习构建的Logistic回归模型具有良好的预测性能,筛选出的预测变量为年龄、BMI、使用ACEIs/ARBs、基础HR和MAP、入室HR和MAP、HR变化值以及MAP变化值,可快速准确评估LC患者PIH的发生风险。
Objective To construct a clinical prediction model of general anesthesia post-induction hypotension(PIH)in patients undergoing laparoscopic cholecystectomy(LC)based on machine learning.Methods The data of patients who underwent LC surgery from May 2019 to September 2023 were retrospectively selected.The training set and the validation set were allocated at a ratio of 7∶3.Logistic regression,random forest(RF),and support vector machine(SVM)were used to construct the prediction model of PIH.The discrimination accuracy of the model was evaluated by the area under the receiver operating characteristic(ROC)curve,and the calibration of the model was evaluated by calibration curve and Hosmer-Lemeshow(H-L)test.The nomogram was used for visual interpretation of the model.Results A total of 260 patients were included,and 82 patients(31.5%)developed PIH.There were 182 patients in the training set and 78 patients in the validation set.Fifty-eight patients(31.9%)in the training set developed PIH,and 24 patients(30.8%)in the validation set developed PIH.Age,BMI,use of angiotensin converting enzyme inhibitors(ACEIs)/angiotensin receptor blockers(ARBs),basal HR and MAP,HR and MAP on entering the operating room,HR change,and MAP change were important factors for the model to predict PIH.The logistic regression model had the best prediction performance,with AUC,accuracy and recall rate of 0.93(95%CI 0.90-0.99),0.81,and 0.81 in validation set,respectively.Conclusion The logistic regression model based on machine learnin has good performance in predicting PIH.The selected predictive variables are age,BMI,use of ACEIs/ARBs,basal HR and MAP,HR and MAP on entering the operating room,HR and MAP change,which can quickly and accurately assess the risk of PIH.
作者
汪宁
王纯辉
张野
WANG Ning;WANG Chunhui;ZHANG Ye(Department of Anesthesiology and Perioperative Medicine,Second Hospital of Anhui Medical University,Hefei 230601,China)
出处
《临床麻醉学杂志》
CAS
CSCD
北大核心
2024年第9期911-916,共6页
Journal of Clinical Anesthesiology
基金
安徽省高校自然科学研究项目(2023AH010081)
安徽医科大学校科研基金(2023xkj049)。
关键词
腹腔镜胆囊切除术
全身麻醉
诱导后低血压
机器学习
预测模型
Laparoscopic cholecystectomy
General anesthesia
post-induction hypotension
Machine learning
Prediction model
