基于人工神经网络技术构建围术期病人低体温风险预测模型  

Construction of prediction model of perioperative hypothermia risk based on artificial neural network technology

作　　者：项海燕[1] 黄立峰[1] 朱锋杰[1] 张浩[1] XIANG Haiyan;HUANG Lifeng;ZHU Fengjie;ZHANG Hao(The Second Affiliated Hospital of Zhejiang University School of Medicine,Zhejiang 310009 China)

机构地区：[1]浙江大学医学院附属第二医院,浙江310009

出　　处：《护理研究》2022年第5期767-772,共6页Chinese Nursing Research

基　　金：浙江省医药卫生科技计划项目,编号:2020KY146。

摘　　要：目的:运用具有深度学习能力的神经网络技术构建并检验围术期病人低体温风险预测模型。方法:回顾性收集临床病例大数据,通过循证和临床相结合的方式筛选出预测模型的建模变量,基于神经网络技术构建预测模型,通过预测模型的准确率、特异度和曲线下面积(AUC)等指标检测其性能。结果:建模的训练集和测试集病例资料的基线特征一致(P>0.05);训练集和测试集预测模型准确率均>75%,阴性预测值>0.9,特异度>0.8,AUC>0.70,模型决策曲线优于随机方案。结论:基于大数据运用神经网络技术构建围术期病人低体温风险预测模型具有可行性,能够为科学预测围术期低体温提供新的方法。Objective:Using artificial neural network(ANN)technology with deep learning capability,prediction model of perioperative hypothermia risk was established and tested.Methods:The big data of clinical cases were retrospectively collected,and the modeling variables of the prediction model were selected through the combination of evidence-based and clinical methods.The prediction model was constructed based on neural network technology,and its performance was tested by the accuracy,specificity and AUC of the prediction model.Results:The baseline characteristics of case data in the model training set and test set were consistent(P>0.05).The prediction model accuracy of training set and test set was>75%,the negative prediction value was>0.9,the specificity was>0.8,and the AUC was>0.70.The model decision curve was better than random scheme.Conclusion:It is feasible to construct a prediction model of perioperative hypothermia risk by using neural network technology based on big data,and it can provide a new method for scientific prediction of perioperative hypothermia.

关 键 词：神经网络技术 围术期 低体温 预测模型 深度学习 人工智能 

分 类 号：R473.6[医药卫生—护理学]