构建COPD再入院患者风险评估神经网络模型  

Construction of Neural Network Model for Risk Assessment of COPD Re-admission Patients

作　　者：严武[1] 叶荔姗[1] 伊菁华 YAN Wu;YE Lishan;YI Qinghua(Zhong Shan Hospital of Xiamen University,Xiamen 361004,China)

机构地区：[1]厦门大学附属中山医院信息中心,厦门市361004

出　　处：《中国卫生信息管理杂志》2018年第4期474-477,共4页Chinese Journal of Health Informatics and Management

摘　　要：目的利用BP神经网络的理论和算法,对COPD患者的历史数据进行分析,构建出COPD再入院患者的风险评估模型,通过对COPD再入院患者各相关因素的敏感度分析和疾病风险评估及分析,为BP神经网络建模在临床诊疗中的应用提供一定的参考,并为医疗资源的合理配置提供较为有效的解决方案。方法编写结构化查询语句,从HIS数据库抽取相关数据,导入Clementine 11.1中,利用BP神经网络算法进行建模,预测结果用SPSS 22.0进行模型的建模效果评估以及模型建模效果的假设检验。结果经过优化后的BP神经网络的拟合度为71.743%,预测准确度93.55%。在所有相关影响因素中,入院次数和入院状态对COPD患者的再入院风险度影响最大。在预测效果上,BP神经网络要优于传统的多元统计分析方法。Objective Based on the theory and algorithm of BP neural network, the historical data of COPD patients were analyzed, and a risk assessment model of COPD readmission patients was constructed. Through sensitivity analysis and risk assessment and analysis of the related factors of the COPD readmitted patients, it provided a reference for the application of BP neural network modeling in clinical diagnosis and treatment, and a more effective solution for the rational allocation of medical resources. Methods SQL statements were written, data extracted from HIS database were imported into Clementine 11.1, and BP neural network algorithm was used for modeling. The result was predicted by SPSS 22, and the hypothesis test of model building effect was carried out. Results The fitting degree of the optimized BP neural network is 71.743%, and the prediction accuracy is 93.55%. Among all the related factors, the number and status of admissions had the greatest impact on the re-admission risk of COPD patients. In terms of the prediction effect, the BP neural network is better than the traditional multivariate statistical analysis method.

关 键 词：再入院率 BP神经网络 多元线性回归 影响因素 

分 类 号：R-39[医药卫生] R319