BP神经网络技术在慢性病患者住院费用研究中的应用  

The Application of BP Neural Network in Hospital Expense Research of Chronic Diseases

作　　者：王敏[1] 张开金[1] 姜丽[1] 黄新[1] 包思敏 

机构地区：[1]东南大学公共卫生学院,南京210009

出　　处：《中国卫生经济》2010年第11期86-88,共3页Chinese Health Economics

基　　金：国家自然科学基金重点项目(70973018)

摘　　要：目的:对慢性病患者住院费用进行分析,了解住院费用的影响因素,为慢性病管理中患者费用控制提供政策建议;方法:运用SPSS Clementine11.1,以年龄、职业等因素为输入神经元,住院费用和住院天数分别为两个模型的输出神经元建立BP神经网络模型,以神经网络模型得出的各变量的敏感度来评价各因素对住院费用和住院天数的影响;同时,运用SPSS13.0建立多元线性回归模型,将两模型分析结果进行比较;结果:两模型结果显示,对住院费用影响排前三位的变量是住院天数、检查费占比和药占比。讨论:BP神经网络与多元线性回归都能很好的运用与住院费用影响因素分析,但两分析工具在适用条件、信息挖掘深度和影响因素分析功能3个方面都还存在着差异。Objectives：On the basis of hospital expense analysis of chronic diseases,to study their impact factors and put forward some expense control policy suggestions in chronic disease management.Methods：Taking age,vocation and other factors as input neuron,hospital expense and hospital stay as output neuron separately to build simulation model of BP neural network,and to evaluate the influence of each input factors to hospital expense and hospital stays by using sensitivity indicators of the model.Comparison between BP neural network and traditional ways of expense analysis has been done through establishing a multiple linear regression model using SPSS13.0.Results：Both the two models results show that the first three leading impact factors are hospital stays,the proportion of infection fee and the proportion of drug fee.Discussion：Both BP neural network and multiple linear regression models are all able to apply well in impact factor analysis of hospital expense,but they still have differences in applicable conditions,information mining depth and analysis function of impact fact.

关 键 词：慢性病 住院费用 BP神经网络 多元线性回归 影响因素 

分 类 号：R197.3[医药卫生—卫生事业管理]