基于支持向量机的医院住院费用影响因素分析  

Analysis of influencing factors of hospitalization expenses based on support vector machine

作　　者：张颖[1] 孙统达[1] 李利杰[2] 刘海容[3] 朱绥 

机构地区：[1]宁波卫生职业技术学院健康服务与管理学院,浙江省宁波市315104 [2]宁波城市职业技术学院信息学院 [3]宁波市第一医院财务科 [4]宁波姜山社区卫生服务中心

出　　处：《中华医院管理杂志》2015年第5期392-396,共5页Chinese Journal of Hospital Administration

基　　金：浙江省医药卫生科技计划项目（2013KYB242）;浙江省卫生经济学会资助课题（2013）

摘　　要：目的利用支持向量机建模，对医院住院费用的主要影响因素进行分析，探讨有效的医疗费用影响因素分析方法。方法随机抽取浙江省6家医院，利用各医院电子病历系统，选取3种典型内外科疾病，建立支持向量机模型、BP神经网络模型、多元线性回归模型，分别对住院费用影响因素及其影响程度进行分析。利用支持向量机模型，分别对3个病种进行分析。结果基于径向基核函数的支持向量机模型对住院费用预测准确度最高，达到96．07％。当对各病种进行混合分析时，3种模型分析均显示住院费用的主要影响因素为住院天数、疾病种类、手术名称、医院代码；当对各病种进行单独分析时，不同病种的影响因素重要程度有所不同，但主要的影响因素基本一致。结论支持向量机在住院费用影响因素分析中具有可行性，通过其分析结果制定合理的单病种付费制度，可有效控制医疗费用的快速增长。Objective To analyze main influencing factors of hospitalization expenses by support vector machine modeling, and explore effective influence factors analysis methods of medical expenses. Methods Random selection of six hospitals in Zhejiang province. Using hospital electronic medical record system of the hospitals and selecting three kinds of typical diseases of internal medicine and surgery, to build the support vector machine model, BP neural network model, and multiple linear regression model for comparison of analysis results. The SVM model is used to analyze three various diseases. Results The support vector machine model based on radial basis kernel function scored the highest prediction accuracy on the hospitalization expenses, up to 96. 07%. In a mixed analysis of different diseases, analysis results of all three models pointed the main influence factors of hospitalization expense as days of stay, disease types, and hospital coding for the surgery. In the analysis by diseases individually, the influencing factors, though varying with diseases, key factors remain the same. Conclusion The support vector machine in the influence factor analysis is feasible in hospitalization expenses. According to the analysis results, the single disease payment system can be made rationally, which can effectively control excessive growth of medical expenses.

关 键 词：住院费用 支持向量机 影响因素 神经网络 多元线性回归 

分 类 号：R197.3[医药卫生—卫生事业管理]