多元岭回归方法在医院成本分析中的应用  

The Application of Ridge Regression to Hospital Cost Analysis

作　　者：胡军[1] 姜潮[1] 图易宸 Jerry.Han 刘启贵[1] 刘宗航[1] 苏玉宏[1] 

机构地区：[1]大连医科大学,116027 [2]辽宁省卫生厅 [3]美国纽约州立大学

出　　处：《中国卫生统计》2000年第6期334-336,共3页Chinese Journal of Health Statistics

摘　　要：目的　确定住院病人成本的影响因素 ,并对其进行定量测量。方法　采用SPSS ,NoSA软件用最小二乘法建立住院病人总成本的多元线性回归模型 ,对其进行共线性诊断 ;与岭回归方法建立的回归模型进行了对比 ;对住院病人规模、利用等医院成本影响因素的相关分析 ;最后用岭回归方法建立了 2 3个变量的大型多元回归计量经济模型。结果由于自变量间多元共线性的效应 ,使用最小二乘法建立的回归模型回归系数有较大的标准误 ,回归模型结果不稳定 ,用岭回归方法改善了多元共线性效应 ,降低了回归系数的标准误 ,使回归系数结果更为可靠。结论　确定了医院住院病人总成本影响因素 ,并提出了医院降低成本的途径。Objective To determine factors which have important effects on hospital cost of inpatients,and measure them quantitatively.Methods we establish hospital cost models by the method of least squares regression and perform collinearity diagnosis with SPSS and NoSA software.The comparison of two models between least squares regression and ridge regression is also done.The correlation analysis of cost affecting factors is studied.At last the large economic model is set up including 23 independent variables.Results There is multi-collinearity between independent variables,so instability of regression coefficients is produced while using least squares method.ridge regression can improve the effect of multi-collinearity and improve the stability of the models.Conclusion Affecting facros on hospital cost are determined by the way of setting up ridge regression model.In the mean time the ways of decreasing cost are provided.

关 键 词：医院成本 多元岭回归方法 多元共线性 

分 类 号：R311[医药卫生—基础医学]