Objective To combine grey relation analysis and partial least squares regression model to establish the forecasting model of per-patient hospitalization expenses .Methods Gray relational analysis was used to filter out the main factors affecting per-patient hospitalization expenses ,and then collinearity was examined between these factors .Partial least squares regression was used to establish prediction model of per-patient hospitalization expenses ,and the prediction accuracy was proved .Results After filtered by gray relational analysis ,the order of the importance of factors affecting per-patient hospitalization expenses was the west-ern medicine fee ,traditional Chinese medicine fees ,diagnosis and treat fees ,other fees ,inspection fees ,bed fees and operation fees . The established partial least squares regression model had a higher accuracy on fitting and prediction ,with low average relative er-ror ,respectively ,-0 .000 2% and 0 .349 3% .Conclusion The gray relational analysis and partial least squares regression are suit-able for the influencing factors and prediction analysis of hospitalization costs .It provides a reference for data with the small sample size and high collinearity between the variables .% 目的结合灰色系统理论中的关联分析法与偏最小二乘回归模型，建立人均住院费用的预测模型。方法采用灰色关联分析筛选出人均住院费用的主要影响因子，对因子间进行共线性诊断，建立人均住院费用与主要影响因子间的偏最小二乘回归预测模型，通过实例证明该模型的预测精度。结果经灰色关联分析筛选出人均住院费用最主要影响因素为西药费、中药费，其次为诊疗费，其他费用、检查费、床费和手术费对人均住院费用影响也较大；偏最小二乘回归模型对住院费用拟合和预测准确率较高，平均相对误差较低，分别为-0．0002％、0．3493％。结论灰色关联分析与偏最小二乘回归适宜于住院费用影响因素与预测分析，可为样本量小、变量间存在严重共线性资料分析提供参考。
展开▼
