基于标准赌博法探索EQ-5D量表综合指数模型
Exploring EQ-5D value set with standard gamble method

目的:采用正交设计典型状态基于标准赌博法(SG)探索构建欧洲生命质量量表(EQ-5D)评分体系。方法:采用分层整群随机抽样的方法,从医学院校按年级、专业分层抽取在校大学生250人,选取L18(2*3^7)正交设计表得到18个标准状态。应用STATA/SE 12.0构建9个加权最小二乘多元线性回归模型,模型包括11个哑变量。结果:总体水平上以中位数为基础构建的WLS回归模型(模型AMWls00)最优,该模型无常数项、无交互作用N3变量,具有简洁、高效的特点,其调整后决定系数为0.998 8,平均绝对误差为0.031 5,18状态的预测值与均值相关系数为0.976 0,逻辑错误数为0。且生命质量评分曲线具有明显的中国人文化特征。结论:总体水平上以中位数为基础无N3、无常数项的WLS回归模型(AMWls00)是最优模型,该建模方法最为合理。
Aim: To find the EQ-5D value set with the typical state of orthogonal design by the standard gamble(SG)for the general Chinese population.Methods: Using a stratified cluster random sampling method, 250 college students were selected from the medical college stratified by the grade and the major.Eighteen states were designed by L18(2*3^7). WLS multivariate linear regression models were constructed by STATA/SE 12.0. In total, 11 dummy variables were included in these models.Results: AMWls00 was estimated using WLS regression based on median at the aggregate level, and it was the best-performing model. It had no constant or N3 term of interactive variable with the advantages of parsimony and high efficiency(adjusted R2=0.998 8, MAE=0.031 5, r=0.976 0, logical error=0). The quality of life score curve had traditional Chinese cultural characteristics.Conclusion: The WLS regression model based on median at the aggregate level without N3 and constant(AMWls00)yields the best-performing model,and is the most reasonable method to SG modeling valuation