映射法在健康效用值测量中的应用研究

发布时间：2018-03-23 12:59

本文选题：成本-效用　切入点：肺癌治疗功能评价量表　出处：《北京中医药大学》2015年硕士论文

【摘要】：研究目的：以肺癌治疗功能评价量表FACT-L为例,对映射法在效用值测量中的应用进行初探,构建基于我国人群偏好的FACT-L到EQ-5D效用值的映射模型,将非偏好的生命质量信息转换为健康效用值从而应用于肺癌治疗领域的成本-效用评价之中。探索不同计量经济学方法、模型类型以及效用积分体系版本的选择对映射结果的影响,并对最佳选择给出建议意见,以期为我国今后开展的映射研究提供经验借鉴。研究方法：本研究资料来自于“肺癌药物经济学评价研究”课题中的历史数据。根据纳排标准选择研究对象共676例,并收集其基本信息、FACT-L中文版(V4.0)以及中文版EQ-5D数据,分别采用中、日、英三版效用值积分体系进行效用值换算。将总体研究对象随机区组分为构建集(n=451)和验证集(n=225),并保证两者之间基本人口学指标的均衡性。运用统计软件,在构建集基础上,以普通最小二乘法、广义线性模型、Tobit模型、截断最小绝对离差以及分位数回归五种计量经济学方法结合四种模型类型建立FACT-L到EQ-5D的备选模型,并以验证集对其进行内部验证。比较后选择最佳方法和模型类型对总样本进行估计,获得最终的映射模型。模型效力主要从差异性检验以及预测精度指标如平均绝对误差、均方根误差等方面进行评价。研究内容：首先,构建基于我国人群偏好的FACT-L效用值映射模型；其次,探讨映射法中不同计量经济学方法、模型类型以及效用积分体系的选择应用特点。研究结果：1.基于我国人群偏好的FACT-L到EQ-5D的映射中,包含有FACT-L维度得分及其两两交叉项和人口学、健康状态变量的模型表现出最佳的预测能力。在最终映射模型中,主效应生理状况PWB、情感状况EWB和癌症子模块LCS维度得分的回归系数具有统计学意义。该模型的预测值与实际值相关系数rs=0.7187,为中偏高度相关；二者的总体分布及个体配对比较均无统计学差异；预测精度指标MAE=0.0886,RMSE=0.1175,与国外映射研究结果相比,处于较优水平。综合来看,可认为该映射模型预测能力良好。2.对于以EQ-5D作为目标量表的映射研究,分位数回归和截断最小绝对离差较为适用,当基于我国人群偏好时,推荐将分位数回归作为首选的计量经济学方法。当起始量表为多维量表时,推荐将以各维度得分作为自变量、以人口学指标等作为协变量的模型类型。中、日、英三版EQ-5D效用积分体系结果的映射相比较,中版最优,其次为日版。因此,在映射中建议使用效用积分体系的本国版本或与本国人群最为贴近的他国版本。研究结论：运用映射法基于我国人群偏好构建的FACT-L到EQ-5D效用值的转换模型,其预测能力良好,具有较高的模型效力,能够较为精确地将我国肺癌患者的非偏好生命质量信息转换为健康效用值。当研究中缺少效用值数据时,可对健康相关生命质量量表进行映射,以此获得的成本-效用分析结果能够为疾病治疗措施的经济性选择提供依据,为政府的卫生资源配置和政策制定提供参考。映射法应用时,应根据研究目的和目标人群选择适合的效用值积分体系版本,建议选择研究对象的国家版本或与其最为贴近的他国版本；对于多维量表的映射,建议以维度得分作为自变量主效应,并加入其二阶变量和人口学指标协变量进行调整与校正；此外还应根据资料数据的类型特征灵活选择计量经济学方法。
[Abstract]:Objective: To study the functional assessment of cancer FACT-L as an example, to make a preliminary study on the application of measurement in utility value in mapping method, construct the mapping model preference of Chinese population FACT-L to EQ-5D based on utility value, the quality of life for the non preference information into the health utility value evaluation from the utility and application in the field of lung cancer cost. To explore different econometric methods, effect and utility model type integral system version selection results of mapping, and to give advice on the best choice, in order to provide good experience for our future work on mapping. Research methods: the research data of historical data from "research" evaluation of lung cancer drug economics. According to the inclusive criteria for the selection of row in 676 subjects, and to collect the basic information, FACT-L Chinese Version (V4.0) and Chinese version of EQ-5D data, divided by In Japan, the British version three utility value integral system utility value conversion. Total subjects were randomly divided into set (n=451) and validation set (n=225), and to ensure the balance between the two basic demographic indicators. Using statistical software, set up in the basis, to the general least square method. The generalized linear model, Tobit model, truncated least absolute deviation and quantile regression five econometric method with alternative models of four kinds of model type FACT-L to EQ-5D, and the validation set to validate the internal comparison. Choose the best method and model types of total samples to obtain final estimation, mapping model the main difference from the model. The effectiveness of inspection and prediction accuracy indicators such as mean absolute error, root mean square error evaluation and so on. Research contents: first, construct the value preference of Chinese population FACT-L mapping based on Utility Model; secondly, explore the different econometric methods in the selection and application of feature mapping method, and the utility model type integral system. The research results: 1. based on the preference mapping population of our country's FACT-L to EQ-5D, including FACT-L score and 22 cross terms and demographic and health status variables model showed the best predictive ability in the final mapping model, the main effect of physiological condition PWB, statistically significant regression coefficients of emotional status EWB and cancer sub module LCS scores. The prediction value and the actual value of the correlation coefficient rs= 0.7187, is highly related to the overall distribution and partial; individual paired comparison of the two were not statistically significant; the prediction accuracy of MAE=0.0886 index, RMSE=0.1175, compared with foreign research results in a mapping, excellent level. In general, we can consider the mapping model good prediction ability for.2. With EQ-5D as the research target mapping scale, quantile regression and truncated least absolute deviation is appropriate when the preference of Chinese population based on quantile regression, is recommended as the preferred method of econometrics. When the initial scale for the multidimensional scale, each dimension score will be recommended as independent variables to demography model type indicators as covariates. In Japan, the British mapping three edition EQ-5D utility compared to the integral system results in the second edition, the best for the Japanese version. Therefore, the Chinese version is close to he in the map suggests the use of utility system and its integral version or its research conclusion: the use of population. China's population preference mapping method conversion to construct FACT-L to EQ-5D utility model based on the prediction ability of good effect, high model can accurately be non preference life quality of patients with lung cancer transfer information in China For the health utility value. When the utility value of lack of data in the research, the health related quality of life scale mapping, in order to obtain the cost utility analysis results can provide the basis for the selection of economic disease treatment measures, provide a reference for the allocation of health resources and policies for the government. The application of mapping method. Should choose according to the research purpose and target population of the utility value of integral system version, recommendations for the selection of research object and the national version or the Chinese version of his close; mapping on a multidimensional scale, to score as independent main effect, and adding second order variables and demographic indicators adjusted for covariates and correction; it should also be the econometrics method flexibly according to types of data.

【学位授予单位】：北京中医药大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R1

【参考文献】