Application Of Mapping Method In The Measurement Of Health Utility Value

Objective:To study on the application of mapping method on utility measures, taking FACT-L for example. To develop a mapping algorithm from FACT-L onto EQ-5D utility values,which is Chinese preference-based, to translate the QoL into health utility values to facilitate economic evalution of lung cancer. To explore the impact of the application of different econometrics,model types and utility value tariffs on the results of mapping.and to provide experience for mapping study in the future.Methods:The research data of this study are from "the study of pharmacoeconomics evaluation of Chinese lung cancer ". According to the inclusion and exclusion criteria of this study,676 cases are chosen whose information of basic demographic, FACT-L and EQ-5D questionnaire were collected. EQ-5D utility values are determined by CN, JP and UK tariff separately. All responses were randomly divided to development set(n=451) and validation set(n=255) with the balance of basic demographic between them. With the use of statistical softwares, relationships between FACT-L and EQ-5D questionnaire about the development set were estimated using ordinary least squares(OLS),generalized linear models(GLM),quantile regression,Tobit models and censored least absolute devistions (CLAD) in four model types. Then the internal validation were carried out by the validation set. The best econometric method and model type were chosen after comparations to estimate the final mapping model from total samples. The performance of each model was assessed in terms of how well responses to FACT-L predicted BQ-5D utilities using mean absolute error (MAE),root mean squared error (RMSE) and the proportion of absolute error greater than 0.05 or 0.10.Contents:On one hand, a mapping algorithm for estimating health utility values from FACT-L was developed. On the other hand,the impact and characteristics of the choice of different econometric methods, model types and tariffs applying to mapping was discussed.Results:1. In the developed algorithm for mapping the FACT-L instrument to the EQ-5D utility values based on Chinese preference, model including FACT-L domain scores variables and interaction terms of domain scores and clinical and demographic variables had the best predictive performance. In final mapping model,PWB.EWB and LCS domain scores of FACT-L significantly predicted the EQ-5D utility indices. The predicted and observed utilities was close to highly correlated according to the correlation coefficient of them was 0.7187. Both overall distribution and individual pair-test were no statistical difference significantly. The accuracy of prediction reached an acceptable degree with MAE equal to 0.0886 and RMSE equal to 0.1175. All above,the FACT-L based mapping function had better predictive ability.2. Quantile regression and CLAD were suitable for mapping with EQ-5D questionnaire as a target instrument. Specifically in mapping based on Chinese preference, quantile regression was recommended as the best econometrics. While start instrument was a multi-dimension instrument,the model type with domain scores as indepedents and clinical and demographic as covariates was recommended. In terms of accuracy of predition, mapping function of CN tariff was the best. Thus,the China tariffs should be recommended in mapping.Conclusions:The model mapping FACT-L onto EQ-5D based on Chinese preference had a good performance,which could translate the HRQol of patients with lung cancer excatly. The cases where the health utility values are missing,using mapping method to attain the economic evaluation results from HRQol measures could provide a reference for economic choice of disease treatment and for government's health decision-making. In the application of mapping, the tariff of target instrument should be chosen in terms of purpose and target people,and own tariff will be a best choice. With multidimensional measures, the model type of dimension scales as main effect demographic indics as covarites should be recommended,and econometrics should be chosen in terms of data types.