A Study Of Modeling For European Quality Of Life(EQ-5D-3L) Value Sets Based On Nationally Representative Sample Of Chinese Population

ObjectivesWorldwide,the disease pattern has shifted from communicable disease taking up the largest proportion to non-communicable disease.People live longer,however,with states not in full health increasing.Thus,quantity measures,such as life expectancy and mortality rate seem insufficient to reflect the state of the population health comprehensively.Quality-adjusted life years(QALYs)which combines both life expectancy(quantity)and health status(quality)is much more qualified to provide essential information for decision makers.In England,as an example,health technology assessments and reimbursement decisions must be based on economic evidence educed using QALYs.China has made great achievements in strengthening its health service system,including public health and medical insurance,but QALYs remain nowhere near widely adopted.The generic EQ-5D-3L instrument consists of the EQ-5D descriptive system and the Visual Analogue Scale(EQ VAS)and is used worldwide.In the year of 2008 and 2013,the EQ-5D-3L has been applied to measure health status among the general population in the National Health Services Survey(NHSS).However experience-based Visual Analog Scale(VAS)value set and Urban Time Trade-Off(TTO)value set are inapplicable to the common population.To date,many domestic studies choose to use Japanese value set instead.Under such circumstances,to conduct an exact valuation study of the EQ-5D-3L is of great importance.The aim of the present study is to obtain a Chinese EQ-5D-3L value set based on a nationally representative sample using VAS and TTO method.Once it succeeds,it would be feasible to figure out longitudinal changes of the Chinese quality of life,which would provide a solid foundation for policy making.MethodsThis study was designed as a field survey and a cross-sectional study.The target population was those Chinese people who aged 15 years and above in 2014.Sample size was 6,000 and the proportions of urban and rural participants were equal.A multi-stage stratified clustered randomized sampling was adopted.EQ-5D-3L was used to describe the health states of the participants.VAS and TTO methods were both carried out to measure the quality of life.A total of 17 socio-economic variables were used to narrate the common status of participants.After data washing,reliability analysis and validity analysis,VAS value sets and TTO value sets were constructed using three kinds of multi-variable regression:ordinary least square(OLS);general least square(GLS);weighted least square(WLS).Best fitting models were identified based on criteria.Sensitivity analysis was implemented using a randomly split data set.After identifying the robustness of the best fitting model,the best-fit VAS and TTO models were used to the sample for prediction.Comparing the predicted results helped to understand the two models.STATA/SE12.0(StataCorp,College Station,TX)was adopted,a=0.05.Nanjing Medical University(NMU)takes the responsibility of designing,training,guiding,quality control,data processing and Modeling with the help of the Center for Statistics and Information of the National Health Commission and EuroQol Group.ResultsA total of 5,939 inhabitants aging 15 years and above were completely interviewed.After data washing 5,884 and 5,503 participants' data were included in constructing VAS and TTO models,respectively.The best fitting VAS model is as below:HRQoL of VAS=1-0.0274-0.0727*MO2-0.1788*MO3-0.1105*SC2-0.2349*SC3-0.0880*UA2-0.1518*UA3-0.0898*PD2-0.1667*PD3-0.0851*AD2-0.1702*AD3The R-square was 0.670.The mean absolute error was 0.0319,and the correlation coefficient between predicted value and mean was 0.9837.The predicted VAS value of verification sample was 0.964±0.092.The 50th,10th and 5th percentile was 1,0.883 and 0.527,respectively.The best fitting TTO model is as below:HRQoL of TTO=1-0.0766*M02-0.2668*MO3-0.0441*SC2-0.2912*SC3-0.0370*UA2-0.0538*UA3-0.0274*PD2-0.0409*PD3-0.0359*AD2-0.1771*AD3The mean absolute error was 0.0830,and the correlation coefficient between predicted value and mean was 0.8990.The predicted VAS value was 0.985±0.052.The 50th,10th and 5th percentile was 1,0.964 and 0.779,respectively.ConclusionIn this study,the best fitting VAS and TTO value sets were obtained,which are robust and effective and have characteristics of Chinese traditional culture.Comparing the two best fitting models,the VAS value set is more sensitive than the TTO value set.To summary,the best fitting model of EQ-5D-3L VAS value set is strongly recommended to researchers for future studies on health-related quality of life of the Chinese population.