SF-6D Health Utility Value Set Study

Background:With China’s economic and social development,population growth,changes in the disease spectrum,and improvement of the medical security system,residents’demand for medical services is growing,and the contradiction between health resources and growing demand for health services is becoming increasingly prominent.China’s future medical and health development needs to further optimize the allocation of health resources and build a medical and health service system that can better meet people’s health service needs.Health economics evaluation is the main method to study the rational allocation of health resources,including cost-benefit analysis,cost-effectiveness analysis,and cost-utility analysis（CUA）.The result of CUA can comprehensively evaluate the output of health preference in physiological,psychological,and social dimensions and allocate health resources more effectively.Many countries in the world have taken it as the main basis for health resource allocation.The health utility value set is an important factor related to the accuracy and effectiveness of evaluation results in CUA research.At present,the most commonly used method used to develop health utility value set is visual-analogue scale（VAS）,time trade-off（TTO）,standard game（SG）,and other direct measurement methods,which usually costs a lot of investigation resources,and has the problems of difficult understanding and heavy investigation burden.Recently,only EQ-5D-3L and EQ-5D-5L utility value sets have been established in China,and the two utility value sets have a serious ceiling effect,which may affect the allocation of health resources.Seven countries and regions have established the corresponding utility value sets of SF-6D,but the health utility value set based on Chinese population preference has not yet been established.Studies have shown that SF-6D is more sensitive than EQ-5D in people with different chronic diseases or health problems,and the difference between the two measurement scales may affect the allocation of health resources.Therefore,it is necessary to build the SF-6D health utility value set based on the preferences of Chinese population,to provide a reasonable calculation method for the CUA research based on the SF-6D scale in China,and to promote it to play its due role in the rational allocation of health resources in China.Objects:This study will use of discrete choice experiment（DCE）to construct SF-6D health utility value sets based on the preference of the Chinese population,to evaluate the health utility of the population more accurately in the research of health resource allocation,provide a reference for the allocation of health resources,and provide experience for the construction of utility value sets of other utility measurement scales in China.Method:The data of this study come from two parts:DCE survey data and Doctoral Fund data.The subjects of DCE were the general population over 16 years old who had the ability of personal cognition,clear thinking,and accurate expression.From September 2019 to April 2020,the snowball sampling method was used for data collection,and the prepared online questionnaire was sent to the respondents in the form of a link or QR code by using network resources.The survey included general demographic characteristics,SF-36v2,and DCE data.The second part is the data of the Doctoral Fundation Program,which comes from the general population quality of life survey in 2012.The subjects were adults over 18 years old who had no cognitive impairment,mental disorder and could express themselves.In this study,respondents were sampled in the central urban area of Chengdu and the surrounding towns by multistage stratified cluster sampling.The questionnaire included general demographic characteristics,health status,SF-36v2,and EQ-5D-3L.In the process of DCE design,we take six dimensions of SF-6D and one time dimensionas the attributes of DCE health states.The level of each attribute is between4 and 7.A total of 126000 health states can be described.After excluding unreasonable health states,the remaining health states are 120750.According to DCE classification,this study belongs to unbalanced design,so orthogonal arrangement fractional-factorial design is adopted.By using Kuhfeld’s SAS program for the D-efficiency orthogonal design,180 health state combinations were generated,and the D-efficiency was 99.52%.To improve the quality and completion rate of DCE survey data,180 selection sets of health status were randomly assigned to 15 blocks,each blocks contains 12 selection sets to reduce the survey burden.In DCE data analysis,to meet the requirements of quality of life and survival time of the QALY concept,we included the survival time as an interaction term into conditional logistic utility model A of DCE.To explore the inconsistency of SF-6D utility coefficient,the interaction effect between dimensions,death option data,and nonlinear model,utility function models B,C,D,and E are constructed based on the model A.According to the ratio of marginal utilities（RMU）,the SF-6D health utility value set of the corresponding model is constructed.AIC and BIC were used to evaluate the model fitting.The smaller AIC and BIC,the better the model fitting.The SF-6D health utility value set of this study was compared with the SF-6D health utility value sets in Australia,Spain,the United Kingdom,China’s Hong Kong,Portugal,and Brazil.The consistency between the SF-6D health utility value set of this study and other utility value sets was evaluated by using a specific health state utility trend chart and Bland-Altman plot.SF-6D utility value set of this study was also compared with the Chinese EQ-5D-3L utility value set in the Chinese general population,and the differences between different scales in the same country and region were evaluated,including ceiling effect,floor effect,consistency analysis,discriminative validity,effect scale,and relative validity analysis.Result:In the DCE part,877 questionnaires were collected,816 of which were valid,the effective rate was 93.04%,and the average completion time was 11.60 minutes.816respondents came from 27 provinces,among which Shandong,Sichuan,Chongqing,Guizhou,and Hubei were the most.The average age was 27.64±9.90 years old.There were 293 males（35.91%）,296 married people（36.27%）,521 urban residents（63.85%）,59 patients with chronic diseases（7.23%）,and 117 patients with two-week illness or discomfort（14.34%）.According to the utility function model,we take level 1 of SF-6D dimensions as the reference level,and construct the DCE utility linear model A,B1,B2,C1,C2,C3,nonlinear model D,and model E with death option data by conditional logistic regression.There are 25 parameters in model A,there are 5 utility coefficients and 3positive utility coefficients which are not consistent with the dimension level.There are five inconsistent utility coefficients in models B1 and B2.Model C is based on models A and B,and the worst health level interaction terms MOST are added.Only the MOST term of C1 model is statistically negative utility,which is-0.0016,and in C2 and C3 is a positive utility coefficient without practical significance.The interaction effect between each dimension and the square of expected life time is added to the nonlinear model D,but 23 positive utility coefficients are found without practical significance.In model E,which contains death option data,two inconsistence coefficients and two positive utility coefficients appear in model E1.There are three inconsistence coefficients in E2,E3,and E4,but there are no positive utility coefficients.The coefficient of inconsistency appeared at level 4,5 of Physical functioning,level 3,4 of Pain and level 3,4 of Vitality.The BIC results of model E3 are the lowest,and the interaction effect between dimension level and survival time better reflects the logic of negative utility of SF-6D dimensions,which indicates that the model fitting effect is better.Although MOST in E4 has a negative effect on health utility（coefficient is less than 0）,the estimated coefficient is not statistically significant,and the model fitting degree has no obvious improved.According to the formula of RMU,the SF-6D health utility sets is constructed.The utility value of model A ranged from-0.2881 to 1,and 303 health states（utility value less than 0）were worse than death,accounting for 1.68%;The B1 utility value set ranged from-0.2620 to 1.00,and 271 health states were worse than death,accounting for 1.51%;The B2 utility value set ranged from-0.2454 to 1.00,and 248 health states were worse than death,accounting for 1.37%.The C1 utility value set range from-0.2824 to 1,and 301 health states were worse than death,accounting for 1.67%.Compared with Model A utility value set,the C1 utility value set had no significant improvement.The utility range of C2 was significantly reduced.The C2 utility value set range was-0.1204～1.00,only 39 health states worse health than death,accounting for 0.22%.The C3 utility value set ranged from-0.2545 to 1.00,and 295（1.64%）health states worse health than death.Depend on different expected survival times of 5 years,10 years,and 15 years,the range of health utility value sets of model D is-0.2112～1.00,-0.4029～1.00,and-0.4691～1.00,respectively.The number of health states worse than death was 155（0.86%）,697（3.87%）and 1130（6.28%）respectively.The model E utility value sets are constructed by incorporating death option data into the analysis.The range of E1,E2,E3,and E4 utility value sets is 0.0774～1,0.0834～1.00,0.0809～1.00,and 0.0983～1.00,respectively.And the range of E1～E4 utility value sets is significantly narrowed.The results show that the utility coefficient of each dimension level of B1 is a better representative of SF-6D health state utility value,the range of utility value is large,and the fitting degree of BIC is better.Thus B1 health utility value set is selected as the health utility value set of SF-6D,represented by SF-6D_CN.In the cross-country comparison,the SF-6D utility value set of different countries and regions is used to calculate 187 health states utility values.The worse the health state is,the greater the utility difference between SF-6D_CN and other countris and regions SF-6D utility value sets are.Bland-Altman plot consistency analysis showed that more than 95%of the health utility values of SF-6D_CN and other countris and regions SF-6D health utility value sets were in the consistent range,and the average difference between SF-6D_CN and Austrilian SF-6D utiltity value set constructed by the DCE method was 0.116;The 95%consistency range was-0.114～0.375,and the utility difference of 886 health states（4.92%）exceeded the 95%consistency range;There is no obvious linear trend between the utility difference and average utility value between SF-6D_CN and Austrilian SF-6D utiltity value set,and the utility difference obeys normal distribution.The average difference between SF-6D_CN and Spanish SF-6D utiltity value set is 0.133;The 95%consistency interval was-0.108～0.373,and 831 health state utility differences exceeded the 95%consistency interval,accounting for 4.62%.The average difference between SF-6D_CN and the UK SF-6D utiltity value set was-0.153,and the95%consistency interval was-0.431～0.125.784 health state utility differences exceeded the 95%consistency interval,accounting for 4.36%;The average difference between SF-6dcn and China’s Hong Kong SF-6D utiltity value set was-0.105,and the 95%consistency interval was-0.339～0.128.854 health state utility differences exceeded the 95%consistency interval,accounting for 4.74%;The average difference between SF-6D_CN and Portugal SF-6D utiltity value set was-0.294,and the 95%consistency interval was-0.544～-0.044.842 health state utility differences exceeded the 95%consistency interval,accounting for 4.68%;The average difference between SF-6D_CNand Brazil SF-6D utiltity value set was-0.141,and the 95%consistency interval was-0.406～0.123.780 health state utility differences exceeded the 95%consistency interval,accounting for 4.33%.The differences between SF-6D_CN and UK,China’s Hong Kong,Portual,Brazil utility value sets are all left-skewed distribution,and there is an obvious linear trend between utility differences and average utility value,with the increase of average utility value.The comparison between SF-6D and EQ-5D-3L in Chinese general population was based on the survey data of the Doctoral Fund.A total of 2186 people were investigated in this study.2182 questionnaires were returned and 2178 were valid,with an effective rate of 99.8%.The average age of the respondents was 46.09±17.49 years old,males accounted for 44.72%,married 1661,accounted for 76.26%;The number of primary schools and junior high school education level accounted for 58.86%;In terms of health status,673 patients（30.9%）had chronic diseases;Two weeks before the survey,753（34.57%）had discomfort or outpatient experience.The correlation coefficient of theoretically related dimensions between SF-6D and EQ-5D-3L range from 0.20 to 0.51,and of theoretically unrelated dimensions from 0.04to 0.18.In terms of ceiling effect and floor effect,the ceiling effect of each dimension of EQ-5D-3L is between 80.62%and 98.35%,and the ceiling effect of each dimension of SF-6D is between 7.58%and 57.81%.The floor effect of the role dimension of SF-6D was 21.03%,and the other dimensions of the two scales had no floor effect.The SF-6D utility value is calculated using the Chinese SF-6D health utility value set(SF-6D_CN),Australia health utility value set(SF-6D_AU)and China’s Hong Kong health utility value set(SF-6D_HK).The EQ-5D-3L utility value is calculated using the health utility value system of Chinese population(EQ-5D-3L_CN).The utility value of EQ-5D-3L_CNhas a high ceiling effect,which is 74.61%.The utility value of SF-6D has no ceiling effect and floor effect.The Spearman correlation coefficients of SF-6D_CN,SF-6D_AU,SF-6D_HK with EQ-5D-3L_CN were 0.44,0.51,and 0.46,respectively.Bland-Altman plot analysis showed that the average utility differences between SF-6D_CN,SF-6D_AU,SF-6D_HK with EQ-5D-3L_CN were 0.208,0.215 and 0.135,respectively;The 95%agreement intervals were[-0.062,0.478],[-0145,0.577]and[-0.092,0.363].SF-6D_CN,SF-6D_AU,SF-6D_HK and EQ-5D-3L_CN can effectively distinguish respondents with different demographic characteristics,and the difference between groups was statistically significant（P<0.05）,and the absolute value of ES between adjacent groups was greater than 0.20;In terms of relative validity,the RV of SF-6D_CNin different gender,age group,marital status,educational level,and employment status was greater than 1,while EQ-5D-3L_CN only had higher discriminant validity in different family annual average income levels than SF-6D_CN,SF-6D_AU,and SF-6D_HK.SF-6D_CN,SF-6D_AU,SF-6D_HK and EQ-5D-3L_CN can effectively distinguish people with different health statuses.SF-6D_CN was better discriminate ability than EQ-5D-3L in subjective health status,chronic disease status,discomfort or outpatients within two weeks,emotional status,and QOL levels（RV>1）.In the EQ-5D-3L fully health states population,the discriminant validity of SF-6D_CN was also higher than that of SF-6D_AU,and SF-6D_HK.Conclusion:The SF-6D health utility value set based on the preference of Chinese general population is constructed by using the discrete choice experiment method,which has high consistency with the SF-6D health utility value sets of other countries and regions,and has higher discriminant validity than EQ-5D-3L in the health utility evaluation of Chinese general population.Discrete choice experiment is feasible in the development of health utility value set,and can be applied in other multi-attribute utility instruments.