Advertisement

Valuing Health State: An EQ-5D-5L Value Set for Ethiopians

Open ArchivePublished:November 02, 2019DOI:https://doi.org/10.1016/j.vhri.2019.08.475

      Highlights

      • There is a growing interest in health technology assessment and economic evaluation in low-resource settings such as Ethiopia.
      • This study developed a value set for the EQ-5D-5L using an Ethiopian general population sample with the EuroQol Group–Portable Valuation Technology (EQ-PVT) protocol administered in Amharic.
      • The new value set provides local users with societal preferences that are relevant.
      • The study also established the feasibility of using the less resource-intensive EQ-PVT, which is relevant for future studies.

      Abstract

      Objectives

      There is a growing interest in health technology assessment and economic evaluations in developing countries such as Ethiopia. The objective of this study was to derive an EQ-5D-5L value set from the Ethiopian general population to facilitate cost utility analysis.

      Methods

      A nationally representative sample (N = 1050) was recruited using a stratified multistage quota sampling technique. Face-to-face, computer-assisted interviews using the EuroQol Portable Valuation Technology (EQ-PVT) protocol of composite time trade-off (c-TTO) and discrete choice experiments (DCEs) were undertaken to elicit preference scores. The feasibility of the EQ-PVT protocol was pilot tested in a sample of the population (n = 110). A hybrid regression model combining c-TTO and DCE data was used to estimate the final value set.

      Results

      In the pilot study, the acceptability of the tasks was good, and there were no special concerns with undertaking the c-TTO and DCE tasks. The coefficients generated from a hybrid model were logically consistent. The predicted values for the EQ-5D-5L ranged from −0.718 to 1. Level 5 anxiety/depression had the largest impact on utility decrement (−0.458), whereas level 5 self-care had the least impact (−0.222). The maximum predicted value beyond full health was 0.974 for the 11112 health state.

      Conclusions

      This is the first EQ-5D-5L valuation study in Africa using international valuation methods (c-TTO and DCE) and also the first using the EQ-PVT protocol to derive a value set. We expect that the availability of this value set will facilitate health technology assessment and health-related quality-of-life research and inform policy decision making in Ethiopia.

      Keywords

      Introduction

      Globally, countries with publicly financed health services struggle to provide universal health coverage. The process of deciding which healthcare technologies and interventions to invest in has thus become increasingly important. This problem is greater in developing countries because they have very limited healthcare budgets.
      • Chalkidudou K.
      • Levine R.
      • Dillon A.
      Helping poorer countries make locally informed health.
      As a result, there is a growing interest in health technology assessment (HTA) in developing countries.
      • Attieh R.
      • Gagnon M.P.
      Implementation of local/hospital-based health technology assessment initiatives in low- and middle-income countries.
      • Dankó D.
      Health technology assessment in middle-income countries: recommendations for a balanced assessment system.
      • Mueller D.
      • Govender B.
      • Basu D.
      Health technology assessment in South Africa—future promise Launch of the Southern African Human Genome Programme.
      An important aspect of HTA is economic evaluation, which involves comparison of 2 or more interventions in terms of costs and outcomes.
      • Attieh R.
      • Gagnon M.P.
      Implementation of local/hospital-based health technology assessment initiatives in low- and middle-income countries.
      • Towse A.
      • Devlin N.
      • Hawe E.
      • Garrison L.
      The Evolution of HTA in Emerging Markets Health Care Systems: Analysis to Support a Policy Response.
      • Babigumira J.B.
      • Jenny A.M.
      • Bartlein R.
      • Stergachis A.
      • Garrison L.P.
      Health technology assessment in low- and middle-income countries: a landscape assessment.
      • Rajan A.
      • Gutierrez-Ibarluezea I.
      • Moharra M.
      Addressing issues in health technology assessment promotion: motives, enablers, and barriers.
      Although economic evaluation can be undertaken using natural outcomes such as life-years saved, these are limited because they capture only 1 aspect of health (eg, mortality). Quality-adjusted life-years (QALYs) combine life-years (mortality) with a quality adjustment that reflects the health-related quality of life (morbidity) on a 0 (dead) to 1 (full health) utility scale.
      • Weinstein M.C.
      • Torrance G.
      • Mcguire A.
      QALYs: the basics.
      QALYs are the recommended health outcome in economic evaluation in guidelines from the United States, Canada, United Kingdom, and many European countries.
      • Brockis E.
      • Marsden G.
      • Cole A.
      • Devlin N.
      A Review of NICE Methods Across Health Technology Assessment Programmes: Differences, Justifications and Implications.
      • Angelis A.
      • Lange A.
      • Kanavos P.
      Using health technology assessment to assess the value of new medicines: results of a systematic review and expert consultation across eight European countries.
      • Weinstein M.
      • Kaiser H.J.
      Cost-per-QALY in the US and Britain: Damned If You do and Damned If You Don’t.
      Canadian Agency for Drugs and Technologies in Health
      Guidelines for the Economic Evaluation of Health Technologies: Canada.
      Quality adjustment for QALYs relies on preferences for different health states derived using scaling methods such as the time trade-off (TTO) and standard gamble.
      • Torrance G.W.
      Measurement of health state utilities for economic appraisal: a review.
      Preference elicitation can be time-consuming and costly if undertaken independently for each study. Therefore, indirect preference-based instruments such as the EQ-5D have been developed. These measures include a descriptive system (eg, EQ-5D has 5 dimensions with 3 or 5 levels of severity) and a tariff that can be applied to generate utility values. The tariff is based on a valuation in a representative sample of the community using methods such as TTO. For specific measurement applications, patients can complete the descriptive system; community-derived tariffs are applied to generate the relevant utility values.
      • van Reenen M.
      • Janssen B.
      EQ-5D-5L User Guide: Basic Information on How to Use the EQ-5D-5L Instrument.
      A number of countries have generated country-specific value sets for the EQ-5D-5L.
      • Devlin N.
      • Shah K.
      • Mulhern B.
      • Feng Y.
      • Tsuchiya A.
      • van Hout B.A.
      Valuing health-related quality of life: an EQ-5D-5L value set for England.
      • Ramos-Goñi J.M.
      • Craig B.M.
      • Oppe M.
      • et al.
      Handling data quality issues to estimate the Spanish EQ-5D-5L value set using a hybrid interval regression approach.
      • Augustovski F.
      • Irazola L.R.V.
      • Morales M.
      • Gibbons L.
      • Manuel J.
      • Ramos-Goñi J.M.
      An EQ-5D-5L value set based on Uruguayan population preferences.
      • Luo N.
      • Liu G.
      • Li M.
      • Guan H.
      • Jin X.
      • Rand-Hendriksen K.
      Estimating an EQ-5D-5L value set for China.
      • Xie F.
      • Nick B.
      • Stirling B.
      • et al.
      A time trade-off-derived value set of the EQ-5D-5L for Canada.
      • Purba F.D.
      • Hunfeld J.A.M.
      • Iskandarsyah A.
      • et al.
      The Indonesian EQ-5D-5L value set.
      • Shiroiwa T.
      • Ikeda S.
      • Noto S.
      • Igarashi A.
      Comparison of value set based on DCE and/or TTO data: scoring for EQ-5D-5L health states in Japan.
      • Kim S.
      • Ahn J.
      • Ock M.
      • et al.
      The EQ-5D-5L valuation study in Korea.
      • Pattanaphesaj J.
      • Thavorncharoensap M.
      • Ramos J.M.
      • Tongsiri S.
      • Ingsrisawang L.
      • Teerawattananon Y.
      The EQ-5D-5L valuation study in Thailand.
      • Ludwig K.
      • Graf von der Schulenburg J.M.
      • Greiner W.
      • Ludwig K.
      German value set for the EQ-5D-5L.
      • Versteegh M.M.
      • Vermeulen K.M.
      • Evers S.M.A.A.
      • De Wit G.A.
      • Prenger R.
      • Stolk E.A.
      Dutch tariff for the five-level version of EQ-5D.
      Given the variation in culture and provision of health and social care, preferences are likely to vary across different populations. It is therefore preferred that economic evaluations should use locally derived value sets.
      • van Reenen M.
      • Janssen B.
      EQ-5D-5L User Guide: Basic Information on How to Use the EQ-5D-5L Instrument.
      In the context of developing countries, the presence of local value sets could facilitate economic evaluation research and can improve the quality of regulatory, coverage, and reimbursement policy decisions in individual and public healthcare.
      • Mueller D.
      • Govender B.
      • Basu D.
      Health technology assessment in South Africa—future promise Launch of the Southern African Human Genome Programme.
      • Rencz F.
      • Gulácsi L.
      • Drummond M.
      • et al.
      EQ-5D in Central and Eastern Europe: 2000–2015.
      • Doaee S.H.
      • Olyaeemanesh A.
      • Emami S.H.
      • Mobinizadeh M.
      • Abooee P.
      • Nejati M.
      Development and implementation of health technology assessment: a policy study.
      • Oortwijn W.
      • Mathijssen J.
      • Banta D.
      The role of health technology assessment on pharmaceutical reimbursement in selected middle-income countries.
      For example, a recent review in Central Europe found more HTA activities in countries with EQ-5D studies, although the direction of causality was not certain.
      • Rencz F.
      • Gulácsi L.
      • Drummond M.
      • et al.
      EQ-5D in Central and Eastern Europe: 2000–2015.
      The Ethiopian Federal Ministry of Health has introduced a public health insurance system and aims to support healthcare financing with principles of HTA.
      • Ali E.E.
      Health care financing in Ethiopia: implications on access to essential medicines.
      The current practice of HTA for regulatory and formulary development is based on expert opinion and published HTA studies from other countries.
      • Babigumira J.B.
      • Jenny A.M.
      • Bartlein R.
      • Stergachis A.
      • Garrison L.P.
      Health technology assessment in low- and middle-income countries: a landscape assessment.
      Although context-specific HTA is needed, local value sets are not yet available. The aim of this study was therefore to obtain an EQ-5D-5L value set from the Ethiopian general population.

      Methods

       Sampling Method and Study Population

      Based on a multicountry pilot study, the recommended total sample size for EQ-5D-5L valuation studies is 1000 participants per country. The total number of health states included for the composite TTO (c-TTO) is 86 and 196 pairs of health state valuation for the discrete choice experiment (DCE) task.
      • Oppe M.
      • van Hout B.
      EuroQol Working Paper Series: Experimental Design of the EQ-VT.
      A representative sample of 1050 respondents was recruited based on multistage stratified quota sampling of geographic area/residence (urban or rural), gender (male or female), age group (18-24, 24-54, 55-64, or >65 years), and religion (Christian, Islam, or other).
      Respondents interviewed in this study were living in Addis Ababa city and Butajira rural area, Southern Ethiopia. These 2 areas were selected because they have mixed populations in terms of ethnicity and culture owing to migration from other regions.
      • Berhane Y.
      • Byass P.
      • Butajira D.S.S.
      Population and health in developing countries.
      • Spaliviero M.C.F.
      • Cheru F.
      The State of Addis Ababa 2017: the Addis Ababa we want.
      In Addis Ababa city, supervisors identified the center of each the 10 subcities and then randomly identified the direction in which households would be identified from this point. One member of each identified household fulfilling the inclusion criteria was interviewed. Interviews were then conducted in other households in that same direction until the required quota sample was obtained. The Demographic Health Surveillance Site list for the Butajira 9 rural area (n = 15 000 households) was used to identify households using their unique household number. The supervisors were responsible for selecting the households from the list based on the quotas that needed to be filled. Supervisors also used personal networks from previous Demographic Health Surveillance Site surveys to mediate contact between the interviewers and respondents. Ethical approval was secured from the Ethics Review Committee of School of Pharmacy at Addis Ababa University, Ethiopia, and prior permission was sought from Demographic Health Surveillance site of Butajira Rural Health Project office, Ethiopia. Written Informed consent was obtained from all study participants to confirm their willingness for participation after explaining the purpose of study.
      The inclusion criteria were (1) ≥18 years old, (2) ability to understand the task (as judged by the interviewer), and (3) ability to give informed consent. Participants with the presence of any illness or cognitive impairment (confirmed and/or interviewer judgement) that would interfere with the study task were excluded. All respondents provided written informed consent.

       Outcome Measure

      The EQ-5D-5L instrument is a generic, multiattribute utility-based health status tool developed by the European quality of life (EuroQol) Group.
      • Brauer C.A.
      • Rosen A.B.
      • Greenberg D.
      • Neumann P.J.
      Trends in the measurement of health utilities in published cost-utility analyses.
      It consists of 5 dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression.
      • Brooks R.
      EuroQol: the current state of play.
      The EQ-5D initially included 3-level response options (no problems, some/moderate problems, extreme problems/unable to/confined to bed) under each dimension but in 2009 was expanded to 5 levels (no problems, slight, moderate, severe, extreme/unable). The EQ-5D-5L therefore includes 3125 health states (see Towse et al
      • Towse A.
      • Devlin N.
      • Hawe E.
      • Garrison L.
      The Evolution of HTA in Emerging Markets Health Care Systems: Analysis to Support a Policy Response.
      ). The change has improved the instrument’s sensitivity and reduced ceiling effects.
      • Herdman M.
      • Gudex C.
      • Lioyd C.
      • et al.
      Development and preliminary testing of the new five-level version of EQ-5D (EQ-5D-5L).
      The EQ-5D-5L has been translated to Amharic, the national language used in Ethiopia, using the standardized approach recommended by the EuroQol group.
      • van Reenen M.
      • Janssen B.
      EQ-5D-5L User Guide: Basic Information on How to Use the EQ-5D-5L Instrument.

       Valuation Techniques

      This study was a population-based, interviewer-administered, face-to-face, cross-sectional survey following a standardized valuation study protocol developed specifically for EQ-5D-5L value set studies.
      • Oppe M.
      • Devlin N.J.
      • Van Hout B.
      • Krabbe P.F.M.
      • De Charro F.
      A program of methodological research to arrive at the new international EQ-5D-5L valuation protocol.
      The data were collected using laptops installed with the EuroQol Portable Valuation Technology (EQ-PVT) software using 2 elicitation techniques: c-TTO and DCE. A combination of both the c-TTO and DCE method has been used in previous EQ-5D valuation studies.
      • Devlin N.
      • Shah K.
      • Mulhern B.
      • Feng Y.
      • Tsuchiya A.
      • van Hout B.A.
      Valuing health-related quality of life: an EQ-5D-5L value set for England.
      • Luo N.
      • Liu G.
      • Li M.
      • Guan H.
      • Jin X.
      • Rand-Hendriksen K.
      Estimating an EQ-5D-5L value set for China.
      • Purba F.D.
      • Hunfeld J.A.M.
      • Iskandarsyah A.
      • et al.
      The Indonesian EQ-5D-5L value set.
      • Pattanaphesaj J.
      • Thavorncharoensap M.
      • Ramos J.M.
      • Tongsiri S.
      • Ingsrisawang L.
      • Teerawattananon Y.
      The EQ-5D-5L valuation study in Thailand.
      • Versteegh M.M.
      • Vermeulen K.M.
      • Evers S.M.A.A.
      • De Wit G.A.
      • Prenger R.
      • Stolk E.A.
      Dutch tariff for the five-level version of EQ-5D.
      The c-TTO involves separate tasks for states considered better than dead and for states worse than dead. For states better than dead, respondents are asked their preference between living in poor health for 10 years and living in full health for a shorter period. The length of time lived in full health is varied until the respondent is indifferent between the 2 options. For states worse than dead, participants complete a lead-time TTO in which they choose between 10 years of full health and the alternative 20 years, which consists of 10 years spent in full health followed by 10 years in an impaired health state.
      • Devlin N.J.
      • Tsuchiya A.K.I.
      • Buckingham K.E.N.
      • Tilling C.
      A uniform time trade off method for states better and worse than dead: feasibility study of the ‘lead time’ approach.
      After completing the TTO tasks, participants provided feedback on the ranking of the states they valued by flagging any states that were in the wrong order.
      DCE involves a set of dichotomous choices over 2 multiattribute poor health states. The 2 methods generate different and complementary preference data. TTO elicits a value for each state, with 1 and 0 defined as anchor points with worse than dead bounded at −1. The DCE approach generates binary data, which allow for the derivation of a scale of nonanchored relative values.
      • Ramos-Goñi J.M.
      • Craig B.M.
      • Oppe M.
      • et al.
      Handling data quality issues to estimate the Spanish EQ-5D-5L value set using a hybrid interval regression approach.
      DCE is an approach that is increasingly used to assess preferences for health states because of the relative simplicity of the tasks.
      • Lancsar E.
      • Louviere J.
      Conducting discrete choice experiment to inform healthcare decision making.

       Data Collection and Eliciting Preferences Methods

      An EQ-VT version 2.1
      • Stolk E.
      • Ludwig K.
      • Rand K.
      • Van Hout B.
      • Ramos-Go J.M.
      Overview, update, and lessons learned from the international EQ-5D-5L valuation work: version 2 of the EQ-5D-5L valuation protocol.
      was used to support computer-aided data collection. The EQ-PVT is a portable version of EuroQol Valuation Technology (EQ-VT), which allows preference data to be collected without requiring direct links to the EuroQol group software, which is necessary for EQ-VT. It runs using similar algorithms to EQ-VT, but these are executed via a program in Microsoft PowerPoint. Data are stored on the computer and can then be uploaded to secure sites when Internet links are available.
      The interviews were undertaken by 10 trained pharmacy master’s degree students from March to May 2018. A 3-day training course on the methodology and study procedures was provided to the interviewers. The training material from the EuroQol group was translated into Amharic. The training material included (1) an introduction to related concepts such as health-related quality of life and the EQ-5D-5L as a generic questionnaire used to value health states, (2) an explanation of the EQ-PVT protocol and interviewer instructions, (3) practice in groups, and (4) pretesting of the tools to ensure that interviewers understood the task.
      During the data collection, all data collectors traveled in 1 group to the study sites. Regular supervision was done on each data collection day by the trained principal investigators (authors A.G.W. and G.B.G.).
      Each interview consisted of a paper-based and a computer-assisted task. The paper-based task included basic background questions, self-reported health measurement using the 5-item EQ-5D-5L, and the companion visual analog scale ranging from 0 (the worst health you can imagine) to 100 (the best health you can imagine).
      The computer-assisted interview task included a practice TTO task (valuation of a mild, moderate, and severe health state). In addition, based on the pilot, participants were asked to provide an example of a better and worse health state than being confined to a wheelchair; where this was not possible, wearing glasses or being confined to bed were provided as examples, respectively. This was followed by 10 TTO tasks and 7 DCE tasks. There were 86 TTO health states in 10 blocks, with each block containing a range of health states across the range of severity from mild to severe as well as the worst health state (55555). Health states were presented randomly within the block. There were 7 DCE pairs in 28 blocks, which were also presented randomly. One block from the TTO and DCE was randomly assigned to each participant.

       Quality Control

      We used a cyclic quality control (QC) process employed using the EQ-PVT QC tool developed by the EuroQol group.
      • Ramos-Goñi J.M.
      • Craig B.M.
      • Oppe M.
      • et al.
      Handling data quality issues to estimate the Spanish EQ-5D-5L value set using a hybrid interval regression approach.
      The EuroQol group’s expert and the Ethiopia team organized Skype-based meetings every 20 interviews for the pilot study and every 100th interview for the main study to discuss the QC reports with the EQ-PVT support team. The QC process consisted of an evaluation of protocol compliance of 40% threshold as cutoff points to assess interviewer effects.
      • Ramos-Goñi J.M.
      • Oppe M.
      • Slaap B.
      • Busschbach J.J.
      • Stolk E.
      Quality control process for EQ-5D-5L valuation studies.
      The QC reports provided a number of statistics related to the quality of the data collected, differentiated by interviewer. Criteria used to evaluate the quality of the interviews included the following:
      • wheelchair time, when the duration of time an interviewer used to explain the “wheelchair example” preceding the actual c-TTO tasks was less than 3 minutes;
      • wheelchair lead-time, when the interviewer did not explain the “worse than dead” element of the wheelchair or worse than wheelchair example;
      • c-TTO duration, if completing the 10 c-TTO tasks took less than 5 minutes; and
      • inconsistency, when the value for state “55555” was not the lowest and it was at least 0.5 higher than that of the state with the lowest value.
      If any of the 4 aforementioned criteria were met, the interview was flagged so that further discussions about the interview process could be undertaken. The distribution of data obtained from different interviewers was also reviewed, and any anomalies such as spikes at critical points (1, 0.5, 0, −0.5, −1) or other differences were highlighted for further discussion. Spikes may indicate poor engagement at the interviewer level, whereas distributions that are different may indicate interviewer effects.

       Pilot Study

      The feasibility of the valuation protocol for EQ-5D-5L valuation studies using the EQ-PVT in Ethiopia was pilot tested in a sample of the population recruited from Addis Ababa. Two Ethiopian principal investigators who were trained in The Netherlands completed 110 interviews in the pilot study from January to February 2018 with QC support. The pilot sample was recruited using national strata on age, gender, and religion. Interviewers paid attention to the understanding of the tasks, level of engagement, acceptability of thinking or talking about death, and acceptability of EQ-5D-5L health states. The concordance between c-TTO and DCE values was also investigated. The data from the pilot study were not part of the final data set.
      The QC process in the pilot identified issues related to the wheelchair example and presenting lead time TTO, and these were discussed and highlighted when training the other interviewers. The acceptability of the tasks was good, and there were no special concerns with thinking and talking about death. Nevertheless, a frequently reported problem was that respondents found health states difficult to imagine. To resolve this issue, practice tasks were tailored to the local context by providing examples that were more meaningful (ie, wearing glasses and confined to bed). Preliminary analysis of the pilot data indicated that DCE data were fully consistent but there were inconsistencies in TTO data, but this was not considered problematic as sample sizes were small.

       Data Analysis

      Statistical analysis was undertaken using the STATA 14.2 statistical package. Descriptive statistics were used to summarize respondents’ characteristics and responses to the c-TTO and DCE tasks, and the relative importance of each attribute was evaluated using logistic regression methods. We used a method developed by Ramos-Goñi et al
      • Ramos-Goñi J.M.
      • Craig B.
      • Oppe M.
      • van Hout B.
      Introducing a hybrid model that combines continuous and dichotomous responses in a single maximum likelihood function: the hyreg command.
      to model value sets. A variety of preference models was produced using both types of data (c-TTO and DCE) individually and together to provide complementary evidence on preferences by taking into account the nature of preference data that are “bounded” (censored), the heterogeneity of respondents’ views in health utilities, and the heteroskedasticity of the error terms. For c-TTO, several models were tested including ordinary least squares, generalized linear models, random coefficient models, and Tobit models, taking into account the panel structure of the data. The dependent variable in the c-TTO part of the model was disutility (defined as 1 minus the c-TTO observed values) for a given health state.
      DCE data were analyzed using the likelihood function of a conditional logit distribution. In the DCE model, the dependent variable was a binary outcome (0/1) indicating the respondent’s choice for each pair of EQ-5D-5L states. Because the coefficients estimated from a conditional logit are expressed on a latent arbitrary utility scale, a rescaled parameter was used, which assumes that the c-TTO model coefficients are proportional to DCE model coefficients. The coefficients represented the utility decrements for the DCE rescaled model.
      • Ramos-Goñi J.M.
      • Craig B.
      • Oppe M.
      • van Hout B.
      Introducing a hybrid model that combines continuous and dichotomous responses in a single maximum likelihood function: the hyreg command.
      Continuous responses from the c-TTO and dichotomous responses from the DCE were combined in a single model using the “hyreg” command developed by Ramos-Goñi et al
      • Ramos-Goñi J.M.
      • Craig B.
      • Oppe M.
      • van Hout B.
      Introducing a hybrid model that combines continuous and dichotomous responses in a single maximum likelihood function: the hyreg command.
      to undertake hybrid models. This command allows the continuous and dichotomous responses to have different distributions (logistic and normal) and different independent variables to model scaling terms. Because the variance of c-TTO data is not homogenous, a heteroskedasticity model was estimated in which c-TTO responses were censored at −1.
      A main effects 20-parameter model consisting of 4 dummies for each EQ-5D-5L dimension was explored using level 1 as the reference. Dummies were constructed to represent the additional utility decrement of moving from one level to another. For instance, the mobility dimension had 4 dummies: MO2 to MO5. The coefficient associated with MO2 indicated the utility decrement of moving from no problems (level 1) to slight problems (level 2), MO3 the additional utility decrement of moving from slight (level 2) to moderate (level 3) problems, and so on. Therefore, the overall decrement of moving from no to severe problems could be calculated as the sum of the coefficients of MO2 to MO5. The same set of dummy variables was defined for each of the remaining dimensions: self-care, usual activities, pain/discomfort, and anxiety/depression.
      Some TTO data were excluded, including data from (1) participants who gave all 10 health states the same value and (2) participants who gave the worst state, 55555, a value that was no lower than the value they gave to the mildest health state in their block.

       Evaluation of Model Performance

      Model performance was evaluated using logical consistency of parameters, goodness of fit, and significance level. Estimated coefficients were said to be logically consistent if the estimated coefficients of the parameters were positive and when the values of the estimated coefficients’ magnitude of logically worse health states were lower than those from logically better health states. Goodness of fit was assessed using the Akaike and the Bayesian information criteria. Finally, the level of significance was assessed based on the P value of the models. Models that met the performance criteria are reported.

      Results

       Respondent Characteristics

      Ten interviewers completed 1050 interviews. No interviewers were excluded from the study; Figure 1 shows the distribution of observed TTO values by interviewer. A total of 1041 responses for the TTO and 1048 for the DCE task formed the sample for analysis. Nine participants were dropped from the TTO task in the main survey because they were nontraders and 2 from the DCE task because of overlap of participant ID during the data collection. The characteristics of sample respondents were similar to the Ethiopian general population in terms of residence, age, gender, and religion (Table 1).
      Table 1Background characteristics of the sample and general population.
      CharacteristicStudy sample (N = 1050), n (%)Ethiopian general population, %
      Data obtained from the 2016 Ethiopian health survey.
      Residence
       Urban300 (28.57)19.92
       Rural750 (71.43)80.08
      Sex
       Female503 (47.90)50.15
       Male547 (52.10)49.85
      Age (in years)
       18-24391 (37.24)36.00
       25-54575 (54.76)52.32
       55-6449 (4.67)6.91
       65+35 (3.33)5.20
      Marital status
       Married559 (53.24)NA
       Unmarried444 (42.23)NA
       Divorced22 (2.10)NA
       Widowed25 (2.40)NA
      Education
       No formal education440 (41.90)NA
       Primary school173 (16.48)NA
       Secondary school265 (25.33)NA
       Higher school171 (16.29)NA
      Employment status
       Employed or self-employed693 (66)NA
       Retired30 (2.86)NA
       Student214 (20.40)NA
       Looking after home or family111 (10.60)NA
       Other/ one of the above2 (0.19)NA
      Experience of serious illness
       In self142 (13.52)NA
       In family329 (31.33)NA
       In self and in family172 (16.36)NA
       In caring for others152 (14.48)NA
       No255 (24.28)NA
      Religion
       Christian679 (64.67)63
       Muslim368 (35.05)34
       Others3 (0.29)3.0
      NA indicates not available.
      Data obtained from the 2016 Ethiopian health survey.

       Self-Reported Health Problems

      As depicted in Table 2, the highest proportion of health problems was reported in the anxiety/depression dimension (43.29%) and the lowest proportion in the self-care dimension (6.57%). The mean self-reported EQ visual analog scale score was 87.27 (SD = 13.63).
      Table 2Self-reported health using the EQ-5D-5L descriptive system and the EQ visual analog scale.
      EQ-5D-5L descriptive system with scores in %
      ParameterMobilitySelf-careUsual activitiesPain/discomfortAnxiety/depression
      No problems88.1993.4383.3358.0956.38
      Slight problems7.614.7011.4327.4333.71
      Moderate problems4.001.614.6612.008.28
      Severe problems0.090.190.572.281.33
      Unable/extreme problems0.090.000.000.190.28
      MeanSD25th PercentileMedian75th Percentile
      Visual analog scale score87.2613.64759095

       Modeling Results

      Very few states were flagged in the TTO feedback module (0.5%), so these were retained in the analysis. There were 837 (8.04%) left-censored c-TTO observations, that is, those in which the respondent gave the lowest possible value (−1) for a health state in the c-TTO task with evidence of a spike at 1 (Fig. 1). An ordinary least squares model was used for c-TTO observation data, and all of the responses were logically consistent. The conditional logistic regression model was used to model the DCE responses. In the conditional logistic regression model, DCE data had 2 inconsistent (negative) values in self-care and usual activity dimensions of level 3.
      Figure thumbnail gr1
      Figure 1Distribution of observed TTO values of every interviewer and in total.
      TTO indicates time trade-off.
      Table 3 shows estimation results from the c-TTO, DCE, and hybrid model. The models’ set of coefficients were in relative agreement; that is, in the disutility amount, the most important dimensions were anxiety/depression and pain/discomfort, and the least important was self-care. The hybrid model parameters were logically consistent, and all parameters except 1 (pain and discomfort level 3) were statistically significant. There was relative agreement between the 20 parameters in the c-TTO, DCE, and hybrid models. The anxiety/depression dimension influenced utility estimates the most (disutility level 5 of 0.4578) and self-care influenced the estimates the least (disutility level 5 of 0.2224) in the final value set of the hybrid model. Figure 2 shows the association between the 3 models. The scatterplots of the different models suggest the compatibility of the TTO model and hybrid model and show the effect of adding the DCE data to the c-TTO valuation in the hybrid model.
      Table 3Estimation results for c-TTO model, DCE rescaled model, and hybrid model in incremental dummies.
      Items with a negative coefficient (in gray) represent inconsistent items. The order of importance is based on the sum of the disutility, which is the disutility associated with level 5.
      Independent variables of the modelc-TTO ordinary least squares modelDCE conditional logistic model rescaledHybrid model censored c-TTO values at −1 (final value set)
      CoefficientSEP valueCoefficientSEP valueCoefficientSEP value
      Mobility (MO)
       MO20.00470.014.7290.47800.061.0000.03370.005.000
       MO30.01660.015.2620.11380.071.1100.03070.009.000
       MO40.17480.016.0000.98100.070.0000.16320.010.000
       MO50.10380.016.0000.74340.074.0000.13220.010.000
      Self-care (SC)
       SC20.00360.013.7850.20440.067.0020.02350.005.000
       SC30.04940.016.002−0.00240.074.9740.01600.008.042
       SC40.11890.015.0000.68490.078.0000.10240.009.000
       SC50.08260.013.0000.42340.073.0000.08040.009.000
      Usual activities (UA)
       UA20.01880.014.1760.34700.063.0000.03230.005.000
       UA30.04410.014.002−0.03910.071.5790.01600.008.042
       UA40.12990.016.0000.58180.071.0000.10910.009.000
       UA50.09360.015.0000.60790.076.0000.11470.010.000
      Pain/discomfort (PD)
       PD20.01400.013.2660.44990.067.0000.03610.004.000
       PD30.01610.017.3310.10900.073.1360.01550.008.061
       PD40.24520.015.0001.13580.077.0000.21870.010.000
       PD50.14210.016.0000.56890.076.0000.13610.011.000
      Anxiety/depression (AD)
       AD20.01110.014.4280.27180.070.0000.02590.004.000
       AD30.03810.015.0120.35160.072.0000.05890.008.000
       AD40.23220.015.0001.18030.079.0000.21390.009.000
       AD50.14140.013.0000.83200.078.0000.15910.010.000
      Akaike information criterion10 587.066498.3014 002.09
      Bayesian information criterion10 739.336650.1714 336.81
      Order of importance
      ADADAD
      PDMOPD
      MOPDMO
      UAUAUA
      SCSCSC
      c-TTO indicates composite time trade-off; DCE, discrete choice experiment; SE, standard error.
      Items with a negative coefficient (in gray) represent inconsistent items. The order of importance is based on the sum of the disutility, which is the disutility associated with level 5.
      Figure thumbnail gr2
      Figure 2The scatterplots show the association between the three models of c-TTO and DCE rescaled predicted utilities, c-TTO and hybrid predicted utilities, DCE rescaled and hybrid predicted utilities, respectively.
      c-TTO indicates composite time trade-off; DCE, discrete choice experiment.
      The observed value of utility ranged from −0.718 for state 55555 to 0.974 for states 11112 in the hybrid model, which is preferred. Values for 11121 and 11122 were found to be 0.964 and 0.938, respectively. The mean observed value was negative for 10 of 86 states that were included in the design.
      In this sample, the mean (standard deviation) EQ-5D-5L values were 0.94 (0.10) based on the hybrid model. To calculate utility values, the sum of the utility decrements for the relevant level (eg, for level 3 this would be the sum of level 2 and level 3) health state 11113 is given by 1 – (0.0259 + 0.0589) = 0.915.

      Discussion

      This study presents social preferences and an EQ-5D-5L value set from the Ethiopian general population. To obtain values attached to 3125 EQ-5D-5L health states, 1050 respondents were interviewed using the computer-assisted valuation protocol EQ-PVT with extensive interviewer training and data inspection. The intensive QC ensured high data quality in terms of few inconsistencies, little clustering of values, and low interviewer effect. The sociodemographic characteristics of the respondents were similar to that of the general population with respect to geographic area/residence, gender, age group, and religion. This makes the EQ-5D-5L suitable for health economic evaluations that will benefit national healthcare financing with the principle of HTA.
      Value sets have been generated based on c-TTO data and DCE data alone. The final value set reported here is derived from a hybrid of both c-TTO and DCE data because the 2 types of data provide different but complementary information about the views of the respondents. The hybrid model maximizes data usage, and it gives the highest validity on parameter estimation because c-TTO measures utilities trading off poor health against time, and the DCE task asked respondents to trade-off between 2 poor health states. A hybrid model was also used by different studies
      • Devlin N.
      • Shah K.
      • Mulhern B.
      • Feng Y.
      • Tsuchiya A.
      • van Hout B.A.
      Valuing health-related quality of life: an EQ-5D-5L value set for England.
      • Ramos-Goñi J.M.
      • Craig B.M.
      • Oppe M.
      • et al.
      Handling data quality issues to estimate the Spanish EQ-5D-5L value set using a hybrid interval regression approach.
      • Purba F.D.
      • Hunfeld J.A.M.
      • Iskandarsyah A.
      • et al.
      The Indonesian EQ-5D-5L value set.
      • Shiroiwa T.
      • Ikeda S.
      • Noto S.
      • Igarashi A.
      Comparison of value set based on DCE and/or TTO data: scoring for EQ-5D-5L health states in Japan.
      • Pattanaphesaj J.
      • Thavorncharoensap M.
      • Ramos J.M.
      • Tongsiri S.
      • Ingsrisawang L.
      • Teerawattananon Y.
      The EQ-5D-5L valuation study in Thailand.
      • Ludwig K.
      • Graf von der Schulenburg J.M.
      • Greiner W.
      • Ludwig K.
      German value set for the EQ-5D-5L.
      to estimate the value set of EQ-5D-5L. Results from the hybrid model had logical consistency of parameters. All dimensions were statistically significant except for 1 parameter (pain and discomfort level 3). This EQ-5D-5L value set considered the complementary data in a hybrid model using an innovative model
      • Devlin N.
      • Shah K.
      • Mulhern B.
      • Feng Y.
      • Tsuchiya A.
      • van Hout B.A.
      Valuing health-related quality of life: an EQ-5D-5L value set for England.
      • Ramos-Goñi J.M.
      • Serrano-Aguilar P.
      • Rivero-Arias O.
      • Cabase J.M.
      Valuation and modeling of EQ-5D-5L health states using a hybrid approach.
      that takes into account the fact that c-TTO data are left censored and the heteroskedasticity of the error terms to decrease biased parameter estimates.
      Similar to other countries such as The Netherlands
      • Versteegh M.M.
      • Vermeulen K.M.
      • Evers S.M.A.A.
      • De Wit G.A.
      • Prenger R.
      • Stolk E.A.
      Dutch tariff for the five-level version of EQ-5D.
      and England,
      • Devlin N.
      • Shah K.
      • Mulhern B.
      • Feng Y.
      • Tsuchiya A.
      • van Hout B.A.
      Valuing health-related quality of life: an EQ-5D-5L value set for England.
      the preferences of the Ethiopian population suggest that anxiety/depression and pain/discomfort are the health problems that are most important. On the other hand, mobility is the most important dimension in Indonesia,
      • Purba F.D.
      • Hunfeld J.A.M.
      • Iskandarsyah A.
      • et al.
      The Indonesian EQ-5D-5L value set.
      Korea,
      • Kim S.
      • Ahn J.
      • Ock M.
      • et al.
      The EQ-5D-5L valuation study in Korea.
      Japan,
      • Shiroiwa T.
      • Ikeda S.
      • Noto S.
      • Igarashi A.
      Comparison of value set based on DCE and/or TTO data: scoring for EQ-5D-5L health states in Japan.
      Canada,
      • Xie F.
      • Nick B.
      • Stirling B.
      • et al.
      A time trade-off-derived value set of the EQ-5D-5L for Canada.
      and Uruguay.
      • Augustovski F.
      • Irazola L.R.V.
      • Morales M.
      • Gibbons L.
      • Manuel J.
      • Ramos-Goñi J.M.
      An EQ-5D-5L value set based on Uruguayan population preferences.
      Although problems with self-care and usual activity dimensions were less important, dimensions with the smallest impact to the utility decrement varied among the countries. In Ethiopia, the maximum predicted value beyond full health occurred for the 11112 state (utility = 0.9741), whereas it was 11112 for Indonesia
      • Purba F.D.
      • Hunfeld J.A.M.
      • Iskandarsyah A.
      • et al.
      The Indonesian EQ-5D-5L value set.
      (utility = 0.921), 11211 and 12111 for England
      • Devlin N.
      • Shah K.
      • Mulhern B.
      • Feng Y.
      • Tsuchiya A.
      • van Hout B.A.
      Valuing health-related quality of life: an EQ-5D-5L value set for England.
      (utility = 0.950), and 11211 for China
      • Luo N.
      • Liu G.
      • Li M.
      • Guan H.
      • Jin X.
      • Rand-Hendriksen K.
      Estimating an EQ-5D-5L value set for China.
      (utility = 0.955).
      There were 837 (8.04%) observed −1 values on the c-TTO observations, whereby respondents gave the lowest possible value (−1) for a health state in the c-TTO task, which higher than that given in other published studies.
      • Devlin N.
      • Shah K.
      • Mulhern B.
      • Feng Y.
      • Tsuchiya A.
      • van Hout B.A.
      Valuing health-related quality of life: an EQ-5D-5L value set for England.
      • Ramos-Goñi J.M.
      • Craig B.M.
      • Oppe M.
      • et al.
      Handling data quality issues to estimate the Spanish EQ-5D-5L value set using a hybrid interval regression approach.
      • Luo N.
      • Liu G.
      • Li M.
      • Guan H.
      • Jin X.
      • Rand-Hendriksen K.
      Estimating an EQ-5D-5L value set for China.
      • Purba F.D.
      • Hunfeld J.A.M.
      • Iskandarsyah A.
      • et al.
      The Indonesian EQ-5D-5L value set.
      • Versteegh M.M.
      • Vermeulen K.M.
      • Evers S.M.A.A.
      • De Wit G.A.
      • Prenger R.
      • Stolk E.A.
      Dutch tariff for the five-level version of EQ-5D.
      In this study, the utility value ranged from −0.718 for state 55555 to 0.974 for state 11112. Compared with other studies that used the hybrid model, the worst health state had a higher value than Indonesia (−0.865)20 but was lower than in the English population (−0.285).
      • Devlin N.
      • Shah K.
      • Mulhern B.
      • Feng Y.
      • Tsuchiya A.
      • van Hout B.A.
      Valuing health-related quality of life: an EQ-5D-5L value set for England.
      This study has several strengths. It is the first study in Ethiopia and in Africa to report a value set for the EQ-5D-5L. The data have been generated using an international standardized protocol developed by the EuroQol group. The value set could also be potentially used by other African countries because although there are important cultural differences, these may be relatively smaller compared with other countries that have EQ-5D-5L tariffs. Furthermore, it provides evidence of the feasibility of valuing the EQ-5D-5L using the EQ-PVT, which uses PowerPoint. This provides evidence that the more cost-effective EQ-PVT approach can be used in low-resource settings. The EQ-PVT is also of advantage where reliable Internet connections are not available, such as in rural areas. The frequent QC meetings through Skype with the EuroQol group team during the course of the study strengthened the approach to high-quality data.
      Although TTO data have good face validity, a potential limitation of the study is that there is evidence of clustering of values at 1 (nontrading for mild health states). This occurred despite assessment of performance of interviewers and may suggest that in this context, participants were unwilling to trade-off time for what they consider to be mild states. There are also limitations in terms of differences in the distribution of background variables in the sample compared with the data provided by the National Bureau of Statistics. Nevertheless, these differences are small and limited to some residence and religion. The strategy of relying on supervisors to act as mediators between interviewers and the respondents based on personal networks may have introduced bias. A further investigation could be conducted to determine whether recruiting respondents via personal networks has an impact on data quality. A final limitation was that the data were collected only in Addis Ababa city and the Butajira rural area of the Gurage zone in the southern part of Ethiopia. This might raise questions about the representativeness of the study sample, but as noted, these areas were selected for their diverse population in terms of ethnicity and culture.

      Conclusions

      This study showed that it was feasible and culturally acceptable to estimate preferences for health states using the EQ-PVT software of EQ-5D-5L valuation. This study established an Ethiopian value set for EQ-5D-5L on the basis of c-TTO and DCE from the general population of Ethiopia. We expect that this work will serve as the foundation for applied health economic evaluation, lead to additional health preference studies, and inform decision making in Ethiopia.

      Acknowledgments

      The authors would like to thank all study participants for their participation. Special thanks to the EuroQol support team, especially Ramos-goñi JM, for providing support in the analysis of the valuation data, and Arnd Jan Prause and Kristina Lidwing, for their advice and support during the study period.
      This project received financial supported from Addis Ababa University , Ethiopia, and the EuroQol foundation, The Netherlands (project ID 20170480).

      Supplemental Material

      References

        • Chalkidudou K.
        • Levine R.
        • Dillon A.
        Helping poorer countries make locally informed health.
        Br Med J. 2010; 16: 1-4
        • Attieh R.
        • Gagnon M.P.
        Implementation of local/hospital-based health technology assessment initiatives in low- and middle-income countries.
        Int J Technol Assess Health Care. 2012; 28: 44451
        • Dankó D.
        Health technology assessment in middle-income countries: recommendations for a balanced assessment system.
        J Mark Access Health Policy. 2014; 2: 1-10
        • Mueller D.
        • Govender B.
        • Basu D.
        Health technology assessment in South Africa—future promise Launch of the Southern African Human Genome Programme.
        South African Med J. 2011; 101: 285-286
        • Towse A.
        • Devlin N.
        • Hawe E.
        • Garrison L.
        The Evolution of HTA in Emerging Markets Health Care Systems: Analysis to Support a Policy Response.
        Office of Health Economics, London, England2011
        • Babigumira J.B.
        • Jenny A.M.
        • Bartlein R.
        • Stergachis A.
        • Garrison L.P.
        Health technology assessment in low- and middle-income countries: a landscape assessment.
        J Pharm Health Serv Res. 2016; 7: 37-42
        • Rajan A.
        • Gutierrez-Ibarluezea I.
        • Moharra M.
        Addressing issues in health technology assessment promotion: motives, enablers, and barriers.
        Int J Technol Assess Health Care. 2011; 27: 55-63
        • Weinstein M.C.
        • Torrance G.
        • Mcguire A.
        QALYs: the basics.
        Value Health. 2009; 12: 5-9
        • Brockis E.
        • Marsden G.
        • Cole A.
        • Devlin N.
        A Review of NICE Methods Across Health Technology Assessment Programmes: Differences, Justifications and Implications.
        Office of Health Economics, London, England2016
        • Angelis A.
        • Lange A.
        • Kanavos P.
        Using health technology assessment to assess the value of new medicines: results of a systematic review and expert consultation across eight European countries.
        Eur J Health Econ. 2018; 19: 123-152
        • Weinstein M.
        • Kaiser H.J.
        Cost-per-QALY in the US and Britain: Damned If You do and Damned If You Don’t.
        Office of Health Economics, London, England2011
        • Canadian Agency for Drugs and Technologies in Health
        Guidelines for the Economic Evaluation of Health Technologies: Canada.
        4th ed. Canadian Agency for Drugs and Technologies in Health, Ottawa, Ontario, Canada2017
        • Torrance G.W.
        Measurement of health state utilities for economic appraisal: a review.
        J Health Econ. 1986; 5: 1-30
        • van Reenen M.
        • Janssen B.
        EQ-5D-5L User Guide: Basic Information on How to Use the EQ-5D-5L Instrument.
        EuroQol Research Foundation, Rotterdam, The Netherlands2015
        • Devlin N.
        • Shah K.
        • Mulhern B.
        • Feng Y.
        • Tsuchiya A.
        • van Hout B.A.
        Valuing health-related quality of life: an EQ-5D-5L value set for England.
        Health Econ. 2014; 27: 7-22
        • Ramos-Goñi J.M.
        • Craig B.M.
        • Oppe M.
        • et al.
        Handling data quality issues to estimate the Spanish EQ-5D-5L value set using a hybrid interval regression approach.
        Value Health. 2018; 21: 596-604
        • Augustovski F.
        • Irazola L.R.V.
        • Morales M.
        • Gibbons L.
        • Manuel J.
        • Ramos-Goñi J.M.
        An EQ-5D-5L value set based on Uruguayan population preferences.
        Qual Life Res. 2016; 25: 323-333
        • Luo N.
        • Liu G.
        • Li M.
        • Guan H.
        • Jin X.
        • Rand-Hendriksen K.
        Estimating an EQ-5D-5L value set for China.
        Value Health. 2017; 20: 662-669
        • Xie F.
        • Nick B.
        • Stirling B.
        • et al.
        A time trade-off-derived value set of the EQ-5D-5L for Canada.
        Med Care. 2016; 54: 98-105
        • Purba F.D.
        • Hunfeld J.A.M.
        • Iskandarsyah A.
        • et al.
        The Indonesian EQ-5D-5L value set.
        Pharmacoeconomics. 2017; 35: 1153-1165
        • Shiroiwa T.
        • Ikeda S.
        • Noto S.
        • Igarashi A.
        Comparison of value set based on DCE and/or TTO data: scoring for EQ-5D-5L health states in Japan.
        Value Health. 2016; 19: 648-654
        • Kim S.
        • Ahn J.
        • Ock M.
        • et al.
        The EQ-5D-5L valuation study in Korea.
        Qual Life Res. 2016; 25: 1845-1852
        • Pattanaphesaj J.
        • Thavorncharoensap M.
        • Ramos J.M.
        • Tongsiri S.
        • Ingsrisawang L.
        • Teerawattananon Y.
        The EQ-5D-5L valuation study in Thailand.
        Expert Rev Pharmacoecon Outcomes Res. 2018; 18: 551-558
        • Ludwig K.
        • Graf von der Schulenburg J.M.
        • Greiner W.
        • Ludwig K.
        German value set for the EQ-5D-5L.
        Pharmacoeconomics. 2018; 36: 663-674
        • Versteegh M.M.
        • Vermeulen K.M.
        • Evers S.M.A.A.
        • De Wit G.A.
        • Prenger R.
        • Stolk E.A.
        Dutch tariff for the five-level version of EQ-5D.
        Value Health. 2016; 19: 343-352
        • Rencz F.
        • Gulácsi L.
        • Drummond M.
        • et al.
        EQ-5D in Central and Eastern Europe: 2000–2015.
        Qual Life Res. 2016; 25: 2693-2710
        • Doaee S.H.
        • Olyaeemanesh A.
        • Emami S.H.
        • Mobinizadeh M.
        • Abooee P.
        • Nejati M.
        Development and implementation of health technology assessment: a policy study.
        Iran J Public Health. 2013; 42: 50-54
        • Oortwijn W.
        • Mathijssen J.
        • Banta D.
        The role of health technology assessment on pharmaceutical reimbursement in selected middle-income countries.
        Health Policy. 2010; 95: 174-184
        • Ali E.E.
        Health care financing in Ethiopia: implications on access to essential medicines.
        Value Health Reg Issues. 2014; 4: 37-40
        • Oppe M.
        • van Hout B.
        EuroQol Working Paper Series: Experimental Design of the EQ-VT.
        EuroQol Research Foundation, Rotterdam, The Netherlands2017
        • Berhane Y.
        • Byass P.
        • Butajira D.S.S.
        Population and health in developing countries.
        Department of community health, faculty of medicine Addis Ababa University and department of public health and clinical medicine UMEÅ University UNIVERSITY, 2002: 0-9 (Ethiopia. http://www.indepth-network.org/dss_site_profiles/butajira.pdf. Accessed October 29, 2019)
        • Spaliviero M.C.F.
        • Cheru F.
        The State of Addis Ababa 2017: the Addis Ababa we want.
        United Nations Human Settlements Programme (UN-Habitat), 2017 (https://openaccess.leidenuniv.nl/bitstream/handle/1887/58263/State_of_Addis_Ababa_2017_Report-web.pdf?sequence=1. Accessed October 29, 2019)
        • Brauer C.A.
        • Rosen A.B.
        • Greenberg D.
        • Neumann P.J.
        Trends in the measurement of health utilities in published cost-utility analyses.
        Value Health. 2006; 9: 213-218
        • Brooks R.
        EuroQol: the current state of play.
        Health Policy. 1996; 37: 53-72
        • Herdman M.
        • Gudex C.
        • Lioyd C.
        • et al.
        Development and preliminary testing of the new five-level version of EQ-5D (EQ-5D-5L).
        Qual Life Res. 2011; 20: 1727-1736
        • Oppe M.
        • Devlin N.J.
        • Van Hout B.
        • Krabbe P.F.M.
        • De Charro F.
        A program of methodological research to arrive at the new international EQ-5D-5L valuation protocol.
        Value Health. 2014; 17: 445-453
        • Devlin N.J.
        • Tsuchiya A.K.I.
        • Buckingham K.E.N.
        • Tilling C.
        A uniform time trade off method for states better and worse than dead: feasibility study of the ‘lead time’ approach.
        Health Econ. 2011; 20: 348-361
        • Lancsar E.
        • Louviere J.
        Conducting discrete choice experiment to inform healthcare decision making.
        Pharmacoeeconomics. 2008; 26: 661-677
        • Stolk E.
        • Ludwig K.
        • Rand K.
        • Van Hout B.
        • Ramos-Go J.M.
        Overview, update, and lessons learned from the international EQ-5D-5L valuation work: version 2 of the EQ-5D-5L valuation protocol.
        Value Health. 2019; 22: 23-30
        • Ramos-Goñi J.M.
        • Oppe M.
        • Slaap B.
        • Busschbach J.J.
        • Stolk E.
        Quality control process for EQ-5D-5L valuation studies.
        Value Health. 2017; 20: 466-473
        • Ramos-Goñi J.M.
        • Craig B.
        • Oppe M.
        • van Hout B.
        Introducing a hybrid model that combines continuous and dichotomous responses in a single maximum likelihood function: the hyreg command.
        EuroQol Research Foundation. 2016;
        • Ramos-Goñi J.M.
        • Serrano-Aguilar P.
        • Rivero-Arias O.
        • Cabase J.M.
        Valuation and modeling of EQ-5D-5L health states using a hybrid approach.
        Med Care. 2017; 55: e51-e58