Modifying the Composite Time Trade-Off Method to Improve Its Discriminatory Power.

OBJECTIVES: In cost-effectiveness analysis of health technologies, health state utilities are needed. They are often elicited with a composite time trade-off (cTTO) method, particularly for the widely used EQ-5D-5L. Unfortunately, cTTO discriminatory power is hindered by (1) respondents' nontrading (NT) of time for quality, (2) censoring of utilities at -1, and (3) poor correlation of negative utilities with state severity. We investigated whether modifying cTTO can mitigate these effects. METHODS: We interviewed online 478 students (February to April, 2021) who each valued the same 10 EQ-5D-5L health states in 1 of 3 arms. Arm A used a standard cTTO, expanded with 2 questions to explore reasons for NT and censoring. Arms B and C used a time trade-off with modified alternatives offered to overcome loss aversion, to unify the tasks for positive and negative utilities, and to enable eliciting utilities < -1. RESULTS: In arms B and C, we observed less NT than in A (respectively, 4% and 4% vs 10%), more strictly negative utilities (38% and 40% vs 25%), and more utilities ≤ -1 (18% and 30% vs 10%). The average utility of state 55555 dropped to -2.15 and -2.52 from -0.53. Enabling finer trades in arm A reduced NT by 70%. Arms B and C yielded an intuitive association between negative utilities and state severity. These arms were considered more difficult and resulted in more inconsistencies. CONCLUSIONS: The discriminatory power of cTTO can be improved, but it may require increasing the difficulty of the task. The standard cTTO may overestimate the utilities, especially of severe states.

Indonesia youth population norms for EQ-5D-Y-3 L, EQ-5D-Y-5 L and the PedsQL generic core scale: lower health related quality of life relates to high economic status and stress.

BACKGROUND: The availability of population norms from generic health-related quality of life (HRQoL) instruments can support the interpretation of health outcomes. This study aimed to provide Indonesian youth population norms for the generic HRQoL measures: EQ-5D-Y-3 L, EQ-5D-Y-5 L, and the PedsQL Generic Core Scales. In addition the opportunity arising from the generation of a large representative sample was taken to explore the relationships between HRQoL, health, and socio-economic factors. METHODS: A representative sample of 1103 Indonesian children (aged 8-16 years) completed EQ-5D-Y-3 L, EQ-5D-Y-5 L, the PedsQL Generic Core Scales, and questions related to demographic data and self-reported health status. A stratified quota sampling design was used to represent Indonesian children in terms of residence, age, gender, and geographical area. Family expenses per capita per month were retrieved from parents to determine a child's economic status. RESULTS: The total sample was representative of the Indonesian youth general population. The proportions of participants who reported problems were 43.35% (EQ-5D-Y-3 L), 44.10% (EQ-5D-Y-5 L), and 94.93% (PedsQL Generic), with 31.7% of children reporting health complaints. Older children (13-16 years) reported more problems than younger children (8-12 years). Children living in urban areas reported more problems than children living in rural areas. The lowest value health state reported was '12332' (valued at 0.54), and the minimum EQ VAS score was 60.00. Moderate correlations were found between EQ-5D-Y-3 L values to EQ VAS scores and to PedsQL Total Score. Hierarchical regression analysis showed that females, older age, and having health complaints contributed to a lower level of HRQoL as measured by EQ-5D-Y-3 L values, EQ VAS, and PedsQL Total Score. Remarkably, children with high economic status had lower EQ VAS and PedsQL Total Scores. Among symptoms, 'having stress' had the largest influence with respect to lower EQ-5D-Y-3L values, EQ VAS, and PedsQL Total Score. CONCLUSIONS: Population norms for children's HRQoL as measured by EQ-5D-Y-3 L, EQ-5D-Y-5 L, and the PedsQL Generic Scales are now available for Indonesia. Age, gender, economic status, and health complaints were related to children's HRQoL. These results provide a basis for health studies and health policy for the youth population of Indonesia.

Development of an EQ-5D Value Set for India Using an Extended Design (DEVINE) Study: The Indian 5-Level Version EQ-5D Value Set.

OBJECTIVES: This study aimed to develop the Indian 5-level version EQ-5D (EQ-5D-5L) value set, which is a key input in health technology assessment for resource allocation in healthcare. METHODS: A cross-sectional survey using the EuroQol Group's Valuation Technology was undertaken in a representative sample of 3548 adult respondents, selected from 5 different states of India using a multistage stratified random sampling technique. The participants were interviewed using a computer-assisted personal interviewing technique. This study adopted a novel extended EuroQol Group's Valuation Technology design that included 18 blocks of 10 composite time trade-off (c-TTO) tasks, comprising 150 unique health states, and 36 blocks of 7 discrete choice experiment (DCE) tasks, comprising 252 DCE pairs. Different models were explored for their predictive performance. Hybrid modeling approach using both c-TTO and DCE data was used to estimate the value set. RESULTS: A total of 2409 interviews were included in the analysis. The hybrid heteroscedastic model with censoring at -1 combining c-TTO and DCE data yielded the most consistent results and was used for the generation of the value set. The predicted values for all 3125 health states ranged from -0.923 to 1. The preference values were most affected by the pain/discomfort dimension. CONCLUSIONS: This is the largest EQ-5D-5L valuation study conducted so far in the world. The Indian EQ-5D-5L value set will promote the effective conduct of health technology assessment studies in India, thereby generating credible evidence for efficient resource use in healthcare.

EQ-5D-Y-3L and EQ-5D-Y-5L proxy report: psychometric performance and agreement with self-report.

BACKGROUND: Self-report is the standard for measuring people's health-related quality of life (HRQoL), including children. However, in certain circumstances children cannot report their own health. For this reason, children's HRQoL measures often provide both a self-report and a proxy-report form. It is not clear whether the measurement properties will be the same for these two forms. We investigated whether it would be beneficial to extend the classification system of the EQ-5D-Y proxy questionnaire from 3 to 5 response levels. The agreement between self-report and proxy-report was assessed for both EQ-5D-Y measures. METHODS: The study included 286 pediatric patients and their caregivers as proxies. At three consecutive measurements-baseline, test-retest and follow-up-the proxies assessed the child's HRQoL using the EQ-5D-Y-3L, EQ-5D-Y-5L, the PedsQL Generic, and matched disease-specific instruments. The proxy versions of EQ-5D-Y-3L and EQ-5D-Y-5L were compared in terms of feasibility, distribution properties, convergent validity, test-retest and responsiveness. Agreement between both EQ-5D-Y proxy versions to their respective self-report versions was assessed at baseline and follow-up. RESULTS: The proportion of missing responses was 1% for the EQ-5D-Y-3L and 1.4% for the EQ-5D-Y-5L. The frequency of health state with no problems in all dimensions (11111) was slightly lower for the EQ-5D-Y-5L (21.3% vs 16.7%). Regarding the convergent validity with the PedsQL and disease-specific measures, the proxy versions of EQ-5D-Y-3L and EQ-5D-Y-5L had similar magnitudes of associations between similar dimensions. The means of test-retest coefficients between the two versions of the EQ-5D-Y proxy were comparable (0.83 vs. 0.84). Regarding reported improved conditions, responsiveness of the EQ-5D-Y-5L proxy (26.6-54.1%) was higher than that of the EQ-5D-Y-3L proxy (20.7-46.4%). Except for acutely ill patients, agreement between the EQ-5D-Y-5L proxy and self-reports was at least moderate. CONCLUSIONS: Extending the number of levels of the proxy version of EQ-5D-Y can improve the classification accuracy and the ability to detect health changes over time. The level structure of EQ-5D-Y-5L was associated with a closer agreement between proxy and self-report. The study findings support extending the EQ-5D-Y descriptive system from 3 to 5 levels when administered by a proxy, which is often the case in the pediatric population.

Comparing measurement properties of EQ-5D-Y-3L and EQ-5D-Y-5L in paediatric patients.

BACKGROUND: The adult versions EQ-5D-3L and EQ-5D-5L have been extensive compared. This is not the case for the EQ-5D youth versions. The study aim was to compare the measurement properties and responsiveness of EQ-5D-Y-3L and EQ-5D-Y-5L in paediatric patients. METHODS: A sample of patients 8-16 years old with different diseases and a wide range of disease severity was asked to complete EQ-5D-Y-3L, EQ-5D-Y-5L, PedsQL Generic Core Scale, and selected, appropriate disease-specific instruments, three times. EQ-5D-Y-3L and EQ-5D-Y-5L were compared in terms of: feasibility, (re-)distribution properties, discriminatory power, convergent validity, test-retest reliability, and responsiveness. RESULTS: 286 participating patients suffered from one of the following diseases: major beta-thalassemia, haemophilia, acute lymphoblastic leukaemia, acute illness. Missing responses were comparable between versions of the EQ-5D-Y, suggesting comparable feasibility. The number of patients in the best health state (level profile 11111) was equal in both EQ-5D-Y versions. The projection of EQ-5D-Y-3L scores onto EQ-5D-Y-5L for all dimensions showed that the two additional levels in EQ-5D-Y-5L slightly improved the accuracy of patients in reporting their problems, especially if severe. Convergent validity with PedsQL and disease-specific measures showed that the two EQ-5D-Y versions performed about equally. Test-retest reliability (EQ-5D-Y-3L 0.78 vs EQ-5D-Y-5L 0.84), and sensitivity for detecting health changes, were both better in EQ-5D-Y-5L. CONCLUSIONS: Extending the number of levels did not give clear superiority to EQ-5D-Y-5L over EQ-5D-Y-3L based on the criteria assessed in this study. However, increasing the number of levels benefitted EQ-5D-Y performance in the measurement of moderate to severe problems and especially in longitudinal study designs.

Peruvian Valuation of the EQ-5D-5L: A Direct Comparison of Time Trade-Off and Discrete Choice Experiments.

OBJECTIVES: (1) To produce Peruvian general population EQ-5D-5L value sets on a quality-adjusted life-year scale, (2) to investigate the feasibility of a "Lite" protocol less reliant on the composite time trade-off (cTTO), and (3) to compare cTTO and discrete choice experiment (DCE) value sets. METHODS: A random sample of adults (N = 1000) in Lima, Arequipa, and Iquitos did a home interview; 300 were randomly selected to complete 11 cTTOs first. All respondents completed a DCE, including 10 latent-scale pairs (A/B) with 5 EQ-5D-5L attributes, and 12 matched pairs (A/B and B/C) with 5 EQ-5D-5L and one lifespan attributes. We estimated a cTTO heteroscedastic tobit (N = 300) model and 3 DCE Zermelo-Bradley-Terry models (N = 300, 700, and 1000). RESULTS: Each model produced a consistent value set (20 positive incremental parameters). Nevertheless, their lowest quality-adjusted life-year values differed greatly (cTTO: -1.076 [N = 300]; DCE: -0.984 [300], 0.048 [700], -0.213 [1000]). Compared with the cTTO, the DCE (N = 300) produced different parameters (Pearson's correlation = 0.541), fewer insignificant parameters (0 vs 8), and fewer values less than 0 (26% vs 44%). Compared with the DCE (N = 300), the DCE (N = 700) produced higher values but similar parameters (Pearson's correlation = 0.800). CONCLUSIONS: Besides producing EQ-5D-5L value sets for Peru, the results casts doubt about the feasibility of a Lite protocol like the one in this study. Additionally, fundamental differences between cTTO and DCE-without the existence of a gold standard-need further clarification. The choice between the two rational value sets produced in the current study is a matter of judgment and may have substantial policy implications.

How Prevalent Are Implausible EQ-5D-5L Health States and How Do They Affect Valuation? A Study Combining Quantitative and Qualitative Evidence.

BACKGROUND: When designing a valuation study, a criterion that has been used for selecting health states for direct valuation is whether the health states are plausible for respondents, because it is assumed that inclusion of implausible states would compromise data quality. OBJECTIVES: To understand which health states are implausible, and how their values differ from the values of the plausible counterparts. METHODS: One thousand six hundred Chinese students valued all 3125 health states of the 5-level EuroQol 5-dimensional questionnaire (EQ-5D-5L) using the EuroQol visual analogue scale. Of these, 890 students also indicated whether each valued state was implausible or not using a binary scale. Additional qualitative interviews were conducted concerning the thought processes involved in the valuation of implausible states. We calculated an implausible score for each state. Then we examined the effect of implausibility on visual analogue scale values by fitting 2 regression models. Two independent researchers analyzed the qualitative transcripts using thematic analysis. RESULTS: Approximately 30% of the EQ-5D-5L health states were rated as implausible by at least 50% of the respondents, but there was substantial heterogeneity in views about the plausibility of EQ-5D-5L states. Health states with dimensions that conflicted were more likely to be judged as implausible states. Health states that respondents deemed as implausible were more difficult to value and had lower values. CONCLUSIONS: When respondents found the EQ-5D states to be implausible, they tended to give them lower values. Nevertheless, completely excluding implausible states from a valuation study is not possible because there is a lack of agreement among respondents on which states are implausible.

Effect of Health State Sampling Methods on Model Predictions of EQ-5D-5L Values: Small Designs Can Suffice.

OBJECTIVE: The current five-level EQ-5D (EQ-5D-5L) valuation protocol requires the valuation of 86 states. It has been demonstrated that the selection of empirically valued health states affects the extrapolated values in three-level EQ-5D (EQ-3D-3L). In this investigation, we aim to compare the performance of the current EQ-5D-5L valuation design with other designs. STUDY DESIGN: 1603 university students participated in a valuation study using a visual analog scale (VAS) to produce values for all EQ-5D-5L states. Different designs were generated to test their prediction accuracy. METHODS: Subsamples of the dataset were used to mimic data obtained from a particular design; the remaining dataset was used as the validation set. In addition to EuroQol Group Valuation Technology (EQ-VT) design, alternative subsamples and designs were created using random, orthogonal, and "optimizing D-efficiency" sampling methods. The root mean squared error (RMSE) was used as the measure of prediction accuracy. RESULTS: The EuroQol Group Valuation Technology (EQ-VT) design showed an average RMSE of 3.44 on EQ-VAS, for all 3125 health states combined. Notably, a 25-state orthogonal design performed similarly to the EQ-VT design, with a smaller RMSE of 3.40, and was thus the most efficient design. One caveat with respect to the orthogonal design was that it did not predict the mild states well. CONCLUSIONS: Our study supports the EQ-VT design. Smaller designs were identified with similar overall prediction accuracy. It is worth investigating whether issues with misprediction of mild states can be resolved, as the use of smaller size designs would reduce the cost of the valuation of EQ-5D-5L considerably.

Toward a Smaller Design for EQ-5D-5L Valuation Studies.

BACKGROUND: To construct an EQ-5D-5L value set, the EuroQol Group developed a standard protocol named EuroQol Valuation Technology (EQ-VT), prescribing the valuation of 86 health states utilizing the composite time trade-off (cTTO) approach, and subsequently modeled the observed values to yield values for all 3125 states. OBJECTIVE: A recent study demonstrated that a 25-state orthogonal design could provide as accurate predictions as the EQ-VT design applying visual analogue scale data. We aimed to test that design using time trade-off (TTO) data. METHOD: We collected TTO values utilizing EQ-VT, orthogonal, and D-efficient designs. The EQ-VT design included 86 health states distributed over 3 blocks of 30 states with some duplicates. The orthogonal and D-efficient designs each comprised 1 block of 30 states. A total of 525 university students were asked to value a random block of health states using EQ-PVT (a PowerPoint replica of EQ-VT software), which generated 100 observations per health state in all 3 designs. We modeled data by design and compared the root mean square error (RMSE) between observed and predicted values within and across the designs. RESULTS: The EQ-VT design had the lowest RMSE of 0.052; the RMSEs for the orthogonal and the D-efficient designs were 0.066 and 0.063, respectively. RMSE results between designs differed for more severe health states. Some coefficients differed between designs. CONCLUSION: Smaller designs did not lead to significant increases in prediction errors when modeling TTO data (measuring 0.01 on a utility scale). Resource-constrained countries may use small designs for valuation studies, especially when other types of preference data, such as those from discrete choice experiments, are collected and modeled jointly.

Overview, Update, and Lessons Learned From the International EQ-5D-5L Valuation Work: Version 2 of the EQ-5D-5L Valuation Protocol.

A standardized 5-level EuroQol 5-dimensional questionnaire (EQ-5D-5L) valuation protocol was first used in national studies in the period 2012 to 2013. A set of problems encountered in this initial wave of valuation studies led to the subsequent refinement of the valuation protocol. To clarify lessons learned and how the protocol was updated when moving from version 1.0 to the current version 2.1 and 2.0, this article will (1) present the challenges faced in EQ-5D-5L valuation since 2012 and how these were resolved and (2) describe in depth a set of new challenges that have become central in currently ongoing research on how EQ-5D-5L health states should be valued and modeled.

Attribute level overlap (and color coding) can reduce task complexity, improve choice consistency, and decrease the dropout rate in discrete choice experiments.

A randomized controlled discrete choice experiment (DCE) with 3,320 participating respondents was used to investigate the individual and combined impact of level overlap and color coding on task complexity, choice consistency, survey satisfaction scores, and dropout rates. The systematic differences between the study arms allowed for a direct comparison of dropout rates and cognitive debriefing scores and accommodated the quantitative comparison of respondents' choice consistency using a heteroskedastic mixed logit model. Our results indicate that the introduction of level overlap made it significantly easier for respondents to identify the differences and choose between the choice options. As a stand-alone design strategy, attribute level overlap reduced the dropout rate by 30%, increased the level of choice consistency by 30%, and avoided learning effects in the initial choice tasks of the DCE. The combination of level overlap and color coding was even more effective: It reduced the dropout rate by 40% to 50% and increased the level of choice consistency by more than 60%. Hence, we can recommend attribute level overlap, with color coding to amplify its impact, as a standard design strategy in DCEs.

Selecting Health States for EQ-5D-3L Valuation Studies: Statistical Considerations Matter.

BACKGROUND: For many countries, the three-level EuroQol five-dimensional questionnaire (EQ-5D-3L) value sets have been established to estimate health state utilities. To generate these value sets, researchers first collect values for a subset of preselected health states from a panel representing the general public, and then use a prediction algorithm to generate values for all 243 states. High prevalence of a health state in daily practice has historically been a key criterion in selecting a subset of health states as the observed set. More recently, other criteria have been suggested, especially approaches based on statistical criteria such as randomization and orthogonality. OBJECTIVES: To evaluate the validity and accuracy of both the earlier and newer criteria, in terms of prediction of values for all the health states and of the values of common health states in particular. METHODS: We used a pre-existing data set that contained visual analogue scale values from 126 students, each of whom valued all 243 EQ-5D-3L states. Then, we generated a series of designs and subsequently modeled the data with respect to each design. Some of these designs were used in the past; for example, the Measurement and Valuation of Health approach was included. Others were newly generated. The performance of different designs was evaluated in terms of the lowest root mean squared error for all health states taken together, and separately for common and rare states. Classification as common or rare was based on the frequency of the states' occurrence in three patient and population data sets pooled together (N = 5269). RESULTS: The orthogonal design with 54 health states produced the lowest root mean squared errors. Over-representation of common health states in a design did not improve the estimations for these states. The published designs performed the worst, whereas the random selection designs were good on average. Nevertheless, the performance of the random selection designs showed more variance compared with orthogonal designs, because some of the former designs did not display appropriate balance. CONCLUSIONS: The published designs gave rise to large estimation errors for the extrapolated EQ-5D-3L health states. The orthogonal design focusing on statistical efficiency showed its superiority. Overall, when weighing up design properties, increased statistical efficiency outweighs an increased error rate, if any, in rare health states.

Effect of Level Overlap and Color Coding on Attribute Non-Attendance in Discrete Choice Experiments.

OBJECTIVE: The aim of this study was to test the hypothesis that level overlap and color coding can mitigate or even preclude the occurrence of attribute nonattendance in discrete choice experiments. METHODS: A randomized controlled experiment with five experimental study arms was designed to investigate the independent and combined impact of level overlap and color coding on respondents' attribute nonattendance. The systematic differences between the study arms allowed for a direct comparison of observed dropout rates and estimates of the average number of attributes attended to by respondents, which were obtained by using augmented mixed logit models that explicitly incorporated attribute non-attendance. RESULTS: In the base-case study arm without level overlap or color coding, the observed dropout rate was 14%, and respondents attended, on average, only two out of five attributes. The independent introduction of both level overlap and color coding reduced the dropout rate to 10% and increased attribute attendance to three attributes. The combination of level overlap and color coding, however, was most effective: it reduced the dropout rate to 8% and improved attribute attendance to four out of five attributes. The latter essentially removes the need to explicitly accommodate for attribute non-attendance when analyzing the choice data. CONCLUSIONS: On the basis of the presented results, the use of level overlap and color coding are recommendable strategies to reduce the dropout rate and improve attribute attendance in discrete choice experiments.

Advocating a Paradigm Shift in Health-State Valuations: The Estimation of Time-Preference Corrected QALY Tariffs.

BACKGROUND: Despite evidence of nonproportional trade-offs in time trade-off exercises and the explicit incorporation of exponential discounting in health technology assessment calculations, quality-adjusted life-year (QALY) tariffs are currently still established under the assumption of linear time preferences. OBJECTIVES: The aim of this study was to introduce a general method of accommodating for nonlinear time preferences in discrete choice experiment (DCE) duration studies and to evaluate its impact on estimated QALY tariffs. METHODS: A parsimonious utility function is proposed that accommodates any discounting function and preserves linear time preferences as a special case. Based on an efficient DCE design and 1775 respondents from a nationally representative scientific household panel, preferences and QALY tariffs for the Dutch SF-6D were estimated while accommodating for nonlinear time preferences via exponential and hyperbolic discounting functions. RESULTS: When the discount rate was estimated directly, we found strong evidence of nonlinear time preferences (with an exponential and hyperbolic discount rate of 5.7% and 16.5%, respectively). When the discount rate was estimated as a function of health state severity, we found that years lived in better health states are discounted minus years lived in impaired health states. Finally, the best statistical fit was obtained when using a hyperbolic discount function, which resulted in smaller QALY decrements and fewer health states classified as worse than immediate death. CONCLUSIONS: Our results highlight the relevance and even necessity of a paradigm shift in health valuation studies in favor of time-preference corrected QALY tariffs, with potentially important implications for health technology assessment calculations and regulatory decisions.

Evaluation of Split Version and Feedback Module on the Improvement of Time Trade-Off Data.

BACKGROUND: EQ-5D-5L valuation studies previously reported many inconsistent responses in time trade-off (TTO) data. A number of possible elements, including ordering effects of the valuation tasks, mistakes at the sorting question, and interviewers' (learning) effects, may contribute to their inconsistency. OBJECTIVES: This study aimed to evaluate the effect of two modifications on consistency of TTO data in The Netherlands (NL) and Hong Kong (HK): (1) separating the valuation of the Better than Dead (BTD) and Worse than Dead (WTD) states; and (2) Implementation of feedback (FB) module by offering an opportunity to review TTO responses. METHODS: A crossover design with two study arms was used to test the effect of the modifications. In each jurisdiction, six interviewers were involved where half the interviewers started using the standard version, and the other half started with the split version. Each version was switched after every 25 (NL) or 30 (HK) interviews until 400 interviews were completed. RESULTS: In the NL and HK, 404 and 403 respondents participated, respectively. With the use of the FB module, the proportion of respondents with inconsistent responses was lowered from 17.8% to 10.6% (P < 0.001) in NL and from 31.8% to 22.3% (P = 0.003) in HK. The result of separating the valuation of BTD and WTD states was not straightforward because it reduced the inconsistency rate in NL but not in HK. CONCLUSIONS: The results support implementation of the FB module to promote the consistency of the data. The separation of the BTD and WTD task is not supported.

Valuation of EuroQol Five-Dimensional Questionnaire, Youth Version (EQ-5D-Y) and EuroQol Five-Dimensional Questionnaire, Three-Level Version (EQ-5D-3L) Health States: The Impact of Wording and Perspective.

BACKGROUND: Valuations of health states were affected by the wording of the two instruments (EQ-5D-3L and EQ-5D-Y) and by the perspective taken (child or adult). OBJECTIVES: There is a growing demand for value sets for the EQ-5D-Y (EQ-5D instrument for younger populations). Given the similarities between EQ-5D-Y and EQ-5D-3L, we investigated whether valuations of health states were affected by the differences in wording between the two instruments and by the perspective taken in the valuation exercise (child or adult). STUDY DESIGN: Respondents were randomly assigned to EQ-5D-3L or EQ-5D-Y (instrument) and further into two groups that either valued health states for an adult or for a 10-year-old child (perspective). The valuation tasks were composite time trade-off (C-TTO) and discrete choice experiments (DCE), including comparisons with death (DCE + death). Members of the adult general population in four countries (Germany, Netherlands, Spain, England) participated in computer-assisted personal interviews. METHODS: Two-way multivariate analysis of variance (MANOVA) and post hoc tests were used to compare C-TTO responses and chi-square tests were conducted to compare DCE + death valuations. RESULTS: A significant interaction effect between instrument and perspective for C-TTO responses was found. Significant differences by perspective (adult and child) occurred only for the EQ-5D-3L. Significant differences in values between instruments (EQ-5D-3L and EQ-5D-Y) occurred only for the adult perspective. Both significant results were confirmed by the DCE + death results. When comparing EQ-5D-3L for adult perspective and EQ-5D-Y for child perspective, values were also significantly different. CONCLUSIONS: The results identified an interaction effect between wording of the instrument and perspective on elicited values, suggesting that current EQ-5D-3L value sets should not be employed to assign values to EQ-5D-Y health states.

Choice Defines Value: A Predictive Modeling Competition in Health Preference Research.

OBJECTIVE: To identify which specifications and approaches to model selection better predict health preferences, the International Academy of Health Preference Research (IAHPR) hosted a predictive modeling competition including 18 teams from around the world. METHODS: In April 2016, an exploratory survey was fielded: 4074 US respondents completed 20 out of 1560 paired comparisons by choosing between two health descriptions (e.g., longer life span vs. better health). The exploratory data were distributed to all teams. By July, eight teams had submitted their predictions for 1600 additional pairs and described their analytical approach. After these predictions had been posted online, a confirmatory survey was fielded (4148 additional respondents). RESULTS: The victorious team, "Discreetly Charming Econometricians," led by Michal Jakubczyk, achieved the smallest χ2, 4391.54 (a predefined criterion). Its primary scientific findings were that different models performed better with different pairs, that the value of life span is not constant proportional, and that logit models have poor predictive validity in health valuation. CONCLUSIONS: The results demonstrated the diversity and potential of new analytical approaches in health preference research and highlighted the importance of predictive validity in health valuation.

Quality Control Process for EQ-5D-5L Valuation Studies.

BACKGROUND: The values of the five-level EuroQol five-dimensional questionnaire (EQ-5D-5L) are elicited using composite time trade-off and discrete choice experiments. Unfortunately, data quality issues and interviewer effects were observed in the first few EQ-5D-5L valuation studies. To prevent these issues from occurring in later studies, the EuroQol Group established a cyclic quality control (QC) process. OBJECTIVES: To describe this QC process and show its impact on data quality. METHODS: A newly developed QC tool provided information about protocol compliance, interviewer effects, and mean values by health state severity. In a cyclic process, this information is initially used to evaluate whether new interviewers meet minimal quality requirements and later to provide feedback about how their performance may be improved. To investigate the impact of this cyclic process, we compared the quality of the data in Dutch and Spanish valuation studies that did not have this QC process with that in the follow-up studies in the same countries that used the QC process. Data quality was measured using protocol violations, variability between interviewers, the proportion of inconsistent responders, and clustering of composite time trade-off values. RESULTS: In Spain, protocol violations were reduced from 87% in the valuation study to 5% in the follow-up study and in the Netherlands from 20% to 8%. In both countries, interviewers performed more homogeneously in the follow-up studies. The number of inconsistent respondents was reduced by 23.2% in Spain and 23.6% in the Netherlands. Values were less clustered in the follow-up studies. CONCLUSIONS: The implementation of a strict QC process in EQ-5D-5L valuation studies increases interviewer protocol compliance and promotes data quality.

Are Health State Valuations from the General Public Biased? A Test of Health State Reference Dependency Using Self-assessed Health and an Efficient Discrete Choice Experiment.

Health state valuations of patients and non-patients are not the same, whereas health state values obtained from general population samples are a weighted average of both. The latter constitutes an often-overlooked source of bias. This study investigates the resulting bias and tests for the impact of reference dependency on health state valuations using an efficient discrete choice experiment administered to a Dutch nationally representative sample of 788 respondents. A Bayesian discrete choice experiment design consisting of eight sets of 24 (matched pairwise) choice tasks was developed, with each set providing full identification of the included parameters. Mixed logit models were used to estimate health state preferences with respondents' own health included as an additional predictor. Our results indicate that respondents with impaired health worse than or equal to the health state levels under evaluation have approximately 30% smaller health state decrements. This confirms that reference dependency can be observed in general population samples and affirms the relevance of prospect theory in health state valuations. At the same time, the limited number of respondents with severe health impairments does not appear to bias social tariffs as obtained from general population samples. Copyright © 2016 John Wiley & Sons, Ltd.