Converting condition-specific measures into preference-based outcomes for use in economic evaluation

A scoping review to create a framework for the steps in developing condition-specific preference-based instruments de novo or from an existing non-preference-based instrument: use of item response theory or Rasch analysis

BACKGROUND

There is no widely accepted framework to guide the development of condition-specific preference-based instruments (CSPBIs) that includes both de novo and from existing non-preference-based instruments. The purpose of this study was to address this gap by reviewing the published literature on CSPBIs, with particular attention to the application of item response theory (IRT) and Rasch analysis in their development.

METHODS

A scoping review of the literature covering the concepts of all phases of CSPBI development and evaluation was performed from MEDLINE, Embase, PsychInfo, CINAHL, and the Cochrane Library, from inception to December 30, 2022.

RESULTS

The titles and abstracts of 1,967 unique references were reviewed. After retrieving and reviewing 154 full-text articles, data were extracted from 109 articles, representing 41 CSPBIs covering 21 diseases or conditions. The development of CSPBIs was conceptualized as a 15-step framework, covering four phases: 1) develop initial questionnaire items (when no suitable non-preference-based instrument exists), 2) establish the dimensional structure, 3) reduce items per dimension, 4) value and model health state utilities. Thirty-nine instruments used a type of Rasch model and two instruments used IRT models in phase 3.

CONCLUSION

We present an expanded framework that outlines the development of CSPBIs, both from existing non-preference-based instruments and de novo when no suitable non-preference-based instrument exists, using IRT and Rasch analysis. For items that fit the Rasch model, developers selected one item per dimension and explored item response level reduction. This framework will guide researchers who are developing or assessing CSPBIs.

The ASAS health index: a new era for health impact assessment in spondyloarthritis

The concept of spondyloarthritis (SpA) encompasses a series of entities that share clinical and imaging characteristics and a common genetic basis. These diseases can affect 0.20 to 1.6% of the general population, limiting functioning and affecting the quality of life of patients. Considering the patient perspective in the management of the disease and ensuring patients are sufficiently prepared to participate in decisionmaking is critical to treatment success as well as for optimal health outcomes. The overall picture of impairments, limitations, and restrictions in activities or social participation of patients with SpA is not adequately assessed in SpA-specific instruments. Therefore, it is quite relevant to measure the broader range of impairments that can affect SpA patients and integrate these into one measure of overall functioning in daily life. The Assessment of SpondyloArthritis international Society-Health Index (ASAS HI) is a recently introduced health instrument for evaluating SpA based on the International Classification of Functioning, Disability and Health, also known as ICF, that could cover a good part of the health metric needs in SpA. This review addresses its origins, its measurement properties, its use in routine clinical practice, as well as its prospects for future use.

Development of one general and six country-specific algorithms to assess societal health utilities based on ASAS HI

Objective

Health utilities represent preference values that persons attach to health states. This study aims to develop one general and six country-specific algorithms to calculate societal preference values for health of patients with spondyloarthritis (SpA), as assessed by the disease-specific Assessment of SpondyloArthritis international Society Health Index (ASAS HI).

Methods

A survey was performed in random population samples from six European countries. In a best-worst choice experiment, subjects were asked to indicate repeatedly which of 4 random aspects of the 17-item ASAS HI was were most and least important. Bayesian analysis provided the relative importance of each of the 17 items. To rescale the relative importance scores on the absolute utility scale between 0 and 1, participants additionally completed two lead time trade-off experiments, one for 'severe SpA' and one for 'best health' without SpA. Six country-specific algorithms and one general algorithm were derived. The general algorithm was tested in 199 patients with axial SpA (axSpA).

Results

3039 subjects, mean age 47 years (SD 15) and 52% female completed the experiments. The population's health utility value for SpA varied between - 0.24 for 'worst' SpA (country range -0.35 to 0.03), and 0.88 for 'best' health (country range 0.81 to 0.90). Among 199 patients with axSpA, the mean utility was 0.36 (SD 0.30, range -0.24 to 0.88) and discriminated well between patients having high (Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) ≥ 4) or low (BASDAI < 4) disease activity (0.18 (SD 0.24) vs 0.51(SD 0.27), p<0.01).

Conclusion

One general and six country-specific algorithms are available to convert scores from the ASAS HI into disease-specific societal utility values.

Condition-specific or generic preference-based measures in oncology? A comparison of the EORTC-8D and the EQ-5D-3L

PURPOSE

It has been argued that generic health-related quality of life measures are not sensitive to certain disease-specific improvements; condition-specific preference-based measures may offer a better alternative. This paper assesses the validity, responsiveness and sensitivity of a cancer-specific preference-based measure, the EORTC-8D, relative to the EQ-5D-3L.

METHODS

A longitudinal prospective population-based cancer genomic cohort, Cancer 2015, was utilised in the analysis. EQ-5D-3L and the EORTC QLQ-C30 (which gives EORTC-8D values) were asked at baseline (diagnosis) and at various follow-up points (3 months, 6 months, 12 months). Baseline values were assessed for convergent validity, ceiling effects, agreement and sensitivity. Quality-adjusted life-years (QALYs) were estimated and similarly assessed. Multivariate regression analyses were employed to understand the determinants of the difference in QALYs.

RESULTS

Complete case analysis of 1678 patients found that the EQ-5D-3L values at baseline were significantly lower than the EORTC-8D values (0.748 vs 0.829, p < 0.001). While the correlation between the instruments was high, agreement between the instruments was poor. The baseline health state values using both instruments were found to be sensitive to a number of patient and disease characteristics, and discrimination between disease states was found to be similar. Mean generic QALYs (estimated using the EQ-5D-3L) were significantly lower than condition-specific QALYs (estimated using the EORTC-8D) (0.860 vs 0.909, p < 0.001). The discriminatory power of both QALYs was similar.

CONCLUSIONS

When comparing a generic and condition-specific preference-based instrument, divergences are apparent in both baseline health state values and in the estimated QALYs over time for cancer patients. The variability in sensitivity between the baseline values and the QALY estimations means researchers and decision makers are advised to be cautious if using the instruments interchangeably.

The External Validity of Mapping MSIS-29 on EQ-5D Among Individuals With Multiple Sclerosis in Sweden

Background: Mapping can be performed to predict utility values from condition-specific measures when preference-based measures are absent. A previously developed algorithm that predicts EQ-5D-3L index values from the Multiple Sclerosis Impact Scale (MSIS-29) has not yet been externally validated. Aim: To examine the external validity of a previously developed mapping algorithm by testing the accuracy of predicting EQ-5D-3L index values from MSIS-29 among multiple sclerosis (MS) patients in Sweden. Methods: Cross-sectional individual-level data were collected from population-based Swedish registers between 2011 and 2014. Health-related quality of life was assessed through MSIS-29 and EQ-5D-3L at one point in time among 767 individuals with known disability level of MS. A previously developed mapping algorithm was applied to predict EQ-5D index values from MSIS-29 items, and the predictive accuracy was assessed through mean absolute error and root mean square error. Results: When applying the algorithm, the predicted mean EQ-5D-3L index value was 0.77 compared to the observed mean index value of 0.75. Prediction error was higher for individuals reporting EQ-5D values <0.5 compared to individuals reporting EQ-5D values ≥0.5. Mean absolute error (0.12) and root mean square error (0.18) were smaller or equal to the prediction errors found in the original mapping study. Conclusion: The mapping algorithm had similar predictive accuracy in the two independent samples although results showed that the highest predictive performance was found in groups with better health. Varied predictive accuracy in subgroups is consistent with previous studies and strategies to deal with this are warranted.

A Systematic Review of the Literature on the Development of Condition-Specific Preference-Based Measures of Health

BACKGROUND

Health state utility values (HSUVs) are required to calculate quality-adjusted life-years (QALYs). They are frequently derived from generic preference-based measures of health. However, such generic measures may not capture health attributes of relevance to specific conditions. In such cases, a condition-specific preference-based measure (CSPBM) may be more appropriate.

OBJECTIVE

This systematic review aimed to identify all published accounts of developing CSPBMs to describe and appraise the methods used.

METHOD

We undertook a systematic search (of Embase, MEDLINE, PsycINFO, Web of Science, the Cochrane Library, CINAHL, EconLit, ASSIA and the Health Management Information Consortium database) to identify published accounts of CSPBM development up to July 2015. Studies were reviewed to investigate the methods used to design classification systems, estimate HSUVs, and validate the measures.

RESULTS

A total of 86 publications were identified, describing 51 CSPBMs. Around two-thirds of these were QALY measures; the remainder were designed for clinical decision making only. Classification systems for 33 CSPBMs were derived from existing instruments; 18 were developed de novo. HSUVs for 34 instruments were estimated using a 'composite' approach, involving statistical modelling; the remainder used a 'decomposed' approach based on multi-attribute utility theory. Half of the papers that described the estimation of HSUVs did not report validating their measures.

CONCLUSION

Various methods have been used at all stages of CSPBM development. The choice between developing a classification system de novo or from an existing instrument may depend on the availability of a suitable existing measure, while the choice between a decomposed or composite approach appears to be determined primarily by the purpose for which the instrument is designed. The validation of CSPBMs remains an area for further development.

Which quality of life measures fit your relative effectiveness assessment?

BACKGROUND

Health-related quality of life (HRQoL) is an important endpoint of many healthcare interventions. This study develops guidance on how to select appropriate HRQoL measures for inclusion in a clinical trial, given the purposes of the HRQoL measurement.

METHODS

The guidance is based on a systematic literature review, discussions with members of the European Network for Health Technology Assessment (EUnetHTA) and two rounds of public consultation.

RESULTS

A set of twelve recommendations was developed, addressing the requirements for HRQoL data for relative effectiveness assessment, for cost-utility analyses and for informing clinical decision making. Recommendations relate to the choice of the type of measure as well as to aspects such as measurement frequency, target population and presentation.

CONCLUSIONS

The purpose and context of HRQoL measurement is crucial for the relevance of the data obtained with a specific HRQoL measure. It is recommended to always include a generic HRQoL instrument in clinical trials to cover a wide range of possible future uses of the HRQoL data.

Development of a preference-based index from the National Eye Institute Visual Function Questionnaire-25

IMPORTANCE

Understanding how individuals value health states is central to patient-centered care and to health policy decision making. Generic preference-based measures of health may not effectively capture the impact of ocular diseases. Recently, 6 items from the National Eye Institute Visual Function Questionnaire-25 were used to develop the Visual Function Questionnaire-Utility Index health state classification, which defines visual function health states.

OBJECTIVE

To describe elicitation of preferences for health states generated from the Visual Function Questionnaire-Utility Index health state classification and development of an algorithm to estimate health preference scores for any health state.

DESIGN, SETTING, AND PARTICIPANTS

Nonintervention, cross-sectional study of the general community in 4 countries (Australia, Canada, United Kingdom, and United States). A total of 607 adult participants were recruited from local newspaper advertisements. In the United Kingdom, an existing database of participants from previous studies was used for recruitment.

INTERVENTIONS

Eight of 15,625 possible health states from the Visual Function Questionnaire-Utility Index were valued using time trade-off technique.

MAIN OUTCOMES AND MEASURES

A θ severity score was calculated for Visual Function Questionnaire-Utility Index-defined health states using item response theory analysis. Regression models were then used to develop an algorithm to assign health state preference values for all potential health states defined by the Visual Function Questionnaire-Utility Index.

RESULTS

Health state preference values for the 8 states ranged from a mean (SD) of 0.343 (0.395) to 0.956 (0.124). As expected, preference values declined with worsening visual function. Results indicate that the Visual Function Questionnaire-Utility Index describes states that participants view as spanning most of the continuum from full health to dead.

CONCLUSIONS AND RELEVANCE

Visual Function Questionnaire-Utility Index health state classification produces health preference scores that can be estimated in vision-related studies that include the National Eye Institute Visual Function Questionnaire-25. These preference scores may be of value for estimating utilities in economic and health policy analyses.

Elicitation of health state utilities in soft tissue sarcoma

PURPOSE

Soft tissue sarcomas (STS) are uncommon tumours with varying histological subtypes. There is a paucity of available data concerning the quality-of-life (QoL) impact of STS which could be used to support economic evaluation of future treatments. This study aimed to elicit societal utility values for health states that depict the impact of STS and its treatment.

METHODS

Following the development of eight health state vignettes, a sample of 100 members of the UK general public participated in a valuation exercise to elicit utility values using the time trade-off procedure.

RESULTS

The treatment response state was valued as the least burdensome by participants followed by the prospect of stable disease (mean utility value: 0.736 SD 0.21). Serious adverse events were associated with a range of disutilities from -0.236 for grade III/IV pain to -0.357 for grade III/IV nausea/vomiting. Progressive disease was deemed the least desirable outcome and was associated with a substantial decline in utility (-0.473).

CONCLUSIONS

Findings suggest advanced STS are associated with significant burden for individuals. Treatment-related adverse events were seen as debilitating, however, progression represents an enormous challenge to QoL. This illustrates the significant value to individuals of extending the progression free survival period.

Setting priorities in and for end-of-life care: challenges in the application of economic evaluation

Health technology assessment processes aim to provide evidence on the effectiveness and cost-effectiveness of different elements of health care to assist setting priorities. There is a risk that services that are difficult to evaluate, and for which there is limited evidence on cost-effectiveness, may lose out in the competition for resources to those with better evidence. It is argued here that end-of-life care provides particular challenges for evaluation. Outcomes are difficult to measure, can take place over short time scales, and services can be difficult to characterise as they are tailored to the specific needs of individuals. Tools commonly used to measure health care outcomes do not appear to discriminate well in the end-of-life care context. It is argued that the assumption that units of time of different quality of life can simply be added to assess the overall experience at the end of life may not apply, and that alternative perspectives, such as the Peak and End Rule, might offer useful perspectives.

Methods and issues associated with the use of quality-adjusted life-years

In this article, we will focus on how preferences and utilities are measured, including the strengths and limitations of various approaches, discuss their use in estimating quality-adjusted life-years (QALYs) and make some recommendations for further research. Preferences are either measured using direct (visual analog scale, time trade-off or standard gamble) or indirect methods. The most commonly used generic indirect measures include the Quality of Well-Being scale, EuroQol-5 Dimension, Health Utilities Index and Short Form-6 Dimension. Disease-specific preference measures are increasingly being developed and applied in studies as more sensitive measures of health status. Preference-based measures and QALY measurement need to be enhanced, and additional research is needed to improve scientific methods for estimating preferences for health assessment. Given the increased focus on comparative effectiveness research, QALYs have the potential for helping researchers, clinicians, health policy-makers and patients to understand the relative effectiveness of alternative interventions for treating medical conditions.

Mapping QLQ-C30, HAQ, and MSIS-29 on EQ-5D

Background

Responses on condition-specific instruments can be mapped on the EQ-5D to estimate utility values for economic evaluation. Mapping functions differ in predictive quality, and not all condition-specific measures are suitable for estimating EQ-5D utilities. We mapped QLQ-C30, HAQ, and MSIS-29 on the EQ-5D and compared the quality of the mapping functions with statistical and clinical indicators.

Methods

We used 4 data sets that included both the EQ-5D and a condition-specific measure to develop ordinary least squares regression equations. For the QLQ-C30, we used a multiple myeloma data set and a non-Hodgkin lymphoma one. An early arthritis cohort was used for the HAQ, and a cohort of patients with relapsing remitting or secondary progressive multiple sclerosis was used for the MSIS-29. We assessed the predictive quality of the mapping functions with the root mean square error (RMSE) and mean absolute error (MAE) and the ability to discriminate among relevant clinical subgroups. Pearson correlations between the condition-specific measures and items of the EQ-5D were used to determine if there is a relationship between the quality of the mapping functions and the amount of correlated content between the used measures.

Results

The QLQ-C30 had the highest correlation with EQ-5D items. Average %RMSE was best for the QLQ-C30 with 10.9%, 12.2% for the HAQ, and 13.6% for the MSIS-29. The mappings predicted mean EQ-5D utilities without significant differences with observed utilities and discriminated between relevant clinical groups, except for the HAQ model.

Conclusions

The preferred mapping functions in this study seem suitable for estimating EQ-5D utilities for economic evaluation. However, this research shows that lower correlations between instruments lead to less predictive quality. Using additional validation tests besides reporting statistical measures of error improves the assessment of predictive quality.

Utility estimation in chronic obstructive pulmonary disease: a preference for change?

Economic evaluations of chronic obstructive pulmonary disease (COPD) incorporate utilities through multi-attribute utility (MAU) measures, most commonly the EQ-5D, to report health-related quality-of-life (HR-QOL) changes or differences. Questions have been raised about the sensitivity of these measures in COPD. Limitations in detecting adequate patient-level changes in HR-QOL over time in stable and exacerbation states compared with disease-specific instruments could also result in underestimation of known treatment benefits. The purpose of this article was to present and discuss the empirical evidence on the validity of generic MAU measures within the COPD population. We built upon a previously conducted validation review for the period 1997-2007 that used 'respiratory disease' and 'EQ-5D' as keywords. For this discussion, PubMed and EMBASE databases were searched for articles in English from 1988 to August 2009, using similar search words. Based on the performance of MAU measures in COPD and exacerbations, they appear to have limited discriminatory ability, particularly between moderate and severe COPD, despite known differences in HR-QOL. Sensitivity to clinically relevant change in stable COPD over time due to treatment also appears limited. Current research suggests adequate sensitivity regarding detecting the onset and resolution of an exacerbation; however, sensitivity is limited in the short term, such as daily changes in health status. The evidence suggests responsiveness of MAU measures may be restricted to large within-patient change, which leads to difficulties in evaluating the subtle but important impact of exacerbations. Studies presenting alternative methods of deriving COPD-related utilities are also discussed. Overall, the insensitivities of generic MAU measures in COPD can lead to biased cost-effectiveness analyses and ill-informed economic decisions. Alternative measures such as condition-specific preference-based measures may be used in circumstances where more sensitivity is needed. The trade-off allows relevant and sensitive matters most important to patients to be evaluated; however inevitable gaps such as those related to adverse events are not considered.

Mapping the QLQ-C30 quality of life cancer questionnaire to EQ-5D patient preferences

BACKGROUND

Although cancer-specific Health-Related Quality of Life (HRQOL) are commonly included in randomized clinical trials or other prospective non-randomized clinical studies, it is rare that preference-based instruments are used that allow the calculation of a utility weight suitable for estimating quality-adjusted life-years gained.

OBJECTIVE

To develop a mapping algorithm to transform the EORTC QLQ-C30 questionnaire responses into EQ-5D derived utilities.

STUDY DESIGN

Retrospective data analysis of a multicentre, multicountry prospective clinical trial in breast cancer patients.

METHODS

Regression analysis of individual pairs of EQ-5D and QLQ-C30 scores.

RESULTS

A model that explained 80% of the variance was developed to estimate EQ-5D Utilities from QLQ-C30 scores at individual level. From this reliable group level means and deviations can be derived.

CONCLUSIONS

Mapping from QLQ-C30 scores to EQ-5D-derived utilities when only QLQ-C30 data are available has been shown to be possible with good accuracy. Validation of the proposed algorithm in other external clinical datasets should be encouraged.

Estimation of a valuation function for a diabetes mellitus-specific preference-based measure of health: the Diabetes Utility Index

Preference-based measures of health (PBMH) provide 'preference' or 'utility' weights that enable the calculation of QALYs for the economic evaluations of interventions. The Diabetes Utility Index (DUI) was developed as a brief, self-administered, diabetes mellitus-specific PBMH that can efficiently estimate patient-derived health state utilities. To describe the development of the valuation function for the DUI, and to report the validation results of the valuation function. Multi-Attribute Utility Theory (MAUT) was used as the framework to develop a valuation function for the DUI. Twenty of 768 possible health states of the DUI classified as anchor states, single-attribute level states including corner states, and marker states were selected and described for preference elicitation interviews. Visual analogue scale and standard gamble (SG) exercises were used to measure preferences from individuals with diabetes recruited from primary care and community settings in and around Morgantown, WV, USA for the 20 health states defined by combinations of DUI attributes and severity levels. Data collected in the interviews were used to develop a valuation function that calculates utilities for the DUI health states and calculates attribute-level utilities. A validation survey of the valuation function was conducted in collaboration with the West Virginia University (WVU) Diabetes Institute. A total of 100 individuals with diabetes were interviewed and their preferences for various DUI health states measured. From data generated in the interviews, a DUI valuation function was developed on a scale where 1.00 = perfect health (PH) and 0.00 = the all worse 'pits' state, and adjusted to yield utilities on the conventional scale 1.00 = PH and 0.00 = dead. A total of 396 patients with diabetes who received care at WVU clinics completed a DUI mail validation survey (response rate = 33%). Clinical data consisting of International Classification of Diseases, 9th edition, diagnosis codes and glycosylated haemoglobin (HbA(1c)) values for the respondents were merged with their responses to the DUI. The utilities calculated by the scoring function of the DUI compared favourably to cardinal SG utilities for three DUI health states for which both assessments were available. The DUI utility function slightly underestimated actual SG utilities for mild and moderate health states (mean absolute difference = 0.05). There was a small but significant correlation between DUI utility scores and average past year HbA(1c) values (r = -0.30; p < 0.001). Respondents with two or more complications had significantly lower DUI utilities than those with no complications (p < 0.001) or one complication (p = 0.015). Insulin users had significantly lower DUI utilities than non-users (p < 0.001), and those with HbA(1c) values <7% had significantly higher DUI utilities than those with HbA(1c) values of >or=7% (p < 0.001). No significant association was found between DUI scores and age or sex. These results show evidence of the feasibility and validity of the DUI. Further research is suggested to demonstrate the generalizability of these findings, to study the responsiveness of the DUI, and to examine the clinical meaningfulness of DUI change scores.