Clara Mukuria

A systematic review, psychometric analysis and qualitative assessment of generic preference-based measures of health in mental health populations and the estimation of mapping functions from widely used specific measures

BACKGROUND Generic preference-based measures of health like the EQ-5D and SF-6D(®) are increasingly being used in economic evaluation and outcome assessment. However, there are concerns as to whether or not these generic measures are appropriate for use in people with mental health problems. OBJECTIVES The EQ-5D and SF-36(®) (including its derivatives the SF-12(®) and SF-6D) were assessed using the psychometric criteria of validity and responsiveness using quantitative and qualitative methods. Another aim was to estimate mapping functions between the EQ-5D and SF-6D and condition-specific measures, where appropriate. DESIGN Four studies were undertaken to examine the appropriateness of the measures: (1) a systematic review of quantitative evidence on validity and responsiveness; (2) a further quantitative assessment of these criteria using existing data sets; (3) a review of qualitative research on the quality of life of people with mental health problems; and (4) qualitative semistructured interviews of people with a full range of problems. A fifth study estimated mapping functions between mental health-specific measures and the EQ-5D and SF-6D. SETTING A choice of venue was offered for the interviews including the participants own home, a room at the university or a centre frequently used by mental health services. PARTICIPANTS The interviews were undertaken with 19 people with a broad range of mental health problems at varying levels of severity. MAIN OUTCOME MEASURES The reviews included the EQ-5D and SF-36 (and the SF-12 and SF-6D). The psychometric analysis included the Hospital Anxiety and Depression Scale (HADS), Clinical Outcomes in Routine Evaluation - Outcome Measure (CORE-OM), Generalised Anxiety Disorder Assessment (GAD-7), General Health Questionnaire (GHQ-12) and Patient Health Questionnaire (PHQ-9). RESULTS (1) and (2) The EQ-5D and SF-36 achieved an adequate level of performance in depression, and to some extent in anxiety and personality disorder. Results from the psychometric analyses in schizophrenia and bipolar disorder have been more mixed. (3) A framework analysis of 13 studies identified six major themes. (4) The interview data fitted the themes from the review well and resulted in minor modifications to the themes. The final set of themes comprised: well-being and ill-being; control, autonomy and choice; self-perception; belonging; activity; hope and hopelessness; and physical health. CONCLUSIONS The EQ-5D and SF-36 achieved mixed results in the quantitative testing against psychometric criteria. The qualitative analysis suggests this is because they provide a very limited coverage of themes identified by people with mental health problems. Recommendations for future work include the development of new preference-based measures in mental health that are based on, or substantially revise, an existing measure. FUNDING The Medical Research Council.

Cost-effectiveness of an Improving Access to Psychological Therapies service

BACKGROUND Effective psychological therapies have been recommended for common mental health problems, such as depression and anxiety, but provision has been poor. Improving Access to Psychological Therapies (IAPT) may provide a cost-effective solution to this problem. AIMS To determine the cost-effectiveness of IAPT at the Doncaster demonstration site (2007-2009). METHOD An economic evaluation comparing costs and health outcomes for patients at the IAPT demonstration site with those for comparator sites, including a separate assessment of lost productivity. Sensitivity analyses were undertaken. RESULTS The IAPT site had higher service costs and was associated with small additional gains in quality-adjusted life-years (QALYs) compared with its comparator sites, resulting in a cost per QALY gained of £29 500 using the Short Form (SF-6D). Sensitivity analysis using predicted EQ-5D scores lowered this to £16 857. Costs per reliable and clinically significant (RCS) improvement were £9440 per participant. CONCLUSIONS Improving Access to Psychological Therapies provided a service that was probably cost-effective within the usual National Institute for Health and Clinical Excellence (NICE) threshold range of £20 000-30 000, but there was considerable uncertainty surrounding the costs and outcome differences.

Using generic preference-based measures in mental health: psychometric validity of the EQ-5D and SF-6D

BACKGROUND Generic preference-based measures (EuroQoL-5D (EQ-5D) and SF-6D) are used in the economic evaluation of mental health interventions. However, there are inconsistent findings regarding their psychometric properties. AIMS To investigate the psychometric properties of the EQ-5D and SF-6D in different mental health conditions, using seven existing data-sets. METHOD The construct validity and responsiveness of the measures were assessed in comparison with condition-specific indicators. RESULTS Evidence for construct validity and responsiveness in common mental health and personality disorders was found (correlations 0.22-0.64; effect sizes 0.37-1.24; standardised response means 0.45-1.31). There was some evidence for validity in schizophrenia (correlations 0.05-0.43), but responsiveness was unclear. CONCLUSIONS EQ-5D and SF-6D can be used in the economic evaluation of interventions for common mental health problems with some confidence. In schizophrenia, a preference-based measure focused on the impact of mental health should be considered.

Eliciting Societal Preferences for Weighting QALYs for Burden of Illness and End of Life

Objectives . Recent proposals for value-based assessment, made by the National Institute of Health and Care Excellence (NICE) in the United Kingdom, recommended that burden of illness (BOI) should replace end of life (EOL) as a factor for consideration when deciding on new health technologies. This article reports on a study eliciting societal preferences for 1) BOI from a medical condition, defined as quality-adjusted life year (QALY) loss due to premature mortality and prospective morbidity, and 2) EOL, defined as expected life expectancy of less than 2 years and expected life expectancy gain from new treatment of 3 months or more. Methods . A discrete choice experiment survey was conducted with an online UK general population sample. Respondents chose whether they thought the health service should treat patient group A or B: life expectancy and health-related quality of life (HRQOL) with current treatment or life expectancy and HRQOL gains from new treatment, respectively. These attributes were used to derive BOI, QALY gain, and EOL. The respondents’ choices were analyzed using conditional logistic regression with a range of specifications examined, including BOI or EOL, QALY gain and QALY gain squared, and robustness. QALY weights were estimated. Results . The sample of 3669 respondents was representative of the UK population for age and sex. QALY gain had a positive and significant coefficient across all models. QALY gain squared term was negative and significant across all models, indicating a diminishing marginal social value from QALY gains. When included, the BOI coefficient was generally small, positive, and significant, but this was not consistent across the different life expectancy variants. EOL was always positive and significant. Conclusions . The social value of a QALY gain is not equal between recipients but depends on whether they are end of life, and it may depend on the prospective burden of illness.

Comparison of Modes of Administration and Alternative Formats for Eliciting Societal Preferences for Burden of Illness

BackgroundProposals for value-based assessment, made by the National Institute of Health and Care Excellence (NICE) in the UK, recommended that burden of illness (BOI) should be used to weight QALY gain. This paper explores some of the methodological issues in eliciting societal preferences for BOI.AimsThis study explores the impact of mode of administration and framing in a survey for eliciting societal preferences for BOI.MethodsA pairwise comparison survey with six arms was conducted online and via face-to-face interviews, involving two different wordings of questions and the inclusion/exclusion of pictures. Respondents were asked which of two patient groups they thought a publically funded health service should treat, where the groups varied by life expectancy without treatment, health-related quality of life (HRQOL) without treatment, survival gain from treatment, and HRQOL gain from treatment. Responses across different modes of administration, wording and use of pictures were compared using chi-squared tests and probit regression analysis controlling for respondent socio-demographic characteristics.ResultsThe sample contained 371 respondents: 69 were interviewed and 302 completed the questionnaire online. There were some differences in socio-demographic characteristics across the online and interview samples. Online respondents were less likely to choose the group with higher BOI and more likely to treat those with a higher QALY gain, but there were no statistically significant differences by wording or the inclusion of pictures for the majority of questions. Regression analysis confirmed these results. Respondents chose to treat the group with larger treatment gain, but there was little support for treating the group with higher BOI. Respondents also preferred to treat the group with treatment gains in life expectancy rather than HRQOL.ConclusionsMode of administration did impact on responses, whereas question wording and pictures did not impact on responses, even after controlling for the socio-demographic characteristics of respondents in the regression analysis.

Mapping Functions in Health-Related Quality of Life Mapping from Two Cancer-Specific Health-Related Quality-of-Life Instruments to EQ-5D-3L

Background. Clinical trials in cancer frequently include cancer-specific measures of health but not preference-based measures such as the EQ-5D that are suitable for economic evaluation. Mapping functions have been developed to predict EQ-5D values from these measures, but there is considerable uncertainty about the most appropriate model to use, and many existing models are poor at predicting EQ-5D values. This study aims to investigate a range of potential models to develop mapping functions from 2 widely used cancer-specific measures (FACT-G and EORTC-QLQ-C30) and to identify the best model. Methods. Mapping models are fitted to predict EQ-5D-3L values using ordinary least squares (OLS), tobit, 2-part models, splining, and to EQ-5D item-level responses using response mapping from the FACT-G and QLQ-C30. A variety of model specifications are estimated. Model performance and predictive ability are compared. Analysis is based on 530 patients with various cancers for the FACT-G and 771 patients with multiple myeloma, breast cancer, and lung cancer for the QLQ-C30. Results. For FACT-G, OLS models most accurately predict mean EQ-5D values with the best predicting model using FACT-G items with similar results using tobit. Response mapping has low predictive ability. In contrast, for the QLQ-C30, response mapping has the most accurate predictions using QLQ-C30 dimensions. The QLQ-C30 has better predicted EQ-5D values across the range of possible values; however, few respondents in the FACT-G data set have low EQ-5D values, which reduces the accuracy at the severe end. Conclusions. OLS and tobit mapping functions perform well for both instruments. Response mapping gives the best model predictions for QLQ-C30. The generalizability of the FACT-G mapping function is limited to populations in moderate to good health.

Deriving a Preference-Based Measure for Myelofibrosis from the EORTC QLQ-C30 and the MF-SAF

BACKGROUND Utility values are required for economic evaluation using cost-utility analyses. Often, generic measures such as the EuroQol five-dimensional questionnaire are used, but this may not appropriately reflect the health-related quality of life of patients with cancer including myelofibrosis. OBJECTIVE To derive a condition-specific preference-based measure for myelofibrosis using appropriate existing measures, the Myelofibrosis-Symptom Assessment Form and the European Organisation for Research and Treatment of Cancer Quality of Life 30 Questionnaire. METHODS Data from the Controlled Myelofibrosis Study with Oral JAK Inhibitor Treatment trial (n = 309) were used to derive the health state classification system. Psychometric and factor analyses were used to determine the dimensions of the classification system. Psychometric and Rasch analyses were then used to select an item to represent each dimension. Item selection was validated with experts. A selection of health states was valued by members of the general population using time trade-off. Finally, health state values were modeled using regression analysis to produce utility values for every state. RESULTS The Myelofibrosis 8 dimensions has eight dimensions: physical functioning, emotional functioning, fatigue, itchiness, pain under ribs on the left side, abdominal discomfort, bone or muscle pain, and night sweats. Regression models were estimated using time trade-off data from 246 members of the general population valuing a total of 33 states. The best performing model was a random effects maximum likelihood model producing utility values ranging from 0.089 to 1. CONCLUSIONS The Myelofibrosis 8 dimensions is a condition-specific preference-based measure for myelofibrosis. This measure can be used to generate utility values for myelofibrosis for any data set containing the Myelofibrosis-Symptom Assessment Form and the European Organisation for Research and Treatment of Cancer Quality of Life 30 Questionnaire data.

The Use of Mapping to Estimate Health State Utility Values

Mapping functions are estimated using regression analyses and are frequently used to predict health state utility values (HSUVs) in decision analytic models. Mapping functions are used when evidence on the required preference-based measure (PBM) is not available, or where modelled values are required for a decision analytic model, for example to control for important sociodemographic variables (such as age or gender). This article provides an overview of the latest recommendations including pre-mapping considerations, the mapping process including data requirements for undertaking the estimation of mapping functions, regression models for estimating mapping functions, assessing performance and reporting standards for mapping studies. Examples in rheumatoid arthritis are used for illustration. When reporting the results of mapping standards the following should be reported: a description of the dataset used (including distributions of variables used) and any analysis used to inform the selection of the model type and model specification. The regression method and specification should be justified, and as summary statistics may mask systematic bias in errors, plots comparing observed and predicted HSUVs. The final model (coefficients, error term(s), variance and covariance) should be reported together with a worked example. It is important to ensure that good practice is followed as any mapping functions will only be as appropriate and accurate as the method used to obtain them; for example, mapping should not be used if there is no overlap between the explanatory and target variables.

What is the evidence for the performance of generic preference-based measures? A systematic overview of reviews

ObjectiveTo assess the evidence on the validity and responsiveness of five commonly used preference-based instruments, the EQ-5D, SF-6D, HUI3, 15D and AQoL, by undertaking a review of reviews.MethodsFour databases were investigated using a strategy refined through a highly sensitive filter for systematic reviews. References were screened and a search for grey literature was performed. Identified citations were scrutinized against pre-defined eligibility criteria and data were extracted using a customized extraction template. Evidence on known group validity, convergent validity and responsiveness was extracted and reviewed by narrative synthesis. Quality of the included reviews was assessed using a modified version of the AMSTAR checklist.ResultsThirty reviews were included, sixteen of which were of excellent or good quality. The body of evidence, covering more than 180 studies, was heavily skewed towards EQ-5D, with significantly fewer studies investigating HUI3 and SF-6D, and very few the 15D and AQoL. There was also lack of head-to-head comparisons between GPBMs and the tests reported by the reviews were often weak. Where there was evidence, EQ-5D, SF-6D, HUI3, 15D and AQoL seemed generally valid and responsive instruments, although not for all conditions. Evidence was not consistently reported across reviews.ConclusionsAlthough generally valid, EQ-5D, SF-6D and HUI3 suffer from some problems and perform inconsistently in some populations. The lack of head-to-head comparisons and the poor reporting impedes the comparative assessment of the performance of GPBMs. This highlights the need for large comparative studies designed to test instruments’ performance.

Mapping the EQ-5D index from the cystic fibrosis questionnaire-revised using multiple modelling approaches

BackgroundThis study was designed to develop a mapping algorithm to estimate EQ-5D utility values from Cystic Fibrosis Questionnaire-Revised (CFQ-R) data.MethodsA cross-sectional survey of adults with cystic fibrosis (CF) was conducted in the UK. The survey consisted of the CFQ-R, the EQ-5D and a background questionnaire. Eight regression models, exploring item and domain level predictors, were evaluated using three different modelling approaches: ordinary least squares (OLS), Tobit, and a two-part model (TPM). Predictive performance in each model was assessed by intraclass correlations, information criteria (Bayesian information criteria and Alkaike information criteria), and root mean square error (RMSE).ResultsThe survey was completed by 401 participants. For all modelling approaches the best performing item level model included all items, and the best performing domain level model included the CFQ-R Physical-, Role- and Emotional-functioning, Vitality, Eating Disturbances, Weight, and Digestive Symptoms domains and a selection of squared terms. Overall, the item level TPM, including age and gender covariates performed best within sample validation, but OLS and TPM domain models with squared terms performed best out-of-sample and are recommended for mapping purposes.ConclusionsDomain and item level models using all three modelling approaches reached an acceptable degree of predictive performance with domain models performing well in out-of-sample validation. These mapping functions can be applied to CFQ-R datasets to estimate EQ-5D utility values for economic evaluations of interventions for patients with cystic fibrosis. Further research evaluating model performance in an independent sample is encouraged.