Samer Kharroubi - Academia.edu (original) (raw)
Papers by Samer Kharroubi
Health economics, 2007
Expected value of sample information (EVSI) involves simulating data collection, Bayesian updatin... more Expected value of sample information (EVSI) involves simulating data collection, Bayesian updating, and re-examining decisions. Bayesian updating in Weibull models typically requires Markov chain Monte Carlo (MCMC). We examine five methods for calculating posterior expected net benefits: two heuristic methods (data lumping and pseudo-normal); two Bayesian approximation methods (Tierney & Kadane, Brennan & Kharroubi); and the gold standard MCMC. A case study computes EVSI for 25 study options. We compare accuracy, computation time and trade-offs of EVSI versus study costs. Brennan & Kharroubi (B&K) approximates expected net benefits to within +/-1% of MCMC. Other methods, data lumping (+54%), pseudo-normal (-5%) and Tierney & Kadane (+11%) are less accurate. B&K also produces the most accurate EVSI approximation. Pseudo-normal is also reasonably accurate, whilst Tierney & Kadane consistently underestimates and data lumping exhibits large variance. B&K computation is 12 times faster t...
Health Economics, 2007
Expected value of sample information (EVSI) involves simulating data collection, Bayesian updatin... more Expected value of sample information (EVSI) involves simulating data collection, Bayesian updating, and re-examining decisions. Bayesian updating in Weibull models typically requires Markov chain Monte Carlo (MCMC). We examine five methods for calculating posterior expected net benefits: two heuristic methods (data lumping and pseudo-normal); two Bayesian approximation methods (Tierney & Kadane, Brennan & Kharroubi); and the gold standard MCMC. A case study computes EVSI for 25 study options. We compare accuracy, computation time and trade-offs of EVSI versus study costs. Brennan & Kharroubi (B&K) approximates expected net benefits to within +/-1% of MCMC. Other methods, data lumping (+54%), pseudo-normal (-5%) and Tierney & Kadane (+11%) are less accurate. B&K also produces the most accurate EVSI approximation. Pseudo-normal is also reasonably accurate, whilst Tierney & Kadane consistently underestimates and data lumping exhibits large variance. B&K computation is 12 times faster than the MCMC method in our case study. Though not always faster, B&K provides most computational efficiency when net benefits require appreciable computation time and when many MCMC samples are needed. The methods enable EVSI computation for economic models with Weibull survival parameters. The approach can generalize to complex multi-state models and to survival analyses using other smooth parametric distributions.
... In Sweeting (1996) some accurate formulae for Bartlett corrections, pos-terior expectations a... more ... In Sweeting (1996) some accurate formulae for Bartlett corrections, pos-terior expectations and predictive distributions are obtained. ... Two alternative fornmlae are given in Sweeting (1996): a product form and a summation form. ...
Statistics in Medicine
Few studies have compared preference values of health states obtained in different countries. Thi... more Few studies have compared preference values of health states obtained in different countries. This paper applies a nonparametric model to estimate and compare EQ-5D health state valuation data obtained from two countries using Bayesian methods. The data set is the US and UK EQ-5D valuation studies where a sample of 42 states defined by the EQ-5D was valued by representative samples of the general population from each country using the time trade-off technique. We estimate a function applicable across both countries which explicitly accounts for the differences between them, and is estimated using the data from both countries. The paper discusses the implications of these results for future applications of the EQ-5D and further work in this field.
... 5. Discussion This paper reports the findings from applying a new approach to modelling healt... more ... 5. Discussion This paper reports the findings from applying a new approach to modelling health state valuation data. ... as the SF-6D, EQ-5D and HUI2. Page 15. 13 ... Level 4 Blind, deaf, or mute Level 4 Requires the help of another person to eat, bathe, dress or use the toilet ...
Sample sizing for survival trials specifies a clinically significant proportional hazard, which i... more Sample sizing for survival trials specifies a clinically significant proportional hazard, which if obtained, implies the adoption of the new therapy. Such adoption decisions are often evaluated using health economic decision models. We develop methods to formally integrate decision modelling with sample size for survival studies using the expected value of sample information approach. The method requires characterisation of prior uncertainty around the proportional hazard, together with a model of survival, quality of life benefits and economic costs. EVSI computation uses Monte Carlo sampling to produce new simulated data-sets with specified size and follow-up. Each simulated dataset is synthesised with existing prior information. Because Bayesian updating for Weibull parameters is not analytically tractable, we use a novel form of Laplace approximation to estimate the decision model outputs. A case study builds on a classic text book example.
Applied health economics and health policy, Jan 21, 2015
There is interest in the extent to which valuations of health may differ between different countr... more There is interest in the extent to which valuations of health may differ between different countries and cultures, but few studies have compared preference values of health states obtained in different countries. We sought to estimate and compare two directly elicited valuations for SF-6D health states between the Japan and UK general adult populations using Bayesian methods. We analysed data from two SF-6D valuation studies where, using similar standard gamble protocols, values for 241 and 249 states were elicited from representative samples of the Japan and UK general adult populations, respectively. We estimate a function applicable across both countries that explicitly accounts for the differences between them, and is estimated using data from both countries. The results suggest that differences in SF-6D health-state valuations between the Japan and UK general populations are potentially important. The magnitude of these country-specific differences in health-state valuation dep...
Value in health : the journal of the International Society for Pharmacoeconomics and Outcomes Research, 2014
Conventionally, parametric models were used for health state valuation data. Recently, researcher... more Conventionally, parametric models were used for health state valuation data. Recently, researchers started to explore the use of nonparametric Bayesian methods in this area. We present a nonparametric Bayesian model to estimate a preference-based index for two condition-specific five-dimensional health state classifications, one for asthma (five-dimensional Asthma Quality of Life Utility Index) and the other for overactive bladder (five-dimensional Overactive Bladder Quality of Life-Utility Index). Samples of 307 and 311 members of the UK general population valued 99 health states selected from a total of 3125 health states defined by each of the measures using the time trade-off technique. The article presents the results of the nonparametric model and compares it with the original model estimated using a conventional parametric random-effects model. The different methods are compared theoretically and in terms of empirical performance across the two data sets. It also reports the ...
Medical Decision Making, 2003
It is well known that the modeling of cost data is often problematic due to the distribution of s... more It is well known that the modeling of cost data is often problematic due to the distribution of such data. Commonly observed problems include 1) a strongly right-skewed data distribution and 2) a significant percentage of zero-cost observations. This article demonstrates how a hurdle model can be implemented from a Bayesian perspective by means of Markov Chain Monte Carlo simulation methods using the freely available software WinBUGS. Assessment of model fit is addressed through the implementation of two cross-validation methods. The relative merits of this Bayesian approach compared to the classical equivalent are discussed in detail. To illustrate the methods described, patient-specific non-health-care resource-use data from a prospective longitudinal study and the Norfolk Arthritis Register (NOAR) are utilized for 218 individuals with early inflammatory polyarthritis (IP). The NOAR database also includes information on various patient-level covariates.
Bayesian Analysis, 2010
ABSTRACT We explore the use of importance sampling based on signed root log-likelihood ratios for... more ABSTRACT We explore the use of importance sampling based on signed root log-likelihood ratios for Bayesian computation. Approximations based on signed root log-likelihood ratios are used in two distinct ways; firstly, to define an importance function and, secondly, to define suitable control variates for variance reduction. These considerations give rise to alternative simulation-consistent schemes to MCMC for Bayesian computation in moderately parameterized regular problems. The schemes based on control variates can also be viewed as usefully supplementing computations based on asymptotic approximations by supplying external estimates of error. The methods are illustrated by a genetic linkage model and a censored regression model.
Value in Health, 2014
Objectives: There is interest in the extent to which valuations of health may differ between diff... more Objectives: There is interest in the extent to which valuations of health may differ between different countries and cultures, but few studies have compared preference values of health states obtained in different countries. The present study applies a nonparametric model to estimate and compare two HK and UK standard gamble values for six-dimensional health state short form (derived from short-form 36 health survey) (SF-6D) health states using Bayesian methods. Methods: The data set is the HK and UK SF-6D valuation studies in which two samples of 197 and 249 states defined by the SF-6D were valued by representative samples of the HK and UK general populations, respectively, both using the standard gamble technique. We estimated a function applicable across both countries that explicitly accounts for the differences between them, and is estimated using the data from both countries. Results: The results suggest that differences in SF-6D health state valuations between the UK and HK general populations are potentially important. In particular, the valuations of Hong Kong were meaningfully higher than those of the United Kingdom for most of the selected SF-6D health states. The magnitude of these country-specific differences in health state valuation depended, however, in a complex way on the levels of individual dimensions. Conclusions: The new Bayesian nonparametric method is a powerful approach for analyzing data from multiple nationalities or ethnic groups to understand the differences between them and potentially to estimate the underlying utility functions more efficiently.
Test, 2003
Some new accurate approximations for posterior expectations and Bartlett corrections are derived.... more Some new accurate approximations for posterior expectations and Bartlett corrections are derived. These approximations are modifications of formulae based on signed root log-likelihood ratios obtained in Sweeting (1996) and are designed to address two problems that arise in the practical application of these formulae in the multiparameter case. The first problem is a computational one associated with inversion of signed
Statistics, 2011
... In Bayesian Statistics , Edited by: Bernardo, JM, Berger, JO, Dawid, AP and Smith, AFM Vol. 5... more ... In Bayesian Statistics , Edited by: Bernardo, JM, Berger, JO, Dawid, AP and Smith, AFM Vol. 5, 427–444. Oxford: Oxford University Press. View all references and Sweeting and Kharroubi 1515. Sweeting, TJ and Kharroubi, SA 2003. ...
Statistics in Medicine, 2010
Few studies have compared preference values of health states obtained in different countries. Thi... more Few studies have compared preference values of health states obtained in different countries. This paper applies a nonparametric model to estimate and compare EQ-5D health state valuation data obtained from two countries using Bayesian methods. The data set is the US and UK EQ-5D valuation studies where a sample of 42 states defined by the EQ-5D was valued by representative samples of the general population from each country using the time trade-off technique. We estimate a function applicable across both countries which explicitly accounts for the differences between them, and is estimated using the data from both countries. The paper discusses the implications of these results for future applications of the EQ-5D and further work in this field.
Statistics and Computing, 2005
We consider exact and approximate Bayesian computation in the presence of latent variables or mis... more We consider exact and approximate Bayesian computation in the presence of latent variables or missing data. Specifically we explore the application of a posterior predictive distribution formula derived in , which is a particular form of Laplace approximation, both as an importance and a proposal distribution. We show that this formula provides a stable importance function for use within poor man's data augmentation schemes and that it can also be used as a proposal distribution within a Metropolis-Hastings algorithm for models that are not analytically tractable. We illustrate both uses in the case of a censored regression model and a normal hierarchical model, with both normal and Student t distributed random effects. Although the predictive distribution formula is motivated by regular asymptotic theory, it is not necessary that the likelihood has a closed form or that it possesses a local maximum.
Social Science & Medicine, 2007
It has long been recognised that respondent characteristics can impact on the values they give to... more It has long been recognised that respondent characteristics can impact on the values they give to health states. This paper reports on the findings from applying a non-parametric approach to estimate the covariates in a model of SF-6D health state values using Bayesian methods. The data set is the UK SF-6D valuation study, where a sample of 249 states defined by the SF-6D (a derivate of the SF-36) was valued by a sample of the UK general population using standard gamble. Advantages of the nonparametric model are that it can be used to predict scores in populations with different distributions of characteristics and that it allows for an impact to vary by health state (whilst ensuring that full health passes through unity). The results suggest an important age effect, with sex, class, education, employment and physical functioning probably having some effect, but the remaining covariates having no discernable effect. Adjusting for covariates in the UK sample made little difference to mean health state values. The paper discusses the implications of these results for policy.
The construction of mapping models is an increasingly popular mechanism for obtaining health stat... more The construction of mapping models is an increasingly popular mechanism for obtaining health state utility data to inform economic evaluations in health care. There is great variation in the sophistication of the methods utilized but to date very little discussion of the appropriate theoretical framework to guide the design and evaluation of these models. In this paper, we argue that recognizing mapping models as a form of indirect health state valuation allows the use of the framework described by Dolan for the measurement of social preferences over health. Using this framework, we identify substantial concerns with the method for valuing health states that is implicit in indirect utility models (IUMs), the conflation of two sets of respondents' values in such models, and the lack of a structured and statistically reasonable approach to choosing which states to value and how many observations per state to require in the estimation dataset. We also identify additional statistical challenges associated with clustering and censoring in the datasets for IUMs, additional to those attributable to the descriptive systems, and a potentially significant problem with the systematic understatement of uncertainty in predictions from IUMs. Whilst recognizing that IUMs appear to meet the needs of reimbursement organizations that use quality-adjusted life years in their appraisal processes, we argue that current proposed quality standards are inadequate and that IUMs are neither robust nor appropriate mechanisms for estimating utilities for use in cost-effectiveness analyses.
This article reports the application of a recently described approach to modeling health state va... more This article reports the application of a recently described approach to modeling health state valuation data and the impact of the respondent characteristics on health state valuations. The approach applies a nonparametric model to estimate a Bayesian Health Utilities Index Mark 2 (HUI 2) health state valuation algorithm. The data set is the UK HUI 2 valuation study where a sample of 51 states defined by the HUI 2 was valued by a sample of the UK general population using standard gamble. The article reports the application of the nonparametric model and compares it to the original model estimated using a conventional parametric random effects model. Advantages of the nonparametric model are that it can be used to predict scores in populations with different distributions of characteristics than observed in the survey sample and that it allows for the impact of respondent characteristics to vary by health state. The results suggest an important age effect with sex, having some effect, but the remaining covariates having no discernable effect. The article discusses the implications of these results for future applications of the HUI 2 and further work in this field.
Health economics, 2007
Expected value of sample information (EVSI) involves simulating data collection, Bayesian updatin... more Expected value of sample information (EVSI) involves simulating data collection, Bayesian updating, and re-examining decisions. Bayesian updating in Weibull models typically requires Markov chain Monte Carlo (MCMC). We examine five methods for calculating posterior expected net benefits: two heuristic methods (data lumping and pseudo-normal); two Bayesian approximation methods (Tierney & Kadane, Brennan & Kharroubi); and the gold standard MCMC. A case study computes EVSI for 25 study options. We compare accuracy, computation time and trade-offs of EVSI versus study costs. Brennan & Kharroubi (B&K) approximates expected net benefits to within +/-1% of MCMC. Other methods, data lumping (+54%), pseudo-normal (-5%) and Tierney & Kadane (+11%) are less accurate. B&K also produces the most accurate EVSI approximation. Pseudo-normal is also reasonably accurate, whilst Tierney & Kadane consistently underestimates and data lumping exhibits large variance. B&K computation is 12 times faster t...
Health Economics, 2007
Expected value of sample information (EVSI) involves simulating data collection, Bayesian updatin... more Expected value of sample information (EVSI) involves simulating data collection, Bayesian updating, and re-examining decisions. Bayesian updating in Weibull models typically requires Markov chain Monte Carlo (MCMC). We examine five methods for calculating posterior expected net benefits: two heuristic methods (data lumping and pseudo-normal); two Bayesian approximation methods (Tierney & Kadane, Brennan & Kharroubi); and the gold standard MCMC. A case study computes EVSI for 25 study options. We compare accuracy, computation time and trade-offs of EVSI versus study costs. Brennan & Kharroubi (B&K) approximates expected net benefits to within +/-1% of MCMC. Other methods, data lumping (+54%), pseudo-normal (-5%) and Tierney & Kadane (+11%) are less accurate. B&K also produces the most accurate EVSI approximation. Pseudo-normal is also reasonably accurate, whilst Tierney & Kadane consistently underestimates and data lumping exhibits large variance. B&K computation is 12 times faster than the MCMC method in our case study. Though not always faster, B&K provides most computational efficiency when net benefits require appreciable computation time and when many MCMC samples are needed. The methods enable EVSI computation for economic models with Weibull survival parameters. The approach can generalize to complex multi-state models and to survival analyses using other smooth parametric distributions.
... In Sweeting (1996) some accurate formulae for Bartlett corrections, pos-terior expectations a... more ... In Sweeting (1996) some accurate formulae for Bartlett corrections, pos-terior expectations and predictive distributions are obtained. ... Two alternative fornmlae are given in Sweeting (1996): a product form and a summation form. ...
Statistics in Medicine
Few studies have compared preference values of health states obtained in different countries. Thi... more Few studies have compared preference values of health states obtained in different countries. This paper applies a nonparametric model to estimate and compare EQ-5D health state valuation data obtained from two countries using Bayesian methods. The data set is the US and UK EQ-5D valuation studies where a sample of 42 states defined by the EQ-5D was valued by representative samples of the general population from each country using the time trade-off technique. We estimate a function applicable across both countries which explicitly accounts for the differences between them, and is estimated using the data from both countries. The paper discusses the implications of these results for future applications of the EQ-5D and further work in this field.
... 5. Discussion This paper reports the findings from applying a new approach to modelling healt... more ... 5. Discussion This paper reports the findings from applying a new approach to modelling health state valuation data. ... as the SF-6D, EQ-5D and HUI2. Page 15. 13 ... Level 4 Blind, deaf, or mute Level 4 Requires the help of another person to eat, bathe, dress or use the toilet ...
Sample sizing for survival trials specifies a clinically significant proportional hazard, which i... more Sample sizing for survival trials specifies a clinically significant proportional hazard, which if obtained, implies the adoption of the new therapy. Such adoption decisions are often evaluated using health economic decision models. We develop methods to formally integrate decision modelling with sample size for survival studies using the expected value of sample information approach. The method requires characterisation of prior uncertainty around the proportional hazard, together with a model of survival, quality of life benefits and economic costs. EVSI computation uses Monte Carlo sampling to produce new simulated data-sets with specified size and follow-up. Each simulated dataset is synthesised with existing prior information. Because Bayesian updating for Weibull parameters is not analytically tractable, we use a novel form of Laplace approximation to estimate the decision model outputs. A case study builds on a classic text book example.
Applied health economics and health policy, Jan 21, 2015
There is interest in the extent to which valuations of health may differ between different countr... more There is interest in the extent to which valuations of health may differ between different countries and cultures, but few studies have compared preference values of health states obtained in different countries. We sought to estimate and compare two directly elicited valuations for SF-6D health states between the Japan and UK general adult populations using Bayesian methods. We analysed data from two SF-6D valuation studies where, using similar standard gamble protocols, values for 241 and 249 states were elicited from representative samples of the Japan and UK general adult populations, respectively. We estimate a function applicable across both countries that explicitly accounts for the differences between them, and is estimated using data from both countries. The results suggest that differences in SF-6D health-state valuations between the Japan and UK general populations are potentially important. The magnitude of these country-specific differences in health-state valuation dep...
Value in health : the journal of the International Society for Pharmacoeconomics and Outcomes Research, 2014
Conventionally, parametric models were used for health state valuation data. Recently, researcher... more Conventionally, parametric models were used for health state valuation data. Recently, researchers started to explore the use of nonparametric Bayesian methods in this area. We present a nonparametric Bayesian model to estimate a preference-based index for two condition-specific five-dimensional health state classifications, one for asthma (five-dimensional Asthma Quality of Life Utility Index) and the other for overactive bladder (five-dimensional Overactive Bladder Quality of Life-Utility Index). Samples of 307 and 311 members of the UK general population valued 99 health states selected from a total of 3125 health states defined by each of the measures using the time trade-off technique. The article presents the results of the nonparametric model and compares it with the original model estimated using a conventional parametric random-effects model. The different methods are compared theoretically and in terms of empirical performance across the two data sets. It also reports the ...
Medical Decision Making, 2003
It is well known that the modeling of cost data is often problematic due to the distribution of s... more It is well known that the modeling of cost data is often problematic due to the distribution of such data. Commonly observed problems include 1) a strongly right-skewed data distribution and 2) a significant percentage of zero-cost observations. This article demonstrates how a hurdle model can be implemented from a Bayesian perspective by means of Markov Chain Monte Carlo simulation methods using the freely available software WinBUGS. Assessment of model fit is addressed through the implementation of two cross-validation methods. The relative merits of this Bayesian approach compared to the classical equivalent are discussed in detail. To illustrate the methods described, patient-specific non-health-care resource-use data from a prospective longitudinal study and the Norfolk Arthritis Register (NOAR) are utilized for 218 individuals with early inflammatory polyarthritis (IP). The NOAR database also includes information on various patient-level covariates.
Bayesian Analysis, 2010
ABSTRACT We explore the use of importance sampling based on signed root log-likelihood ratios for... more ABSTRACT We explore the use of importance sampling based on signed root log-likelihood ratios for Bayesian computation. Approximations based on signed root log-likelihood ratios are used in two distinct ways; firstly, to define an importance function and, secondly, to define suitable control variates for variance reduction. These considerations give rise to alternative simulation-consistent schemes to MCMC for Bayesian computation in moderately parameterized regular problems. The schemes based on control variates can also be viewed as usefully supplementing computations based on asymptotic approximations by supplying external estimates of error. The methods are illustrated by a genetic linkage model and a censored regression model.
Value in Health, 2014
Objectives: There is interest in the extent to which valuations of health may differ between diff... more Objectives: There is interest in the extent to which valuations of health may differ between different countries and cultures, but few studies have compared preference values of health states obtained in different countries. The present study applies a nonparametric model to estimate and compare two HK and UK standard gamble values for six-dimensional health state short form (derived from short-form 36 health survey) (SF-6D) health states using Bayesian methods. Methods: The data set is the HK and UK SF-6D valuation studies in which two samples of 197 and 249 states defined by the SF-6D were valued by representative samples of the HK and UK general populations, respectively, both using the standard gamble technique. We estimated a function applicable across both countries that explicitly accounts for the differences between them, and is estimated using the data from both countries. Results: The results suggest that differences in SF-6D health state valuations between the UK and HK general populations are potentially important. In particular, the valuations of Hong Kong were meaningfully higher than those of the United Kingdom for most of the selected SF-6D health states. The magnitude of these country-specific differences in health state valuation depended, however, in a complex way on the levels of individual dimensions. Conclusions: The new Bayesian nonparametric method is a powerful approach for analyzing data from multiple nationalities or ethnic groups to understand the differences between them and potentially to estimate the underlying utility functions more efficiently.
Test, 2003
Some new accurate approximations for posterior expectations and Bartlett corrections are derived.... more Some new accurate approximations for posterior expectations and Bartlett corrections are derived. These approximations are modifications of formulae based on signed root log-likelihood ratios obtained in Sweeting (1996) and are designed to address two problems that arise in the practical application of these formulae in the multiparameter case. The first problem is a computational one associated with inversion of signed
Statistics, 2011
... In Bayesian Statistics , Edited by: Bernardo, JM, Berger, JO, Dawid, AP and Smith, AFM Vol. 5... more ... In Bayesian Statistics , Edited by: Bernardo, JM, Berger, JO, Dawid, AP and Smith, AFM Vol. 5, 427–444. Oxford: Oxford University Press. View all references and Sweeting and Kharroubi 1515. Sweeting, TJ and Kharroubi, SA 2003. ...
Statistics in Medicine, 2010
Few studies have compared preference values of health states obtained in different countries. Thi... more Few studies have compared preference values of health states obtained in different countries. This paper applies a nonparametric model to estimate and compare EQ-5D health state valuation data obtained from two countries using Bayesian methods. The data set is the US and UK EQ-5D valuation studies where a sample of 42 states defined by the EQ-5D was valued by representative samples of the general population from each country using the time trade-off technique. We estimate a function applicable across both countries which explicitly accounts for the differences between them, and is estimated using the data from both countries. The paper discusses the implications of these results for future applications of the EQ-5D and further work in this field.
Statistics and Computing, 2005
We consider exact and approximate Bayesian computation in the presence of latent variables or mis... more We consider exact and approximate Bayesian computation in the presence of latent variables or missing data. Specifically we explore the application of a posterior predictive distribution formula derived in , which is a particular form of Laplace approximation, both as an importance and a proposal distribution. We show that this formula provides a stable importance function for use within poor man's data augmentation schemes and that it can also be used as a proposal distribution within a Metropolis-Hastings algorithm for models that are not analytically tractable. We illustrate both uses in the case of a censored regression model and a normal hierarchical model, with both normal and Student t distributed random effects. Although the predictive distribution formula is motivated by regular asymptotic theory, it is not necessary that the likelihood has a closed form or that it possesses a local maximum.
Social Science & Medicine, 2007
It has long been recognised that respondent characteristics can impact on the values they give to... more It has long been recognised that respondent characteristics can impact on the values they give to health states. This paper reports on the findings from applying a non-parametric approach to estimate the covariates in a model of SF-6D health state values using Bayesian methods. The data set is the UK SF-6D valuation study, where a sample of 249 states defined by the SF-6D (a derivate of the SF-36) was valued by a sample of the UK general population using standard gamble. Advantages of the nonparametric model are that it can be used to predict scores in populations with different distributions of characteristics and that it allows for an impact to vary by health state (whilst ensuring that full health passes through unity). The results suggest an important age effect, with sex, class, education, employment and physical functioning probably having some effect, but the remaining covariates having no discernable effect. Adjusting for covariates in the UK sample made little difference to mean health state values. The paper discusses the implications of these results for policy.
The construction of mapping models is an increasingly popular mechanism for obtaining health stat... more The construction of mapping models is an increasingly popular mechanism for obtaining health state utility data to inform economic evaluations in health care. There is great variation in the sophistication of the methods utilized but to date very little discussion of the appropriate theoretical framework to guide the design and evaluation of these models. In this paper, we argue that recognizing mapping models as a form of indirect health state valuation allows the use of the framework described by Dolan for the measurement of social preferences over health. Using this framework, we identify substantial concerns with the method for valuing health states that is implicit in indirect utility models (IUMs), the conflation of two sets of respondents' values in such models, and the lack of a structured and statistically reasonable approach to choosing which states to value and how many observations per state to require in the estimation dataset. We also identify additional statistical challenges associated with clustering and censoring in the datasets for IUMs, additional to those attributable to the descriptive systems, and a potentially significant problem with the systematic understatement of uncertainty in predictions from IUMs. Whilst recognizing that IUMs appear to meet the needs of reimbursement organizations that use quality-adjusted life years in their appraisal processes, we argue that current proposed quality standards are inadequate and that IUMs are neither robust nor appropriate mechanisms for estimating utilities for use in cost-effectiveness analyses.
This article reports the application of a recently described approach to modeling health state va... more This article reports the application of a recently described approach to modeling health state valuation data and the impact of the respondent characteristics on health state valuations. The approach applies a nonparametric model to estimate a Bayesian Health Utilities Index Mark 2 (HUI 2) health state valuation algorithm. The data set is the UK HUI 2 valuation study where a sample of 51 states defined by the HUI 2 was valued by a sample of the UK general population using standard gamble. The article reports the application of the nonparametric model and compares it to the original model estimated using a conventional parametric random effects model. Advantages of the nonparametric model are that it can be used to predict scores in populations with different distributions of characteristics than observed in the survey sample and that it allows for the impact of respondent characteristics to vary by health state. The results suggest an important age effect with sex, having some effect, but the remaining covariates having no discernable effect. The article discusses the implications of these results for future applications of the HUI 2 and further work in this field.