Methods for the Indirect estimation of health state utilities
Lead Research Organisation:
University of Leeds
Department Name: Medicine
Abstract
Decisions made by NICE on the use of treatments in the NHS are based on the evidence for both clinical and cost-effectiveness. The methods used by NICE include the quality adjusted life year (QALY) which is a measure of both the length and quality of life gained from a treatment. A QALY can be measured using a health related quality of life questionnaire. But information from these sorts of questionnaires is not always available and often must be inferred from another source. This research aims to improve the methods used for making such inferences (often referred to as mapping) by reducing bias in the estimates that are produced and accurately describing how uncertain these estimates are.
We aim to achieve our objectives in three ways. First, we will look at how closely each instrument of health related quality of life measures the same domains, or aspects of health. This will help us determine the degree to which mapping from one instrument to the other will be reliable. We will then examine ways in which we can make more explicit the uncertainty in the estimates that arise from mapping between instruments. Finally, we will look at ways that we can combine information from both instruments to improve the mapping process.
To conduct our research we will be using a range of existing data sets. These data sets contain information about a number of different instruments for measuring health related quality of life. These instruments are the EQ-5D, EQ-5D Child friendly version, SF-36, EORTC-QLQ30, KIDSCREEN26 and HUI2. The data in these data sets has been collected for a number of different studies, including the Health Survey for England 1996 and clinical trials for the treatment of cancer. No new data will be collected, all patient data will be anonymised and unlinked so that no individuals can be identified and ethical approval will be sought as appropriate before any data is used.
As a result of this research we will develop a tool kit for researchers to use when making inferences about health related quality of life data from alternative sources. We will also work to promote the understanding of these methods to the public to improve the transparency of NICE?s decision making processes.
We aim to achieve our objectives in three ways. First, we will look at how closely each instrument of health related quality of life measures the same domains, or aspects of health. This will help us determine the degree to which mapping from one instrument to the other will be reliable. We will then examine ways in which we can make more explicit the uncertainty in the estimates that arise from mapping between instruments. Finally, we will look at ways that we can combine information from both instruments to improve the mapping process.
To conduct our research we will be using a range of existing data sets. These data sets contain information about a number of different instruments for measuring health related quality of life. These instruments are the EQ-5D, EQ-5D Child friendly version, SF-36, EORTC-QLQ30, KIDSCREEN26 and HUI2. The data in these data sets has been collected for a number of different studies, including the Health Survey for England 1996 and clinical trials for the treatment of cancer. No new data will be collected, all patient data will be anonymised and unlinked so that no individuals can be identified and ethical approval will be sought as appropriate before any data is used.
As a result of this research we will develop a tool kit for researchers to use when making inferences about health related quality of life data from alternative sources. We will also work to promote the understanding of these methods to the public to improve the transparency of NICE?s decision making processes.
Technical Summary
Mapping between health-related quality of life instruments is an increasingly common technique for estimating utility scores for use in economic evaluations. Mapping refers to the process of estimating utility values for use in an economic evaluation through predicting results for a target preference based measure from data collected using a non-preference based measure. This process allows economists to include information on preferences for health states within economic evaluation where direct evaluation is not possible, thereby improving resource allocation decisions. The use of mapping algorithms is increasing in the literature and is set to grow further as the National Institute for Health and Clinical Excellence (NICE) has proposed the use of mapped estimates as a legitimate approach to estimating utility values in the 2008 edition of their methodology guidelines for technology appraisals. Despite increasing levels of interest in this area, the use of such approaches has not yet been fully theoretically justified.
This research addresses three key theoretical problems relating to mapping. The first is the process of determining the most appropriate statistical approach for estimating utility scores from non-preference based instruments. Which method to use will be dependent on the concordance between the domains of the two instruments, the properties of the data produced by the outcome measure (whether outcomes are ordinal or cardinal for example) and the population group being studied. The second problem is how to best capture and characterise uncertainty in the final estimates of utility values. At present only random error is captured in mapping model. Yet there are two other sources of uncertainty one arising from the information lost in the mapped estimates owing to discordance between the instruments and heterogeneity in the health state concordance between the measures at the individual respondent level. The final problem to be addressed will be how to combine information on global health from two different instruments.
The completion of this work will result in a toolkit for researchers to enable them to develop consistent methods for mapping that account fully for the nature of the data that is being mapped, use the most appropriate statistical methods and capture fully the uncertainty that arises when mapping. Such a toolkit will enable more reliable, consistent and transparent decision making with respect to the allocation healthcare resources.
This research addresses three key theoretical problems relating to mapping. The first is the process of determining the most appropriate statistical approach for estimating utility scores from non-preference based instruments. Which method to use will be dependent on the concordance between the domains of the two instruments, the properties of the data produced by the outcome measure (whether outcomes are ordinal or cardinal for example) and the population group being studied. The second problem is how to best capture and characterise uncertainty in the final estimates of utility values. At present only random error is captured in mapping model. Yet there are two other sources of uncertainty one arising from the information lost in the mapped estimates owing to discordance between the instruments and heterogeneity in the health state concordance between the measures at the individual respondent level. The final problem to be addressed will be how to combine information on global health from two different instruments.
The completion of this work will result in a toolkit for researchers to enable them to develop consistent methods for mapping that account fully for the nature of the data that is being mapped, use the most appropriate statistical methods and capture fully the uncertainty that arises when mapping. Such a toolkit will enable more reliable, consistent and transparent decision making with respect to the allocation healthcare resources.
