IMPROVING UTILITY ASSESSMENT METHODS FOR PRIORITY SETTING IN HEALTH CARE

Economic evaluation is increasingly used for priority setting in health care. For economic evaluation, health care programs' impact on symptoms and functional capacity has to be translated into improvements in health-related quality of life (HRQoL). HRQoL is usually measured by means of "instruments". Here, the respondents (patients, the general population, health personnel) fill in questionnaires with items measuring levels of pain, immobility, anxiety, etc. Answers are subsequently translated into a 0 -1 scale by means of an algorithm. There is uncertainty in the scientific community about how best to develop such algorithms, e.g. what population to sample, how to administer valuation tasks, and how to analyze and interpret the results. Because preferences for health may vary across cultures, populations and countries, national algorithms are recommended for use in economic evaluation. The EQ-5D, the most commonly used instrument worldwide, and 15D are two such HRQOL instruments, neither of which have Norwegian algorithms. In 2014 we performed a study interviewing 600 individuals from the general population valued EQ-5D health states using different versions of the time trade-off valuation method. The study suggested that tweaking the valuation task in ways that, in theory, should be irrelevant to the value of a health state, influenced respondents. A qualitative study investigating how respondents with spinal cord injuries interpret the mobility items of 7 generic HRQoL instruments, including EQ-5D and 15D is under review. Results of this study indicate that the wording of the items across the different instruments is an important source of discrepancy of values obtained this patient group across different instruments. Further, the project has used postal and novel web-based approaches in general population samples to develop and test Norwegian algorithms for the 15D. A paper on this topic will be submitted by the end of 2016. This paper also identifies inconsistencies in the current methods to estimate the 15D algorithm, and a subsequent paper to suggest improved methods is planned. A paper in which we test the predictive accuracy of different regression models for EQ-5D health states has been published. The conclusion from this paper based on large datasets form China, Singapore and Spain, is that the standard regression models are over specified, and that simpler models may have better predictive accuracy. Using different methods and data, we have identified potential sources of biases for health state values. This is the first step in the scientific work to improve the methods. In some cases, this will be the basis to test new methods that take into account the potential biases. In other cases, like the choice of regression models, and inconsistencies of 15D algorithm estimation methods, we have explicit recommendations on how to improve current standard practice. This project involves international collaboration with leading international experts. Co-authors of papers produced are from Canada, Spain, Singapore, China, USA and Finland. We face challenges in refining current methods and in communication of the results to a conservative research community, where the existing methods clearly have limitations. The project will be useful for health authorities because it will provide a basis for better priority setting based on national values.

Economic evaluation is increasingly used for priority setting in health care. For economic evaluation, health care programs' impact on symptoms and functional capacity has to be translated into improvements in health-related quality of life (HRQoL). HRQoL is usually measured by means of "instruments". Here, the respondents (patients, the general population, health personnel) fill in questionnaires with items measuring levels of pain, immobility, anxiety, etc. Answers are subsequently ?translated? into a 0 -1 scale by means of an algorithm. There is uncertainty in the scientific community about how best to develop such algorithms, e.g. what population to sample, how to administer valuation tasks, and how to analyze and interpret the results. Because prefer ences for health may vary across cultures, populations and countries, national algorithms are recommended for use in economic evaluation. The EQ-5D, the most commonly used instrument worldwide, and 15D are two such HRQOL instruments, neither of which have Norwegian algorithms. This project will use interviews and novel web-based approaches in general population samples to develop and test Norwegian algorithms for the EQ-5D and 15D. In addition, the project will investigate other relevant and practical me thodological problems, such as whether algorithms can be developed more cheaply through shorter interviews, new ways of phrasing questions, or by web-based data collection. Finally the project will study to the extent to which valuations are influenced by personality traits or attitudes among the respondents. This study involves international collaboration with leading international experts. We face challenges in refining current methods and in communication of the results to a conservative research commu nity, where the existing methods clearly have limitations. The project will be useful for health authorities because it will provide a basis for better priority setting based on national values.