Use of generic and condition-specific measures of health-related quality of life in NICE decision-making: systematic review, statistical modelling and survey.

Abstract

Background: The National Institute for Health and Care Excellence recommends the use of generic preference-based measures (GPBMs) of health for its Health Technology Assessments (HTAs). However, these data may not be available or appropriate for all health conditions. Objectives: To determine whether GPBMs are appropriate for some key conditions and to explore alternative methods of utility estimation when data from GPBMs are unavailable or inappropriate. Design: The project was conducted in three stages: (1) A systematic review of the psychometric properties of three commonly used GPBMs [EQ-5D, SF-6D and Health Utilities Index Mark 3 (HUI3)] in four broadly defined conditions: visual impairment, hearing impairment, cancer and skin conditions. (2) Potential modelling approaches to ‘map’ EQ-5D values from condition-specific and clinical measures of health [European Organisation for Research and Treatment of Cancer Quality-of-life Questionnaire Core 30 (EORTC QLQ-C30) and Functional Assessment of Cancer Therapy – General Scale (FACT-G)] are compared for predictive ability and goodness of fit using two separate data sets. (3) Three potential extensions to the EQ-5D are developed as ‘bolt-on’ items relating to hearing, tiredness and vision. They are valued using the time trade-off method. A second valuation study is conducted to fully value the EQ-5D with and without the vision bolt-on item in an additional sample of 300 people. Main outcome measures: Comparisons of EQ-5D, SF-6D and HUI3 in four conditions with various generic and condition-specific measures. Mapping functions were estimated between EORTC QLQ-C30 and FACT-G with EQ-5D. Three bolt-ons to the EQ-5D were developed: EQ + hearing/vision/tiredness. A full valuation study was conducted for the EQ + vision. Results: (1) EQ-5D was valid and responsive for skin conditions and most cancers; in vision, its performance varied according to aetiology; and performance was poor for hearing impairments. The HUI3 performed well for hearing and vision disorders. It also performed well in cancers although evidence was limited and there was no evidence in skin conditions. There were limited data for SF-6D in all four conditions and limited evidence on reliability of all instruments. (2) Mapping algorithms were estimated to predict EQ-5D values from alternative cancer-specific measures of health. Response mapping using all the domain scores was the best performing model for the EORTC QLQ-C30. In an exploratory analysis, a limited dependent variable mixture model performed better than an equivalent linear model. In the full analysis for the FACT-G, linear regression using ordinary least squares gave the best predictions followed by the tobit model. (3) The exploratory valuation study found that bolt-on items for vision, hearing and tiredness had a significant impact on values of the health states, but the direction and magnitude of differences depended on the severity of the health state. The vision bolt-on item had a statistically significant impact on EQ-5D health state values and a full valuation model was estimated. Conclusions: EQ-5D performs well in studies of cancer and skin conditions. Mapping techniques provide a solution to predict EQ-5D values where EQ-5D has not been administered. For conditions where EQ-5D was found to be inappropriate, including some vision disorders and for hearing, bolt-ons provide a promising solution. More primary research into the psychometric properties of the generic preference-based measures is required, particularly in cancer and for the assessment of reliability. Further research is needed for the development and valuation of bolt-ons to EQ-5D.