Developing and testing methods for deriving preference-based measures of health from condition-specific measures (and other patient-based measures of outcome)

Abstract

Objectives: Generic preference-based measures such as EQ-5D are widely used to estimate quality-adjusted life-years but may not be available or, more importantly, appropriate in some medical conditions. Condition-specific preference-based measures (CSPBMs) provide an alternative to generic measures that may be more relevant in some conditions. This project conducted five studies to examine issues in the development and use of CSPBMS: (1) literature review of measures; (2) deriving health states values for classifications with highly correlated dimensions; (3) impact of condition labelling; (4) impact of add-on dimensions; and (5) comparative performance of measures.

Design: (1) Systematic search and literature review; (2) and (5) psychometric analyses on existing data; (2), (3) and (4) valuation surveys and survey analyses.

Setting: Valuation surveys conducted using face-to-face interviews in the respondents' homes.

Participants: Valuation surveys conducted using representative samples of the UK general population.

Interventions: Not applicable.

Main outcome measures: The project developed a CSPBM CORE-6D and analyses AQL-5D, CORE-6D, EORTC-8D, EQ-5D, OAB-5D and SF-6D data.

Results: (1) There was substantial variability in methods used to develop CSPBMs. (2) A new method for generating states using Rasch analysis was undertaken, which successfully dealt with the problem of highly correlated domains. (3) Condition labels affected utility values but this was dependent on the condition and severity of the health state. (4) Adding on an extra dimension affected health-state values and preference weights for other dimensions. (5) The performance of CSPBMs was comparable with that of their parent instrument and of generic preference-based measures with better performance for discrimination between severity groups.

Conclusions: CSPBMs have an important role for economic evaluation, for which generic measures are inappropriate. However, their use in economic evaluation may be compromised by naming the condition; the exclusion of side effects and comorbidities; and focusing effects. Whether a reduction in comparability should be accepted depends on the extent of any gain in validity and responsiveness. This will depend on the condition and measure in question. Research agenda: (1) The appropriateness of generic preference-based measures should be examined in more conditions (and compared with CSPBMs). (2) Further quantitative and qualitative work is requested into the impact of, and reasons for labelling effects. (3) Use of add-ons for condition-specific measures (for side effects and comorbidities) and as a solution to the limitation of generic measures should be explored.

Funding: The National Institute for Health Research Health Technology Assessment programme.