Improving valuation sampling of EQ-5D health states

BACKGROUND: The original valuation exercise which formed the basis of the UK EQ-5D time trade-off social tariff of health states, employed a sampling scheme involving 43 health states. Neither that study, nor other published international valuations studies have used explicit quantifiable criteria to justify the choice of sampled states. New criteria are proposed and methods described to aid researchers in designing improved sampling schemes for future EQ-5D sampling exercises. METHOD: Four such criteria are described, and applied to assess the merits of four sampling schemes previously reported, using three large observational databases to quantify relative performance. An alternative sampling design conforming to these criteria is described, which aims to generate improved performance. RESULTS: Previous published approaches are shown to perform poorly against the measured criteria. The alternative sampling design is demonstrated to provide superior performance on all measures. CONCLUSION: Future valuation exercises using sampled health states based on this approach may be expected to offer benefits in terms of greater precision, avoidance of bias in favour of less severe states, and a higher proportion of research observations valued directly rather than dependent on extrapolation modelling.