Valuation of preference-based measures: can existing preference data be used to select a smaller sample of health states?

Background: Different countries have different preferences regarding health, and there are different value sets for popular preference-based measures across different countries. However, the cost of collecting data to generate country-specific value sets can be prohibitive for countries with smaller population size or low- and middle-income countries (LMIC). This paper explores whether existing preference weights could be modelled alongside a small own country valuation study to generate representative estimates. This is explored using a case study modelling UK data alongside smaller US samples to generate US estimates.
Methods: We analyse EQ-5D valuation data derived from representative samples of the US and UK populations using time trade-off to value 42 health states. A nonparametric Bayesian model was applied to estimate a US value set using the full UK dataset and subsets of the US dataset for 10, 15, 20 and 25 health states. Estimates are compared to a US value set estimated using US values alone using mean predictions and root mean square error.
Results: The results suggest that using US data elicited for 20 health states alongside the existing UK data produces similar predicted mean valuations and RMSE as the US value set, while 25 health states produce the exact features.
Conclusions: The promising results suggest that existing preference data could be combined with a small valuation study in a new country to generate preference weights, making own country value sets more achievable for LMIC. Further research is encouraged.