METHODS 112 people were recruited from the UK electoral register and familiarised with the standard gamble (bottom-up titration variant). Health state descriptions were developed from disease-specific quality of life measures and presented in bullet format.

RESULTS 83 of the panel members constributed at least one value. 4,054 preferences were obtained on 73 separate health states, collected in 18 sets of valuation tasks over 18 months. Thirteen specific conditions and EQ5D states were depicted. Participants provided an average of 50 values during the project. No consistent non-gamblers were identified, although four participants were unwilling to take the gamble option at any risk of death in more than 20% of valuations. Overall, 4.7% of preferences for states were 1.0. Interestingly, no negative values were obtained. Mean health state values ranged from 0.37 to 0.99 and were broadly consistent with values from other studies on similar states. The most notable exception was in EQ5D states. The panel's data was highly correlated with results of the UK EQ5D valuation study on the same states, but showed only moderate agreement with a possible upward bias in more severe states greater than would be expected by methodological variation in preference elicitation techniques.

CONCLUSION It is feasible to use an internet panel to provide health state valuations in response to the needs of economic evaluations, and the approach provides an opportunity to explore the determinants of preferences across a wide range of health states.