BACKGROUND/ PURPOSE: EQ-5D (EuroQol-5 dimensions) is a preference-based measure of health, which is widely used in cost-utility analyses. It has been suggested that each country should develop its own value set. We therefore sought to develop the quality weights of the EQ-5D health states with the time trade-off (TTO) method in Taiwan. METHODS: A total of 745 respondents consisting of employees and volunteers in 17 different hospitals were recruited and interviewed. Each of them valued 13 of 73 EQ-5D health states using the TTO method. Based on the three exclusion criteria for valuation data, only 456 (61.21%) respondents were considered eligible for data analysis. The quality weights for all EQ-5D health states were modeled by generalized estimating equations (GEEs). RESULTS: Over half of the responses were given negative values, and the medical personnel seemed to have a significantly higher TTO value (+0.1) than others after controlling for other predictors. The N3 model (level 3 occurred within at least 1 dimension) yielded an acceptable fit for the observed OTT data [mean absolute error (MAE) = 0.056, R(2) = 0.35]. The magnitude of mean absolute differences (MADs) between Taiwan data and those from the UK, Japan, and South Korea ranged from 0.146 to 0.592, but the rank correlation coefficients were all above 0.811. CONCLUSION: This study reaffirms the differences in health-related preference values across countries. The high proportion of negative values might indicate that we have also partially measured the intensity of fear in addition to the utility of different health states.
BACKGROUND/ PURPOSE: EQ-5D (EuroQol-5 dimensions) is a preference-based measure of health, which is widely used in cost-utility analyses. It has been suggested that each country should develop its own value set. We therefore sought to develop the quality weights of the EQ-5D health states with the time trade-off (TTO) method in Taiwan. METHODS: A total of 745 respondents consisting of employees and volunteers in 17 different hospitals were recruited and interviewed. Each of them valued 13 of 73 EQ-5D health states using the TTO method. Based on the three exclusion criteria for valuation data, only 456 (61.21%) respondents were considered eligible for data analysis. The quality weights for all EQ-5D health states were modeled by generalized estimating equations (GEEs). RESULTS: Over half of the responses were given negative values, and the medical personnel seemed to have a significantly higher TTO value (+0.1) than others after controlling for other predictors. The N3 model (level 3 occurred within at least 1 dimension) yielded an acceptable fit for the observed OTT data [mean absolute error (MAE) = 0.056, R(2) = 0.35]. The magnitude of mean absolute differences (MADs) between Taiwan data and those from the UK, Japan, and South Korea ranged from 0.146 to 0.592, but the rank correlation coefficients were all above 0.811. CONCLUSION: This study reaffirms the differences in health-related preference values across countries. The high proportion of negative values might indicate that we have also partially measured the intensity of fear in addition to the utility of different health states.