OBJECTIVE: The objective of this study is to select a suitable EQ-5D value set model for Taiwan population. METHODS: Data for valuation of EQ-5D health states were obtained via three methods: a) through postal survey value analog scale (VAS); using the standard EQ-5D format; A random sampling 12,924 adults was surveyed. b) VAS and c) Time Trade-Off (TTO) interview methods were used to interview patients and care givers in a medical center. There were 228 adults interviewed. Four methods were used to exclude responses from individuals to better fit the model. METHODS A: According to EuroQol exclusion criteria. B: when A+(7). C: when A+(5)+(6). D: C+(7). The study used four criteria to select the proper EQ-5D model: the higher adjust R2, smaller actual and estimated value difference, higher correlation with EuroQol model, larger sample size, the preference value (PV) graph had the similar pattern as EuroQol value set chart. RESULTS: In VAS method either by postal survey or interview, the study indicate that D method was proper and the PV adjusted R2 achieving 0.53~0.57. In TTO interview method indicate that all A~D exclusion criteria results in the preference value adjusted R2 of 0.48. In N2N3/VAS/D and N2N3/TTO/A models the actual and estimated value difference was smaller and higher correlation with EuroQol model than N3/VAS/D and N3/TTO/A. The N2N3/VAS/D and N2N3/TTO/A model had the similar pattern as EuroQol value set chart. CONCLUSIONS: The N2N3/VAS/D/Postal and N2N3/TTO/A models adjusted R2 achieve 0.53~0.48 and the correlation with EuroQol model achieve 0.907~0.966. The two models can be used to interpolate values of the states for which there is no direct observation.