OBJECTIVES: This study was conducted to obtain a Korean EQ-5D valuation set with the time-trade off method from a representative sample of South Korean population.  METHODS: We selected ‘33333’ state and 100 health states that were much more than in previous studies to minimize the interpolation space. These 100 health states include 25 mild states, 50 moderate states, and 25 severe states according to the proposal of EuroQol group. We interviewed 1307 residents using the Korean version of EQ-5D questionnaire and TTO method. Each respondent valued ‘33333’ state and 12 health states. The exclusion criteria for valuation data as follows: if less than 3 states were valued; if all states were valued the same; if all states were valued worse than dead; if there were 4 or more logical inconsistencies. The final valuation sample included 1170 respondents. We consider several interaction terms as well as main effect term in modeling. RESULTS: Main effects model with random effects had a best goodness-of-fit for the observed values with a mean absolute error (MAE) of 0.0416. Because 33333 state was included for all respondents, there will be a larger volume of information for this health state than for others. Therefore we also used alternative approaches excluding the state 33333 from the model. The MAE of this model was identical with that of original model. CONCLUSIONS: We selected the main effects model by alternative approaches excluding 33333 state. The Korean valuation formula of EQ-5D are as follows; TTO value=0.995-(0.1175*M2)-(0.626*M3)-(0.0601*SC2)-(0.2051*SC3)-(0.0637*UA2)-(0.2788*UA3)-(0.0601*PD2)-(0.2307*PD3)-(0.071*AD2)-(0.2329*AD3).