OBJECTIVES: The study was to examine the impact of health state selection on EQ-5D valuation using the recently collected Korean valuation dataset. Most of valuation studies concentrate on the effect of eliciting preference method or elicited subject. Little attention has been paid to the selection of health state. Unlike majority of EQ-5D valuation studies, the Korean valuation study collected values for 101 out of total 243 possible health states using time-tread-off method in a national representative sample of 1307 individuals. The objective of the study is to investigate the effect of number of health state in estimating EQ-5D value set. METHODS: A simulation approach was adopted. Fourteen categories for the number of states were distinguished: all 101 health states, 90, 80, 70, 60, 50, 45, 40, 35, 30, 25, 20 and 15 states randomly selected out of the total 101 states. For each investigated category, 100 random samples were drawn from the dataset. Note that the category with all 101 health states, only one sample was possible. For each sample, the random effect regression was applied to estimate coefficients for calculating the predicted values for all 101 health states. The evaluation criterion for performance of each category was the average value of mean absolute errors (MAE) of 100 samples. RESULTS: The average value of MAE for each category was 0.0416, 0.0417, 0.0419, 0.0421, 0.0424, 0.0430, 0.0434, 0.0439, 0.0450, 0.0460, 0.0474, 0.0517, and 0.0699. As the number of health states decreased, the size of average MAE increased. The trend, however, was relatively stable until the number of health state drop into 30 (elbow point). CONCLUSIONS: The threshold for number of health state required, in order to estimate effectively and obtain reliable EQ-5D value set, was 30 based on the current finding.  Such threshold, however, could vary depending on the chosen criteria.