BACKGROUND: Two types of mapping exercises have been suggested to translate data from the short-form health instrument SF-12 into measurements for the EQ-5D instrument. One of the trends associates SF-12 data directly to the EQ-5D index, whereas a second approach suggests a response mapping method where SF-12 data have been mapped to the five domains of the EQ-5D questionnaire. OBJECTIVES: To improve the current response mapping approach and provide a comparison between the direct and the response mapping approaches to the EQ-5D. METHODS: Five multinomial logit regressions were implemented to estimate the association between SF-12 variables and each of the EQ-5D domains. Predicted EQ-5D responses were estimated using a Monte Carlo method. A parameter uncertainty approach was introduced to calculate confidence intervals for the predicted EQ-5D index. The direct mapping approach for the comparison was also conducted. Several large datasets were used for internal and external validation. Actual versus predicted EQ-5D index were compared using mean square error (MSE), mean absolute error (MAE) and confidence intervals. RESULTS: In the internal validation dataset, both the response and direct mapping predicted a similar mean EQ-5D index but the response mapping yielded a smaller MSE of 0.018 compared to 0.020 in the direct mapping, and a smaller MAE of 0.091 in the response mapping compared with 0.105 in the direct mapping method. Using the external validation dataset, the MSE and MAE were one decimal point less in the response compared to the direct mapping. CONCLUSIONS: The revised response approach provided marginally better results than the direct mapping method. The response method can be implemented to country-specific EQ-5D data with available value sets. To facilitate the implementation of the revised response mapping algorithm, a Stata command has been programmed.