OBJECTIVES: This study aimed to develop a function to predict the SF-6D index score from SF-8 scores. METHODS: This study was a secondary analysis of data collected in a population health survey in Singapore in which respondents (n=7,529) completed both the SF-36 and the SF-8 questionnaires. Four multiple linear regression models with different specifications were compared for their performance in predicting SF-6D scores calculated using responses to SF-8 items.  RESULTS: The model in which responses to all SF-8 items were coded into dummy variables achieved the best prediction outcomes. The model explained 62% (other models: 56% to 62%) of the variance in the SF-6D index score with the mean absolute error being 0.056 (other models: 0.056 to0.06100) and root mean square error being 0.076 (other models: 0.077 to 0.083). The absolute error between predicted and observed SF-6D scores was less than 0.1 and 0.05 among 84% and 59% of the respondents, respectively; this results for other modes were 77% to 84% and 51% to 56%.  CONCLUSIONS: It is possible to generate a utility-based index score from the SF-8. The function developed in this study should be further tested in other populations.