OBJECTIVES: The valuation of health is fundamental to measuring health outcomes. Health status index measures such as HUI, QWB and EQ-5D typically assign a value of 1 to full health and a value of zero to a state equivalent to dead. Values for intermediate health are generally not known in advance and may be the subject of investigation through surveys of the general population. These tend to be expensive in terms of labour and cost. An alternative approach to generating national value sets for EQ-5D was developed by adjusting pre-existing value sets from “donor” countries using domestic population health survey data. METHODS: Self-rated health status was collected as part of the US Medical Expenditure Panel Survey in 2001 and 2003 using visual analogue scale rating within EQ-5D. Combined data from some 38,000 respondents, of whom 4,200 were Spanish-speaking Hispanics, were analysed. These data yielded values for 25 real EQ-5D health states. Regression analysis was used to linking observed values for these EQ-5D states in the two subgroups. This model was then applied to a larger dataset for 43 states that had been valued as part of a US national survey designed to construct values for hypothetical EQ-5D states. This 2-stage procedure yielded a mean set of values calibrated on the preferences of Spanish-speaking Hispanics RESULTS: Self-rated VAS scores for Spanish-speaking Hispanics [VASh] were linked via an estimation function VASh = 0.869*VASw + 7.262, where the independent term is the observed VAS score from non-Hispanic respondents (R2 = 0.90). The mean absolute difference between estimated and observed scores for the 43 hypothetical health states was 1.73. CONCLUSION: The results yield a viable proxy estimate of social preference weights for EQ-5D derived from a US Hispanic population who speak Spanish. The extent of its generalisability remains to be tested elsewhere.