OBJECTIVES: Generic health status indexes conventionally require that full health and dead are valued as 1 and 0 respectively. When social preference weights for health states are obtained using a visual analogue scaling (VAS) technique their raw scores often lie on a scale with different endpoints (such as "best" and "worst" health). The transformation of such "raw" scores to a 0-1 scale leads to the exclusion of respondents who fail, for whatever reason, to record a value for dead or full health. This study reports on an alternative approach that does not impose such strict inclusion criteria. METHODS: Data from a postal valuation survey (n=686) conducted in Belgium are used to compare 2 different approaches: (a) values for EQ-5D health states are transformed on a within-respondent basis and median values then computed; (b) median values for EQ-5D health states are computed from the observed data and then transformed to a 0-1 scale. Rankings, health state values and value differences between pairs of states resulting from both methods are compared. RESULTS: The exclusion rate approaches 29% in the conventional approach and 5% in the alternative approach. Transformed median valuations are not significantly different from median transformed valuations. Correlation is 0.994 (p<0.0005). Health state rankings are identical in both approaches. Although the value differences between pairs of states differ between both methods, their correlation is very high (0.998). CONCLUSIONS: Using median original valuations in the analysis of EQ-5D valuation data are theoretically appealing. It builds on the "median voter" model and has the advantage of a lower respondent exclusion rate. Given the need for transparency and the inclusive nature of the mode of analysis, aggregation of observed VAS scores with subsequent transformation is considered an acceptable alternative to the conventional method with its associated data loss.