OBJECTIVES: The objective was to develop utility tariffs for Malaysia using expressed health preference data from a patient population. METHODS: Health Preference survey data was collected in 2004/5 from dialysis patients, their careers and dialysis centre staff (N=153) at 9 Ministry of Health hospitals in Malaysia. Forty-five health vignettes from the EQ-5D descriptive system (5 health dimensions by 3 severity levels per dimension) were presented to respondents who each valued a set of 15 vignettes on a visual analogue scale (VAS) that ranged from 0 to 100 (worst to best health imaginable). A linear regression model was used to generate predicted utility. It incorporated all EQ-5D health dimensions and a D1 interaction term that represented the number of departures from perfect health beyond the first. The model predictions were compared against health valuations by 686 dialysis patients drawn from the same population. RESULTS: The model covariates possessed properties found in tariff valuation models developed elsewhere and were statistically significant with R2=0.752. Covariates possessed the right sign (negative) and the worst severity levels had greater negative values than moderate severity levels. Among a test set of 654 dialysis patients, utility values changed as expected with differences in age. Utility declined from 0.90 to 0.88 and 0.84 for patients under 40, 40 to 55, and above 55  years respectively while the equivalent mean raw VAS scores were 76, 77 and 74. CONCLUSIONS: The utility tariff developed from the D1 model performed well statistically and appeared to provide better utility estimates than raw VAS scores.