OBJECTIVES: To carry out cost utility analysis, utility values can be derived using generic preference based measures such as EQ-5D or SF-6D.  In some settings generic measures are not used, and mapping functions are being developed to predict utility scores from condition specific measures.  The aim of this study is to map the DHP-18 - a diabetes-specific HRQoL patient reported outcome measure - onto EQ-5D and SF-6D utility scores for type 1 and type 2 diabetes mellitus populations.  METHODS: The data used was pooled from a longitudinal study of quality of life in diabetes.  OLS, GLS and Tobit models regressing DHP dimensions and, separately, DHP items onto EQ-5D and SF-6D index scores for both type 1 (n=236) and type 2 (n=2358) diabetes populations were applied . RESULTS: For both the EQ-5D and SF-6D, the GLS model mapping selected DHP-18 item scores, squared item scores, age and gender onto the utility index provided the best fit, and this was the case for both the type 1 and type 2 populations (R2 EQ-5D type 1: 0.516; EQ-5D type 2: 0.290; SF-6D type 1: 0.647; SF-6D type 2: 0.396).  The models under predict utility when the state is severe and over predict when the state is mild.  The error associated with the models was lower for SF-6D than for EQ-5D due to differences in the range of the measures. CONCLUSIONS: The DHP-18 items can predict both the EQ-5D and SF-6D utility scores with acceptable precision with the mapping algorithm for the SF-6D displaying a higher level of precision.  The mapping functions developed from the models can be used to predict utility scores in settings where the EQ-5D or SF-6D have not been used alongside the DHP-18. However mapping should be considered second best in comparison to using generic measures in research studies.