OBJECTIVES: There already exists an algorithm to generate utility scores from SF-36 data. This has a serious floor affect and does not accurately describe EQ-5D data. The purpose of this study was to develop an alternative model that better accommodates utility scores from the entire range of ill-health. METHODS: Data on 1,946 inpatients and outpatients were abstracted from the Health Outcomes Data Repository (HODAR) in Cardiff, UK and utilised in the model building. Validation of the model was made using independent data from 554 respondents from a survey of people with diabetes in Cardiff that included the EQ-5D, SF-36. Both surveys also included complete inpatient records, diagnoses, procedures, in and out of hospital mortality, biochemistry, unit costs, inpatient medications, risk factors and demography. Various multivariate parametric and non-parametric techniques were applied to calculate an optimised model on the HODAR data set. RESULTS: When applied to the diabetes data, the calculated and actual values were as follows. Data are the minimum and maximum utility values, the inter-quartile range, median, mean, and standard deviation, respectively. Results from the survey were as follows: -0.48, 1, 0.52, 0.85, 0.71, 0.65 and 0.32, respectively. For the existing model, 0.26, 1, 0.61, 0.85, 0.74, 0.72 and 0.16, respectively. For the revised model, 0.05, 1, 0.52, 0.94, 0.76, 0.71 and 0.25, respectively. CONCLUSION: This new algorithm represents a notable improvement on the existing model. Proper evaluation of new medicines is vital, and the floor effect tends to reduce incremental cost utility ratios when achieving certain thresholds is important for drug reimbursement. HODAR will evolve to be the largest series of outcome data anywhere.