OBJECTIVES: Chronic obstructive pulmonary disease (COPD) is a debilitating disease. The purpose of this study was to characterise the association between various clinical and quality of life outcome parameters typically used to characterise respiratory illnesses, and utility, as measured by the EQ5Dindex. METHODS: Data were abstracted from the Health Outcomes Data Repository (HODaR), an ongoing survey of patients in a UK region that included the EQ5D and the SF36 and the St. George's Respiratory Questionnaire (StGRQ). Detailed lung function data (LFD) were also available on the same patients for the 12-year period. Generalised linear models were used to estimate utility from predictor variables and goodness of fit was ascertained by the proportion of predictions lying within 0.03 utility units. RESULTS: Complete data were available on 444 subjects. A multinomial regression model mapped the clinical data from 315 (71%) of these subjects to EQ5Dindex and the resulting model was then applied to an independent set of 129 (29%) subjects. In the 315 subjects, 40%, 41% and 52% of predictions were within 0.03 units when using i) the SF36 alone, ii) the SF36 with LFD and iii) the SF36 with LFD and St. George's data, respectively. The same models applied to the test data set resulted in 33%, 33% and 29% of predictions within 0.03 units, respectively. For subjects with moderate to severe COPD (based on the GOLD classification) 49%, 53% and 79% of predictions were within 0.03 units, respectively. The corresponding figures for none or mild COPD (GOLD classification) were 47%, 49% and 86%, respectively. CONCLUSIONS: It was possible to reliably model the EQ5Dindex from the SF36 in respiratory disease; however, models that included the lung function data provided a marginal improvement, and the availability of the StGRQ conferred further improvement within the construction data set.