OBJECTIVE: The purpose of this study was to conduct a statistical mapping between standard assessments of disease severity in atopic dermatitis (AD) and a generic quality of life measure and use optimised algorithms to estimate health utility. METHODS: Multinomial logistic regression was used to estimate response probabilities to the SF-12 health survey from assessment of AD severity using the modified Eczema Area & Severity Index (mEASI) in patients enrolled into a clinical trial studying the long-term use of tacrolimus ointment. A random sample of 70% of patient responses was used for the regression analysis, with the remaining 30% used to test the predictive ability of the response mapping. Predicted and actual SF-12 responses were converted to SF-6D and EQ-5D health utilities using published algorithms (Brazier 2004 and Gray 2006) by Monte Carlo simulation. RESULTS: Evaluable data were available for 255 patients, 40% of whom were male, and for which the median age at screening was 28 years (IQR 22 to 38). The percentage of variance in item response attributable to predictor variable change ranged from 19.8% (SFCALM) to 37.5% (SFENER). The mean predicted utility in the test dataset was 0.799 (sd 0.083) by the SF-6D method and 0.778 (0.203) by EQ-5D. The mean squared error (MSE) between the actual and predicted utilities were 0.012 for SF-6D and 0.075 for EQ-5D. Predicted SF-6D utility was within 10% of the actual utility in 64.8% of cases while for predicted EQ-5D utility this fell to 39.9%. CONCLUSION: In this study response mapping of AD symptom severity to SF-12 produced more reliable estimates of SF-6D utility than EQ-5D utility. In young adults with low co-morbidity the sensitivity of SF-6D at higher utilities make it be a more appropriate choice than EQ-5D where cases would otherwise be classified in perfect health.