PurposeTo derive a mapping algorithm to predict SF-6D utility scores from the non-preference-based LupusQoL and test the performance of the developed algorithm on a separate independent validation data set.
MethodLupusQoL and SF-6D data were collected from 320 patients with systemic lupus erythematosus (SLE) attending routine rheumatology outpatient appointments at seven centres in the UK. Ordinary least squares (OLS) regression was used to estimate models of increasing complexity in order to predict individuals' SF-6D utility scores from their responses to the LupusQoL questionnaire. Model performance was judged on predictive ability through the size and pattern of prediction errors generated. The performance of the selected model was externally validated on an independent data set containing 113 female SLE patients who had again completed both the LupusQoL and SF-36 questionnaires.
ResultsFour of the eight LupusQoL domains (physical health, pain, emotional health, and fatigue) were selected as dependent variables in the final model. Overall model fit was good, with R 2 0.7219, MAE 0.0557, and RMSE 0.0706 when applied to the estimation data set, and R 2 0.7431, MAE 0.0528, and RMSE 0.0663 when applied to the validation sample.
ConclusionThis study provides a method by which health state utility values can be estimated from patient responses to the non-preference-based LupusQoL, generalisable beyond the data set upon which it was estimated. Despite concerns over the use of OLS to develop mapping algorithms, we find this method to be suitable in this case due to the normality of the SF-6D data.3