Predicting EQ-5D Values Using the SGRQ

Abstract

Objectives The purpose of this study was to develop and validate an algorithm that predicts EQ-5D utility from the St. George's Respiratory Questionnaire (SGRQ) in subjects with chronic obstructive pulmonary disease and to examine the effect of using this algorithm in predicting quality-adjusted life-years (QALYs). Methods In the TORCH (Towards a Revolution in COPD Health) trial, the SGRQ and EQ-5D were administered at baseline and every 24 weeks for 3 years. To map EQ-5D utility from the SGRQ, ordinary least squares (OLS), generalized linear models (GLMs) and two-part models were used. Algorithms were developed in a fitting sample and used to predict utility scores in a validation sample and selected based on root-mean-square error (RMSE). QALYs were estimated from the algorithm and compared to QALYs derived from EQ-5D utility scores collected in the trial. Results A simple OLS algorithm was found to perform as well as algorithms developed using more complex modeling structures. The resulting model was (RMSE 0.1723): EQ-5D = 0.9617 − 0.0013 × SGRQ total − 0.0001 × SGRQ total 2 + 0.0231 × male. Ordering of treatments by QALY gain was dependent on the method of utility estimation. Conclusion A mapping algorithm can be used to predict EQ-5D utility scores from the SGRQ and may be useful in some situations; however, for use in a health technology assessment (HTA) submission in which precision of estimation is important, it is in the interests of both the manufacturer and the HTA body that utility scores be directly derived from the clinical trial population.