OBJECTIVES: The objective of this study was to assess the accuracy of two methods for predicting EQ-5D-5L utilities from EQ-5D-3L data.

METHODS: The research employed two data sets: an observational study of 100 patients treated for non-small cell lung cancer (NSCLC) and a clinical trial of 168 patients treated for irreversible pulpitis. The EQ-5D-3L and EQ-5D-5L were administered at multiple time points in both studies. EQ-5D-3L data were mapped to English EQ-5D-5L utilities using algorithms developed by the National Institute for Health and Care Excellence Decision Support Unit (DSU) and EuroQol Group. The accuracy of the two algorithms was evaluated by comparing predicted utilities and associated quality-adjusted life years (QALYs) against utilities and associated QALYs derived from EQ-5D-5L data.

RESULTS: Due to differences in algorithm covariates as well as missing data, the NSCLC study sample size was smaller for both utilities and QALYs when applying the DSU algorithm as compared to the EuroQoL algorithm. The mean absolute error and root mean square error of predicted utilities were marginally lower for the DSU algorithm (0.088, 0.116) than the EuroQoL algorithm (0.107, 0.129). Similar results were obtained when basing comparisons on QALYs. The results appeared to be influenced by the limited distribution of EQ-5D-3L health states (i.e., 36% of observations were distributed across only four health states). Additional analyses explored alternate specifications of the EuroQol mapping algorithm as well as the influence of age, gender, and missing data. Results of analyses of the dental trial data will be presented at the meeting.

CONCLUSIONS: In an observational study including paired EQ-5D-3L and EQ-5D-5L data, the DSU algorithm yielded marginally more accurate predictions than the EuroQol algorithm, though differences were slight. More research is needed to benchmark the two algorithms against the EQ-5D-5L utility index and understand implications for health economic modeling.