Updated algorithms for direct and indirect mapping of WHOQOL-BREF to EQ-5D-5L utility scores using data from a...

Mapping techniques can estimate EQ-5D-5L utility scores from non-preference-based WHOQOL-BREF responses. A previous Thai study used a reciprocal linear model, but it may not adequately address the left-skewed distribution of utility scores. This study aimed to develop improved mapping algorithms using alternative regression models that better account for the distributional characteristics of utility scores in the Thai general population.

A 2022 national survey dataset of paired WHOQOL-BREF and EQ-5D-5L responses (n = 2,000), representative of the Thai general population, was used. Five predictor sets were fitted using eight regression models, including ordinary least squares, Tobit, censored least absolute deviations, generalized linear model (GLM), two-part models, adjusted limited dependent variable mixture model, and beta regression-based mixture model for the direct mapping approach; and multinomial logistic regression (MLOGIT) for the indirect mapping approach. The best-performing models were identified based on the lowest ten-fold cross-validated mean absolute error (MAE) and root mean square error (RMSE), together with minimal prediction bias from graphical assessment.

The GLM-poisson model, fitted using a predictor set comprising age, general WHOQOL-BREF items, and selected Physical Health, Psychological Health and Environment items, emerged as the best-performing model, with predictions closely aligning with observed values. A MLOGIT model performed less well, with overpredictions for utility scores > 0.6.

This study advances the literature on mapping WHOQOL-BREF to EQ-5D-5L by providing updated algorithms for estimating EQ-5D-5L utility scores from WHOQOL-BREF responses in the Thai general population. The improved algorithms offer more accurate predictions, supporting their use in health economic analyses in Thailand.