Direct and indirect mapping of the 12-item Short Form Survey version 2 (SF-12v2) onto the EQ-5D-5L utility scores in general Thai population

Abstract

The SF-12v2 is widely used to measure health-related quality of life (HRQoL) in the general Thai population because it is brief and places a low burden on respondents. However, it does not provide utility scores required for economic analyses. The mapping approach is a solution to estimate the utility scores from the SF-12v2 response using several regression models. Therefore, this study aimed to develop a mapping algorithm to estimate EQ-5D-5L utility scores from the SF-12v2 items using a 2022 national dataset of 2000 Thai respondents. Four predictor sets incorporating SF-12v2 items/subscales and age as a covariate were investigated using direct and indirect mapping approaches. Direct mapping approaches included ordinary least squares, Tobit, censored least absolute deviations, generalized linear model (GLM), two-part models, adjusted limited dependent variable mixture model (ALDVMM), and beta mixture model, while multinomial logistic regression (MLOGIT) was investigated for indirect mapping. Model performance was evaluated using ten-fold cross-validated (post-CV) mean absolute error (MAE) and root mean square error (RMSE). The ALDVMM-1 component with a predictor set comprising age and selected SF-12 items as categorical variables demonstrated the best predictive performance, yielding the lowest post-CV MAE and RMSE (0.0410 and 0.0625, respectively). However, MLOGIT showed poorer predictive performance. Therefore, the proposed ALDVMM-1 component-based mapping algorithm may facilitate the estimation of utility scores in studies where only the SF-12v2 was collected. Given that the mapping algorithm was developed from a predominantly healthy sample, it may be less reliable in populations with poorer health status, and further validation in populations with less healthy or clinical samples is warranted in future studies.