Mapping KDQOL-36 Onto EQ-5D-5L and SF-6Dv2 in Patients Undergoing Dialysis in China

Objectives

To develop mapping algorithms based on Kidney Disease Quality of Life-36 (KDQOL-36) scores to the EQ-5D-5L and SF-6Dv2 utility values among dialysis patients in China.

Methods

We used data from a cross-sectional multicenter survey in China to map the KDQOL-36 to the EQ-5D-5L and SF-6Dv2. The conceptual overlap between the KDQOL-36 and the EQ-5D-5L or SF-6Dv2 was evaluated using Spearman’s correlation coefficients. Direct mapping, including ordinary least squares, generalized linear model, beta regression model, Tobit regression model (TRM), censored least absolute deviations model, adjusted limited dependent variable mixture model (ALDVMM), response mapping, and seemingly unrelated ordered probit model, were used to derive mapping functions using the data set. Model performance was assessed by the mean absolute error (MAE) and root mean square error (RMSE) using cross-validation.

Results

A total of 378 patients (50.53% female; mean [SD] age: 49.05 [13.34]) were included in this study. The mean utility values of EQ-5D-5L and SF-6Dv2 were 0.72 and 0.57, respectively. When mapping to the EQ-5D-5L, the ALDVMM with 1 component was the best-performing model (MAE = 0.1579, RMSE = 0.2289). When mapping to SF-6Dv2, TRM was the best-performing model (MAE = 0.1108, RMSE = 0.1508). Generally, KDQOL-36 subscale scores and their squares were the optimal predictor set for each model. Overall, the models using KDQOL-36 subscale scores showed a better fit than those using the Kidney Disease Component Summary.

Conclusions

The ALDVMM and TRM models with the KDQOL-36 scores can be used to predict the EQ-5D-5L and SF-6Dv2 utility values, respectively, among dialysis patients in China.