Mapping the NEI-VFQ-25 scores to the EQ-5D-Y-3L and CHU-9D utility values among adolescents with myopia.

Abstract

Purpose: To develop mapping algorithms for predicting EQ-5D-Y-3L and CHU-9D utility values from the National Eye Institute 25-Item Visual Function Questionnaire (NEI-VFQ-25) among adolescents with myopia to support economic evaluation.

Methods: Data were collected from 2,198 adolescents with myopia in China. Conceptual overlap was assessed using Spearman's rank correlation. Six regression models, including ordinary least squares (OLS), Tobit, censored least absolute deviations (CLAD), generalized linear model (GLM), two-part model (TPM), and adjusted limited dependent variable mixture model (ALDVMM), were evaluated. The predictors included the NEI-VFQ-25 total score and the subscale scores. Model performance was assessed via five-fold cross-validation and random sample validation based on mean absolute error (MAE), root mean squared error (RMSE), Akaike information criterion (AIC), Bayesian information criterion (BIC), and the proportions of absolute error (AE) > 0.05. Sensitivity analyses of model validation were conducted.

Results: The mean (SD) utility values were 0.962 (0.070) for the EQ-5D-Y-3L and 0.851 (0.160) for the CHU-9D. The NEI-VFQ-25 total score correlated moderately with the CHU-9D (r = 0.50) but weakly with the EQ-5D-Y-3L (r = 0.37). The OLS and GLM using stepwise-selected subscale scores demonstrated the best predictive performance for EQ-5D-Y-3 L (MAE = 0.045, RMSE = 0.061, AE > 0.05 [%] = 30.37%) and CHU-9D (MAE = 0.103, RMSE = 0.139, AE > 0.05 [%] = 61.50%), respectively. Sensitivity analyses generally supported the primary model-selection results.

Conclusion: This study established algorithms for estimating EQ-5D-Y-3L and CHU-9D utility values based on NEI-VFQ-25 subscale scores for adolescents with myopia, thus facilitating future economic evaluation for myopia interventions. However, given the limited conceptual overlap between the NEI-VFQ-25 and EQ-5D-Y-3L, the EQ-5D-Y-3L algorithm should be applied with caution.