Psychometric properties comparison of EQ-5D-Y-3L and CHU-9D in children with adenoid hypertrophy.

Abstract

Purpose: This study aims to assess the psychometric properties of two preference-based instruments, EQ-5D-Y-3 L and CHU-9D, in assessing the health-related quality of life of children with adenoid hypertrophy.

Methods: EQ-5D-Y-3 L and CHU-9D were completed in a random sequence by children with adenoid hypertrophy (aged 8-15years) from three hospitals in XX city. The instruments were systematically compared for distribution, ceiling effects, agreement, construct validity, known‑groups validity, and discriminative ability (via AUC, relative efficiency, orthogonal polynomial contrasts, and ANCOVA) in distinguishing disease severity levels.

Results: This study compared the performance of the EQ-5D-Y-3 L and CHU-9D in 644 children with adenoid hypertrophy (mean age 10.63 years). The two instruments showed poor agreement (ICC = 0.691). In discriminating disease severity, the CHU-9D demonstrated superior sensitivity, with larger effect sizes (η²=0.257), and its relative efficiency suggested approximately 1.39 times greater overall discriminative ability (reference: EQ-5D-Y-3 L = 1). However, pairwise Cohen's d values were small for both instruments (range:0.108-0.163). ANCOVA confirmed these advantages persisted after adjusting for covariates.

Conclusions: The EQ-5D-Y-3L and CHU-9D were not interchangeable in children with adenoid hypertrophy. The CHU-9D showed superior discriminatory sensitivity, larger global effect sizes, and better model fit, particularly in detecting mild to moderate disease changes. Therefore, in health economic evaluations, the CHU-9D may be more suitable when the study aim is to identify early-stage or subtle health changes. In contrast, the EQ-5D-Y-3L provides a more stable model for assessing a consistent linear relationship between disease severity and health utility. Researchers should select the instrument based on clearly defined study objectives, while acknowledging that the observed pairwise effect sizes were modest and warrant further validation.