Abstract
BACKGROUND: Multiple diseases, such as Adolescent Idiopathic Scoliosis (AIS), present at adolescent age and the impact on quality of life (QoL) prolongs into adulthood. For the EQ-5D, a commonly used instrument to measure QoL, the current guideline is ambiguous whether the youth or adult version is to be preferred at adolescent age. To assess which is most suitable, this study tested for equivalence along predefined criteria of the youth (EQ-5D-5L) and adult (EQ-5D-Y-5L) version in an adolescent population receiving bracing therapy for AIS. METHODOLOGY: 107 adolescents were recruited from 4 scoliosis centers in the Netherlands between March 2022 and January 2023; they completed both EQ-5D's and the SRS-22r (scoliosis-specific questionnaire). The following criteria were evaluated using the individual and sum of domains (level-sum-score (LSS)). Our primary criterion for non-equivalence of the EQ-5D's was less than excellent (≤ 0.9) intra-individual agreement using Intraclass Correlation Coefficient (ICC) analysis for LSS and weighted (quadratic) kappa for domains. Secondary criteria were differences in ceiling using McNemar test; a different number of quantified hypotheses for construct validity achieved using the SRS-22r as comparator; differences in test-retest reliability by comparing ICC/kappa values using a Z-test. RESULTS: Adolescents had a mean age of 14 years (range 12-18), and 78% were female. Ceiling was mostly comparable between EQ-5D's, ranging from 78 to 81% for mobility and self-care, 52-54% for usual activities, and 31-36% for pain/discomfort. The EQ-5D-5L showed more ceiling (57%) compared to the EQ-5D-Y-5L (41%) on anxiety/depression (p = 0.006). Agreement between the EQ-5D's did not meet our criterion for the LSS (ICC 0.79 (95% confidence interval 0.70-0.85)), and decreased further at the domain-level. Both EQ-5D's achieved 5/7 validity hypotheses. Test-retest reliability was slightly better for EQ-5D-5L LSS (ICC 0.76 (0.64-0.84)) compared to EQ-5D-Y-5L LSS (ICC 0.69 (0.55-0.79)), although this was statistically insignificant (p = 0.284). This pattern was similar for most domains. CONCLUSIONS: The EQ-5D versions showed insufficient agreement, and cannot be considered fully equivalent. While they were similar in terms of validity and test-retest reliability, differences in score distribution were present. Taken together, we advise using the EQ-5D-5L to monitor the QoL in adolescent patients with AIS, as it avoids switching instruments and thus data discontinuities. Future studies should verify these findings in different patient groups and the general population.