Abstract
Purpose :
We applied Rasch analysis to National Eye Institute-Visual Function Questionnaire (NEI VFQ-25) data from the Age-Related Eye Disease Study (AREDS) to estimate calibrated item and person measures.
Methods :
AREDS, a randomized clinical trial (1992-98) to evaluate the effect of nutritional supplements on risk of developing late age-related macular degeneration (AMD), included yearly visits with color fundus photography graded for AMD by a masked central reading center; the NEI VFQ-25 was administered at study visits after 1997. We included participants with at least two visits with NEI VFQ-25 data and AMD gradings and applied the method of successive dichotomizations (MSD) (Bradley et al, 2018, R package 'msd') to estimate item and person measures (M2C) and standard errors.
Results :
3619 individuals (56% female; mean age 73 y; better-eye visual acuity (VA) mean, 83 letters (~20/25); better-eye AMD none (28%), early (22%), intermediate (26%), or late (23%)) were included. The MSD cannot estimate person measures for individuals who endorse all NEI VFQ-25 items at the highest ability level (n=281, 8%); this group was significantly younger (72.0 vs 73.0, p<.001), less likely to be female (47% vs 57%, p<.001) or to have late AMD (4% vs 25%, p<.001), and had better VA (87 vs 82 letters, p<.001) and higher measures on the NEI VFQ-25 overall composite score (97.8 vs 86.8, p<.001). Estimated person measures (logit scale) ranged from -3.67 to 5.44; item measures ranged from -2.20 to 3.70. Many participants were at the ceiling of the item measures (Figure 1), indicating a lack of items with sufficient difficulty to characterize the visual ability of this population. Item measures derived from the AREDS population were generally consistent in rank order (p=0.5) with those estimated by Goldstein et al (2022). We compared the M2C person measures with the NEI VFQ-25 composite score (Figure 2); for the subgroup with greater visual ability ((i.e., NEI VFQ-25 scores >90), the M2C person measures appeared to capture a wider range of visual ability than the NEI VFQ-25 composite score alone.
Conclusions :
Use of item calibrations on the NEI VFQ-25 may improve its measurement properties; however, in this group of patients with AMD, there was not sufficient item difficulty to further characterize persons with the highest visual ability.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.