Abstract
PURPOSE: To improve the usefulness of the National Eye Institute Visual Function Questionnaire-25 (NEI VFQ-25) by enabling estimation of measures on an invariant scale and comparisons between patients and across studies. METHODS: Datasets of baseline NEI VFQ-25 responses from nine studies (seven retina randomized trials, n = 2770; two low vision studies, n = 572) were combined. The method of successive dichotomizations was applied to patient ratings of the main NEI VFQ-25 and six supplemental items to estimate Rasch model parameters using the R package 'msd.' Calibrated item measures and rating category thresholds were estimated for the NEI VFQ-25, as well as for two domain-specific versions: the NEI VFQ-VF that includes only visual function items and the NEI VFQ-SE that includes only socioemotional items. RESULTS: Calibrated item measures were estimated from study participants (n = 3342) ranging in age from 19 to 103 years, with mean (SD) age of 69.3 (11) years and a mean logMAR visual acuity of 0.30 (Snellen 20/40). Item measure estimates had high precision (standard error range, 0.026-0.085 logit), but person measure estimates had lower precision (standard error range, 0.108-0.499 logit). Items were well targeted to most persons, but not to those with higher levels of function. CONCLUSIONS: Calibrated item measures and rating category thresholds enable researchers and clinicians to estimate visual, socioemotional, and combined measures on an invariant scale using the NEI VFQ-25. TRANSLATIONAL RELEVANCE: Applying NEI VFQ 25C calibrated item measures (software provided) to the NEI VFQ-25, users can estimate overall, visual, and socioemotional function measures for individual patients.