The National Eye Institute developed a visual functioning questionnaire (NEI-VFQ) designed to assess health-related quality of life of patients with visual impairments. The developers of the NEI-VFQ distributed the original 52 items into 13 different domains. The recommended method for scoring the NEI-VFQ is to linearly transform the sum of the ordinal ratings to each item within each domain to produce 13 scores. The major shortcoming of this scoring method is that sums of ordinal numbers do not necessarily generate valid measurement scales. However, Rasch models can be used to estimate interval measurement scales from ordinal responses to items. We administered 27 items from the 52-item NEI-VFQ to 341 patients with low vision. Rasch analysis was used to estimate the ‘visual ability’ required by each item for a particular response (item measures) and to estimate the ‘visual ability’ of each patient (person measures). The validity of the model was evaluated by examining the distributions of residuals for item and person measures. We observed that the 17 items we tested from the NEI-VFQ that require difficulty ratings produce a valid interval scale for low-vision patients. The estimated person measures of visual ability are linear with log MAR acuity. The ten items that require frequency or level of agreement ratings do not work together to produce a valid interval scale. Rather, these items appear to be confounded by other variables distributed in the patient sample (e.g. psychological state). The visual ability scale estimated from the 17 NEI-VFQ items is proportional to the visual ability scales estimated from two earlier studies that also elicited difficulty ratings from low-vision patients.
