Remediating serious flaws in the National Eye Institute Visual Function Questionnaire

PURPOSE: To test the assumption that the National Eye Institute Visual Function Questionnaire (NEI VFQ) measures visual functioning, assess the validity of its subscales, and, if flawed, revise the questionnaire and derive a shortened version with sound psychometric properties. SETTING: Flinders Medical Centre, Adelaide, Australia. METHODS: Patients from the cataract surgery waiting list self-administered and completed the 39-item NEI VFQ (NEI VFQ-39). Rasch analysis was applied, and the psychometric performance of the entire questionnaire and each subscale was tested. Instrument revision was performed in the context of Rasch analysis statistics. RESULTS: Five hundred thirty-six patients (mean age 73.8 years) completed the questionnaire. Response categories for 2 question types were not used as intended so dysfunctional categories were combined. The NEI VFQ-39 and the 25-item version (NEI VFQ-25) had good precision but evidence of multidimensionality (more than 1 construct in 1 score), questions that did not fit the construct, suboptimum targeting of item difficulty to person ability, and dysfunctional subscales (8 NEI VFQ-39; 12 NEI VFQ-25). Questions could be reorganized into 2 constructs (a visual functioning scale and a socioemotional scale) that, after misfitting questions were removed, gave valid measurement of each construct and preserved 3 subscales. Removing redundancy from these long-form subscales yielded valid short-form scales. CONCLUSIONS: Several NEI VFQ subscales were not psychometrically sound; as an overall measure, it is flawed by multidimensionality. This was repaired by segregation into visual functioning and socioemotional scales. Valid long and short forms of the scales could enhance application of the questionnaire.