Background: To evaluate the reliability and construct validity of a Greek version of the NEI-VFQ-25 in patients with chronic ophthalmic diseases.
Methods: We developed the Greek version of the instrument using forward and backward translation. One hundred-eighty-six patients responded to the questionnaire. To examine reliability, Cronbach's alpha for each subscale was used as an index of internal consistency. Test-retest reliability was evaluated with intraclass correlation coefficients. Regarding construct validity, both convergent and discriminant validities were calculated by means of multi-trait analysis. Rasch analysis was used to estimate the visual ability required by each item for a particular response, and each patient's visual ability. Correspondingly, instrument validity was evaluated by estimating the distribution of residuals for item and subject measures.
Results: Four patient groups were studied, each including participants with a single cause of visual impairment. Group 1 consisted of 84 glaucoma subjects. Group 2 included 30 subjects with age-related macular degeneration (ARMD); group 3 included 25 subjects with dry-eye syndrome, whereas group 4 included 18 cataract patients. Twenty-nine healthy individuals comprised the control group. NEI-VFQ scores (mean +/- SD) for the glaucoma, ARMD, dry-eye, cataract and control groups were: 76.9 +/- 20.2, 70.9 +/- 20.2, 81.6 +/- 16.5, 73.5 +/- 24.0 and 93.7 +/- 8.9 respectively. Item analysis revealed no significant data skewing. Cronbach's alpha ranged from 0.678 to 0.926, with most subscales having high internal consistency. Intraclass correlation coefficient ranged from 0.717 to 0.910 for all subscales. All items passed the convergent and discriminant validity tests. Strong correlations were detected between visual acuity and "general vision", "distant activities" and "near activities" subscales. Significant correlations were also detected between visual field deficits and the "peripheral vision" and "general vision" subscales. Rasch analysis revealed potential weaknesses of the instrument that are associated with the assumptions of the model itself. Specifically, low precision of the "agreement" items was detected in the estimation of visual ability. Twenty-three percent of the subjects had fit statistics that fell outside the tolerance box.
Conclusion: Although traditional validation methods indicated that the Greek version of the NEI-VFQ-25 is a valid and reliable instrument for VS-QoL assessment, Rasch analysis detected significant misfits to the model, especially of the "agreement" items. This means that results of the corresponding subscales should be interpreted with extreme caution.