Abstract
Purpose: :
To compare the psychometric performances of the IVI and NEI VFQ-25.
Methods: :
The 28-item IVI and 25-item NEI VFQ-25 were administered to participants with low vision. Both questionnaires were assessed using Rasch analysis (Winsteps v3.67). The key psychometric parameters used to compare were: person separation reliability (PSR); person-item targeting; item fit; unidimensionality; and differential item functioning (DIF). The psychometric performance of subscales [IVI (3); NEI VFQ-25 (12)] was also assessed.
Results: :
Two hundred and thirty four participants were recruited (mean±SD age 67.4±16 yr; 48% male). 63% had visual acuity <6/18. Initial fit to the Rasch model for the NEI VFQ-25 showed 6 misfitting items (Infit mean square (MNSQ)>1.3). Ten items had disordered thresholds. To improve fit, items overall health, pain or discomfort and doing work or hobbies were removed and the 10 items with disordered thresholds were dichotomized. Thereafter, targeting was excellent (mean person measure=0.1) and PSR (0.92). There was no evidence of DIF and multidimensionality (unexplained variance in the 1st contrast =2.9 Eigenvalue units (EU)). Seven NEI VFQ-25 subscales could not be fitted to the Rasch model due to insufficient items (<3 items) and the remaining 5 displayed poor item fit characteristics. Initial fit of the IVI showed ordered thresholds for all items. Only one item showed marginal misfit (worry about eyesight MNSQ=1.33). PSR was excellent (0.94). DIF and multidimensionality (3.1 EU) were not evident. Targeting was moderate (mean person measure=0.8). All 3 IVI subscales had ordered thresholds, no DIF and were unidimensional (EU=1.6-2.1). PSR ranged from 0.81 to 0.87 and targeting from 0.72 to 0.90 Logits.
Conclusions: :
The NEI VFQ-25 offers good overall measurement of functioning but none of the subscales are viable. In contrast, the IVI provides measurement of overall and specific aspects of vision-related quality of life in people with low vision.
Keywords: low vision • quality of life