Purchase this article with an account.
B. Nafees, A. Rentz, D. A. Revicki, M. Meguro, J. Kowalski, J. Walt, J. Brazier, R. D. Hays; National Eye Institute Visual Function Questionnaire - 25-Item Reduction Using Rasch Analysis. Invest. Ophthalmol. Vis. Sci. 2008;49(13):4462.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
This study uses Rasch analysis (RA) to identify a subset of items in the NEI-VFQ-25 (VFQ-25) to derive a preference-based measure of health.Background: The VFQ-25 is a 25-item targeted profile measure that assesses vision-related functioning in patients across a range of eye diseases. Previous research demonstrates that generic measures such as SF-36 (Ware et al, 1993) have been converted into preference measures (Brazier et al, 2002) and have led to sensitive descriptive systems to classify people into health states. As an initial step in deriving a preference-based score from the VFQ-25, we used RA to identify a subset of items free of differential item functioning (DIF).
5 central vision loss (CVL) and 3 peripheral vision loss (PVL) datasets of VFQ-25 were used for these analyses. Datasets of each disease group were examined individually, and then combined to identify question items with the best psychometric properties, using specified criteria. A random subset of cases was randomly extracted to have approximately equal groups by disease type and to conform to limitations of RA. The individual disease group analysis was conducted on a sample of 970 CVL patients (mean age=69, male=49%, best eye ETDRS=55) and 1164 PVL patients (mean age=65, male 41%, best eye ETDRS=83). The combined disease analysis was conducted on a sample of 800 patients (mean age=67, male=45%). DIF analyses were also conducted to show significant differences between CVL and PVL, age and gender with a Bonferroni-adjusted p value of 0.002.
14 questions were identified with the best item fit in the CVL sample while 12 items were identified for PVL. In the combined dataset, driving items were excluded due to high missing (not applicable) rates (>30%). Analyses were conducted for the combined dataset to identify the items that showed significant DIF between CVL and PVL, age and gender. 9 items were selected that included items from near vision, distance vision, peripheral vision, social function, role difficulty, dependency and mental health. The selected items were reviewed for content validity by ophthalmologists.
This study illustrates the use of Rasch and DIF analyses to reduce the number of items in VFQ-25 for subsequent health state classification. This provides the basis for a health state classification that can then be valued to derive a preference-based measure in vision.
This PDF is available to Subscribers Only