Psychometric validation techniques applied to the IND-VFQ-33 visual function questionnaire: the Hyderabad ocular morbidity in the elderly study (HOMES)

• Main Authors: William Mitchell, Srinivas Marmamula, Nazlee Zebardast, Weiwen Ng, Joseph J. Locascio, Thirupathi Kumbam, Satya Brahmanandam, Navya Rekha Barrenkala
• Format: Article
• Language: English
• Published: BMC 2021-02-01
• Series: BMC Medical Research Methodology
• Subjects: Ophthalmology; Psychometric validation; Factor analysis; Item response theory; Differential item functioning
• Online Access: https://doi.org/10.1186/s12874-021-01217-w

Abstract Background Over 2 billion people suffer from vision impairment or blindness globally, and access to validated visual measurement tools in imperative in accurately describing and managing the burden of eye disease. The present study applies contemporary psychometric validation techniques to the widely used 33-item Indian Visual Function Questionnaire (IND-VFQ-33). Methods We first estimated the polychoric correlation between each pair of items. Next, an unrotated and oblique Promax rotated factor analysis, item response theory (IRT, using a graded response model (GRM)), and differential item functioning (DIF) testing were applied to the IND-VFQ-33. We subsequently propose a validated IND-VFQ-33 questionnaire after psychometric testing, data reduction, and adjustment. Results Exploratory unrotated factor analysis identified two factors; one with a particularly high eigenvalue (18.1) and a second with a lower eigenvalue still above our threshold (1.1). A subsequent oblique Promax factor rotation was undertaken for a 2-factor solution, revealing two moderately correlated factors (+ 0.68) with clinically discrete item loadings onto either Factor 1 (21 items; collectively labelled “daily activities”) or Factor 2 (5 items; collectively labelled “bright lights”). IRT confirmed high item discrimination for all remaining items with good separation between difficulty thresholds. We found significant DIF on depression for six items in Factor 1 (all uniform DIF, except item 21 (non-uniform DIF) with no substantive difference in beta thresholds for any item and no substantive difference in expected individual or sum score, by depression at baseline. For Factor 2, only one item demonstrated significant uniform DIF on gender, similarly without major differences in beta thresholds or expected total score between gender at baseline. Consequently, no further item recalibration or reduction was undertaken after IRT and DIF analysis. Conclusion Applying IRT and DIF validation techniques to the IND-VFQ-33 identified 2 discrete factors with 26 uniquely-loading items, clinically representative of difficulty performing daily activities and experiencing difficulty due to bright lights/glare respectively. The proposed modified scale may be useful in evaluating symptomatic disease progression or response to treatment in an Indian population.