Psychometric properties of the Vision Impairment in Low Luminance (VILL) questionnaire in age-related macular degeneration in the MACUSTAR study

Abstract

Purpose : Patient-reported difficulties caused by the characteristic functional impact of AMD, in particular in its early and intermediate stages, are only captured in selected patient-reported outcomes (PROs), none of which currently meet the FDA criteria for PRO development. The Vision Impairment in Low Luminance (VILL) questionnaire was designed to meet these criteria as well as to capture the characteristic impact of all AMD stages on patients’ lives. Here we report its psychometric functioning and test-retest reliability in the European multi-center study MACUSTAR.

Methods : Using the baseline VILL data from participants of the MACUSTAR study (intermediate, early, late or no AMD) psychometric characteristics including model fit, person reliability (PR), person separation index (PSI), targeting and dimensionality of the existing VILL subscales (reading, VILL-R; mobility, VILL-M; emotional, VILL-E) were investigated using Rasch analysis (Winsteps, Chicago, IL). In a subset with repeated data test-retest reliability was assessed using mean deviation (MD), intra-class correlation coefficients (ICC) and Bland-Altman analysis.

Results : A baseline dataset including 716 participants was available for analysis. The VILL subscales were unidimensional. Dropping five misfitting items in an iterative process (VILL-R, 3 items; VILL-M, 1 item; VILL-E, 1 item) improved model fit. After removal of a total of 39 single misfitting responses from three additional misfitting items, 32 items could be retained and no item showed misfit. PR and PSI were 0.91 and 3.27 (VILL-R), 0.87 and 2.58 (VILL-M) and 0.73 and 1.65 (VILL-E), respectively. The mean person measure-item measure differences were 2.0 (VILL-R), 2.1 (VILL-M) and 1.6 (VILL-E). 289 participants were included in the evaluation of test-retest reliability of the VILL-32. MD and ICC (95% confidence interval) were 0.30 and 0.92 (0.90-0.94, VILL-R), 0.07 and 0.93 (0.91-0.94, VILL-M) and 0.30 and 0.85 (0.81-0.88, VILL-E).

Conclusions : Our results support internal consistency and test-retest reliability of the VILL in AMD in a large sample. The reading and mobility subscales are reliable measures while the emotional subscale is noticeably less precise. Thus, the emotional subscale was dropped, resulting in 29 items overall. The validation process of the instrument using functional data is ongoing.

This is a 2021 ARVO Annual Meeting abstract.