Abstract Purpose To develop a new comprehensive vision-related low luminance performance-based measure, the ‘Assessment of Low Luminance Vision-Related Activities’ (ALLVA), and evaluate its construct validity and test-retest reliability using Rasch analysis. Methods A cross-sectional observational study of 75 adults with vision impairment from various ocular conditions (mean age 70 years SD: 15 years; mean binocular visual acuity 0.63 logMAR SD: 0.45 logMAR) was conducted. Seventeen tasks were developed as items and administered to participants under low luminance, with completion time and number of errors recorded. As some items could not be completed by all participants, five categories of completion time were created for analysis (quartiles and a fifth category representing non-completion). The ‘method of successive dichotomisations’—a polytomous Rasch model that always estimates ordered response category thresholds, enabling its application to binned continuous data—was applied to create a single combined measure of performance. Errors were not analysed as they occurred infrequently and generally increased completion time. Eleven participants with age-related macular degeneration were retested after 2–4 weeks. Clinical vision measures, including visual acuity, contrast sensitivity and visual fields, were also collected. Results Initial analysis of the 17-task ALLVA led to removal of one item, walking a mobility course, due to infit and outfit mean square statistics being outside the acceptable range. For the remaining 16 tasks, item difficulty was well targeted to person ability, with only a minor floor effect. Item and person reliability values were 0.98 and 0.93, respectively. Clinical vision measures were significantly correlated with person measures. Bland–Altman analysis indicated a mean difference between test and retest person measures of −0.08 logits (95% limits of agreement 2.16 to −2.32 logits). Conclusion The ALLVA is the first comprehensive vision-related low luminance performance-based measure. It demonstrated strong Rasch psychometric properties, validity and good test-retest reliability.
Sailaja et al. (Wed,) studied this question.