Impact of general health status on validity of visual impairment measurement

Purpose: To examine the effects of bias and distortion from general health status on Rasch model-based measurements of visual function, and establish measurement equivalence across different health status groups in low vision patients. Methods: Self-reported data were obtained from 1746 low vision patients who completed the Activity Inventory (AI) and an intake health-related questionnaire prior to their first visit to the low vision rehabilitation service. Differential Item Functioning (DIF) analysis by health status and separate Rasch analyses adjusted by health status were performed on the responses to both a DIF-free scale and the full scale of the AI. Results: Of 48 Goal-level items, only 15 items showed significant DIF (P < 0.001). Comparing the vision-related estimates from the original full set of items with those from the DIF free scale; we found that only 25% of the person measure estimates differed by 0.5 logits or more and there was a strong intraclass correlation between the two scales in measuring the patients' vision-related functional ability (IC = 0.75). Patients' health status influenced the Rasch model based estimation of visual function by the AI (ANOVA, p = 0.005), but this effect was within accepted range of MISFIT statistics. Conclusion: Taken together, our results indicate that despite confounding effects of general health status on vision-related ability, we can still regard visual ability to be a single theoretically constructed variable for the low vision population. It appears that self-perceived comorbidities add to vision-related disability, but do not distort its measurement.