Jem Martin ^1* Chris Bradley ² Barry Kran ¹ Nicole Ross ^1,3

• ¹ New England College of Optometry, Boston, United States
• ² Wilmer Eye Institute, School of Medicine, Johns Hopkins Medicine, Baltimore, Maryland, United States
• ³ Northeastern University, Boston, Massachusetts, United States

Purpose: Cerebral Visual Impairment (CVI) is the leading cause of pediatric visual impairment. Given the diversity of clinical presentations of CVI, we are interested in whether questionnaires appropriately target the spectrum CVI cases, specifically the Teach-CVI Screening Tool. Rasch analysis is a standard psychometric technique for assessing the targeting of questionnaire items, however this analysis technique has not yet been applied to this questionnaire. Methods: We performed a retrospective review of clinical CVI cases from the NECO Center for Eye Care at Perkins School for the Blind from January 2016 to December 2022. Electronic medical records were reviewed to identify patients with an ICD-9 or ICD-10 code of CVI or other neurological visual impairment. Age, gender, diagnoses, visual acuity, contrast sensitivity, visual fields, ocular alignment, and Teach-CVI responses were collected. We applied the method of successive dichotomizations, a polytomous Rasch model, to estimate item measures and person measures from the survey. Targeting of questionnaire items to the sample population was explored by comparing estimated item measures to person measures. Multiple linear regression was used to determine which factors influence patient visual ability (i.e., Teach-CVI person measure). Results: 119 records were included, 54% of which were male. The mean age was 8.9 years (SD = 6.12) with a range of 0 to 33 years of age. Mean visual acuity was 0.46 logMAR (SD = 0.40), or 20/57. The majority of patients in the sample had a co-occurring visual disorder in addition to CVI (84%), the most frequent being strabismus (69.9%) or visual field loss (25.3%). Item measures ranged from -2.67 to 1.77 logits (SD = 0.76), with a mean of 0 logit by convention. Estimated person measures ranged from -2.19 to 3.08 logits (SD = 1.10) with a mean of -0.03 logit. The range of item measures covered 93.3% of the person measures, and all person measures, except one, were within one logit of an item measure. Visual measures were not statistically significantly associated with Teach-CVI person measures. Conclusion: The findings from this study suggest that the Teach-CVI survey is well targeted and an appropriate patient reported outcome measure for CVI.