Diagnostic sensitivity and specificity of the Teach CVI screening tool for identifying cerebral visual impairment in children
Elsie Bush, Terry Schwartz, Melissa Rice, Li Lin, Reena Donofe, and Karen Harpster
Main Research Question
This study set out to answer a practical and clinically important question: how well does the Teach CVI screening tool identify children with Cerebral Visual Impairment (CVI)? More specifically, the researchers examined whether the tool’s three versions could accurately flag children who either do, or do not, have CVI, and whether screening outcomes align with formal diagnoses.
Background Context
Cerebral Visual Impairment (CVI) is a neurological condition that affects how the brain processes visual information. Unlike ocular vision problems, CVI originates in the brain’s visual pathways. It is now recognized as the leading cause of visual impairment in children in developed countries, yet it remains widely underdiagnosed.
One of the central challenges lies in its variability. Some children exhibit profound visual difficulties, while others struggle only in visually complex environments. These differences, combined with co-existing conditions like autism or ADHD, make CVI difficult to detect early. As a result, clinicians often need better tools to determine when a child should be referred for further evaluation.
Screening tools like the Teach CVI questionnaire are designed to bridge this gap, offering a structured way to capture caregiver observations and identify concerning visual behaviors.
Methodology
To evaluate the effectiveness of the Teach CVI tool, researchers conducted a retrospective chart review at a pediatric academic medical center in the United States. They analyzed records from children who had completed the screening between 2018 and 2023.
The final sample included 257 children aged 2 to 21, with an average age of just over 6 years. Notably, nearly three-quarters of the participants had already received a CVI diagnosis. Many also presented with additional neurological or developmental conditions, reflecting the complexity of this population.
The Teach CVI tool itself consists of three versions: one designed for non-ambulatory children (Teach 1), one for younger, ambulatory children ages 2–6 (Teach 2), and one for older, ambulatory children ages 6–12 (Teach 3). Caregivers complete the questionnaire, rating observed visual behaviors across a range of situations.
Researchers evaluated the tool using several statistical measures, including sensitivity (how well it detects true cases), specificity (how well it rules out non-cases), and overall diagnostic performance (Area Under the Curve - AUC). They also examined whether screening results were statistically associated with confirmed CVI diagnoses.
Image Description: Optometrist examining the eyesight of a young boy with cerebral palsy, using a small flashlight. Photo credit: Jodi Mora Igaul.
Key Findings/Results
The results revealed mixed results: each version of the tool performed differently, with its own strengths and limitations.
Teach 1, designed for non-ambulatory children, demonstrated strong sensitivity. In other words, it was relatively effective at identifying children who do have CVI. However, this came at a cost: its specificity was quite low, leading it to frequently flag children who ultimately did not have the condition. This tendency toward false positives may be influenced by the complex motor and developmental challenges present in this population.
Teach 2 showed a different pattern. Its sensitivity was lower, indicating that some children with CVI might be missed. At the same time, it performed better in terms of specificity, suggesting it was more reliable at identifying children without CVI. Importantly, results from this version were significantly associated with actual diagnoses, lending it some clinical credibility.
Teach 3 offered the most balanced performance. With moderate levels of both sensitivity and specificity, it struck a middle ground, reasonably effective at identifying true cases while also avoiding excessive false positives. Like Teach 2, its results showed a meaningful statistical relationship with CVI diagnosis.
Across all versions, overall diagnostic accuracy (as measured by AUC) ranged from poor to only marginally acceptable. This finding underscores a key point: the tool is better suited for screening than for definitive diagnosis.
One particularly practical insight emerged from the analysis: a smaller subset of key screening questions performed just as well as the full questionnaire. This raises the possibility of using a shorter, more efficient version of the tool in busy clinical settings without sacrificing effectiveness.
Significance of the Study: What This Means for CVI
This study fills an important gap in the literature. While several CVI screening tools exist, few have been rigorously evaluated in clinical populations. By examining real-world data, the authors provide early evidence about how the Teach CVI tool performs in practice.
The findings are especially relevant given the current underdiagnosis of CVI. Early identification is critical because it enables timely interventions that can significantly improve a child’s development and quality of life. Screening tools, even imperfect ones, play a vital role in this process by helping clinicians decide when further evaluation is warranted.
At the same time, the study highlights the inherent trade-offs in screening. A tool that prioritizes sensitivity may over-identify cases, while one that emphasizes specificity risks missing children who need support. Understanding this balance is essential for clinicians using the tool.
Main Conclusions
The authors conclude that the Teach CVI screening tool shows promise, but with clear limitations. It should not be used as only a diagnostic instrument. Instead, its value lies in complementing clinical judgment, medical history, and multidisciplinary assessment.
Among the three versions, Teach 1 and Teach 3 demonstrated the strongest ability to detect potential cases, while Teach 3 offered the most balanced overall performance. Teach 2, although less sensitive, showed meaningful associations with diagnosis and may still be useful in certain contexts.
The study also suggests that a shorter set of screening questions could serve as a practical alternative to the full tool, particularly in settings with limited time.
Ultimately, the Teach CVI tool is best understood as a first step, a way to raise awareness and prompt further investigation rather than to provide definitive answers.
Limitations and Future Directions
As an exploratory, retrospective study, the findings should be interpreted with caution. The sample was drawn from a population already suspected of having CVI, which may limit generalizability. Additionally, reliance on caregiver reports introduces subjectivity, and the sample size, while meaningful, was not large enough to fully validate the tool.
Future research will need to take a more prospective approach, ideally involving larger and more diverse populations. Studies that include children at varying levels of risk and that refine how questions are structured and scored will be especially important.
Overall Summary
In summary, this study offers an important first look at the Teach CVI screening tool’s performance in the real world. While it does not provide definitive diagnostic accuracy, it demonstrates clear potential as a supportive screening instrument. When used thoughtfully and in combination with broader clinical evaluation, the tool can help clinicians identify children who may otherwise go unnoticed, an essential step toward improving outcomes for those with cerebral visual impairment.
Elsie Bush, Terry Schwartz, Melissa Rice, Li Lin, Reena Donofe, Karen Harpster. Diagnostic sensitivity and specificity of the teach CVI screening tool for identifying cerebral visual impairment in children. Research in Developmental Disabilities, Volume 166, 2025, 105125, ISSN 0891-4222, https://doi.org/10.1016/j.ridd.2025.105125