Researchers at the HSE University Centre for Language and Brain, in cooperation with a professor of neuropsychology from Lomonosov Moscow State University and specialists from the Centre for Speech Pathology and Neuro-rehabilitation, evaluated the diagnostic validity of the Standardized Assessment of Reading Skills in Russian (SARS) and checked whether the available normative data are current. The results of the study, the updated levels for reading speed, as well as the control levels for evaluating these indicators, were published in The Russian Journal of Cognitive Science.
This is the first-ever experimental evaluation of the diagnostic validity of the SARS. Previously, there was no scientific verification of how accurately children were diagnosed with dyslexia and children without reading disorders were classified as normal.
A. Kornev developed the Standardized Assessment of Reading Skills in Russian (SARS) in 1982. It was first published in 1997. Speech therapists and other specialists use it to diagnose dyslexia. However, the normative data for this test were collected more than 10 years ago and, therefore, requires updating. Such updates are a general requirement for most standardized tests.
The authors of the study tested 90 children of normal development who had mastered basic reading skills, as well as 50 children whom experienced specialists had clinically diagnosed as dyslexic. They then analysed the correlation between the clinical diagnoses and the children’s indicators according to the SARS using both the original norms given in the published guide to the test as well as the new normative data.
The study showed that the test has high specificity. This means that the SARS makes it possible to identify typical reading indicators and the absence of dyslexia with high accuracy: from 100% (when using standard data obtained in 2007-2008) to 94.4% (when using control levels calculated with data from 2018). However, the sensitivity of the SARS — that is, its ability to identify a child with dyslexia as having a reading disorder — largely depends on the normative data used to calculate ‘cut-off levels.’
When using normative data obtained in 2007-2008, the sensitivity of the test ranged from 36.6% to 39.0% (depending on the text), and when using new data on reading speed obtained in 2018, the sensitivity of the test reached from 60.0% to 72.0%. However, even 72.0% is a very low degree of sensitivity. It means that up to 28.0% of children will be misdiagnosed as having no reading disorders when, in fact, they might be dyslexic.
The relatively low sensitivity of the SARS is the result of using only one (albeit the most important) diagnostic criterion — decoding speed. However, the SARS actually allows for the use of two additional diagnostic criteria: the level of reading comprehension and a weighted assessment of reading errors.