A Comprehensive Framework for the Development and Evaluation of Linguistically Fair Universal Reading Screening in Educational Settings

Early and accurate identification of students who struggle with learning to read is essential for providing timely and effective interventions. This requires universal screening tools that not only produce reliable and valid results across linguistically diverse student populations but also support scalability and allow teachers and clinicians to interpret the results meaningfully. In this study, we present a comprehensive framework for the development and evaluation of linguistically fair universal reading screening tools for use in educational settings. As part of this framework, we consider (i) classification accuracy, (ii) instructional relevance, (iii) clinical utility, and (iv) operational feasibility. We apply this framework and propose a series of reading screening algorithms that accurately predict end-of-year reading risk in English, with sensitivity exceeding 80% in kindergarten and grade 1, and 90% in grade 2. The Spanish algorithm achieves similar sensitivity levels, though kindergarten screening remains more challenging for bilingual students. Notably, the inclusion of oral language and expressive vocabulary measures enhances specificity when administered in students' first language, while these same measures may introduce systematic bias (in the form of reduced specificity) when administered in English to students classified as English learners. While very promising, these findings underscore the critical need for developing and validating reading screeners in languages beyond English, as well as creating comprehensive multilingual screening protocols that leverage students' full linguistic repertoires across all their languages.