Early Prediction of Educational Support Needs Through Machine Learning: Protocol and State of the Art in Early Childhood and Primary Education
Background: Timely identification of Educational Support Needs (NEAE / Special Educational Needs) during early developmental stages (ages 3 to 12) is decisive for preventing learning gaps and optimizing inclusive school pathways. However, traditional support models operate predominantly reactively. Objective: To systematically synthesize scientific evidence regarding the efficacy, accuracy, predictive variables, ethical challenges, and level of explainability of Machine Learning (ML) models designed for the early detection and prediction of educational support needs in early childhood and primary education. Eligibility Criteria: Peer-reviewed empirical quantitative or mixed-methods studies implementing supervised, unsupervised, or multimodal ML algorithms to predict or identify early educational support needs in students aged 3 to 12, published in English or Spanish between 2015 and 2026. Information Sources: Comprehensive literature search across a strategic interdisciplinary core of 4 databases: Scopus, Web of Science (WoS), IEEE Xplore, and ERIC (with optional complementary searches in PubMed/PsycINFO). Risk of Bias Assessment and Synthesis: The adapted PROBAST (Prediction model risk of bias assessment tool) will be used to evaluate methodological risk of bias and applicability. A qualitative and quantitative narrative synthesis grouped by predictive variable taxonomy, algorithmic architecture, and educational level will be conducted.