A healthcare system overwhelmed by demands hampers the timely diagnosis of specific learning disabilities (SLDs) in children. The need for improved school-based screening is highlighted, as teachers often lack clinical training. This study analyzed data from 364 children referred for clinical consultation, utilizing a 96-item screening questionnaire completed by teachers. Researchers computed a severity score across 19 learning sub-domains using item response theory and employed cluster analysis to categorize children based on their capabilities. Two distinct profiles emerged from the analysis, differentiated by severity, though the proportion of children entering the clinical pathway did not significantly differ. The study emphasizes that even children with fewer difficulties may have SLDs that warrant attention. Machine learning models, specifically a Support Vector Classifier (SVC) and Naive Bayes (NB), were used to assess outcomes, yielding a median area under the precision-recall curve of 0.96 for one cluster and 0.69 for the other. Findings indicate varying complexities in children’s profiles and underscore the importance of early intervention and comprehensive assessment in managing SLDs.