Multidimensional Cognitive and Linguistic Profiles in Adolescent Schizo-Obsessive Presentations: A Machine Learning and Explainable Artificial Intelligence Approach.

Schizo-obsessive presentations in adolescence remain insufficiently characterized, particularly with respect to their cognitive, linguistic, and social-cognitive features. This study explored clinical, neurocognitive, social-cognitive, and thought-language characteristics associated with schizophrenia, obsessive-compulsive disorder (OCD), and schizo-obsessive presentations in adolescents using a Support Vector Machine-based explainable machine learning framework. A total of 182 adolescents were included: schizophrenia with comorbid OCD (n = 27), schizophrenia without obsessive-compulsive symptoms (n = 55), and OCD without psychotic symptoms (n = 100). The model demonstrated high discriminative performance overall, although classification performance was lower for the schizo-obsessive group, suggesting partial overlap with both schizophrenia and OCD presentations. Shapley Additive exPlanations (SHAP) analyses indicated that obsessive-compulsive symptom burden and negative symptom burden were the primary dimensions associated with classification, while social cognition, verbal fluency, verbal memory, executive functioning, processing speed, thought-language organization, and prosodic features also contributed to group differentiation. The findings suggest that schizo-obsessive presentations may involve broader cognitive and linguistic characteristics beyond symptom-level overlap and support a multidimensional approach to their clinical evaluation.