Stud Health Technol Inform. 2025 Jun 26;328:126-130. doi: 10.3233/SHTI250686.
ABSTRACT
This study analyzes a sanitized cohort of 152 delinquency-referred adolescents (11-21 yrs; mean 16.3 ± 2.4) using two complementary, data-driven steps. First, we rank and plot the 20 most frequent ICD-10 codes, exposing a dominance of hyperkinetic (F90.x) and family-related psychosocial (Z63.5) diagnoses. Second, age and one-hot ICD flags feed a K-Means model whose four-cluster solution-visualised via t-SNE-delineates clinically coherent subgroups. The lightweight Python workflow shows how routinely collected records can swiftly surface high-risk profiles and guide targeted adolescent mental-health interventions.
PMID:40588894 | DOI:10.3233/SHTI250686
