Machine Learning Insights for Adolescent Mental Health: A Data-Driven Approach for Functional Communication, Therapy,...

Area of Science:

• Adolescent Mental Health
• Clinical Informatics
• Data Science in Healthcare

Background:

• Adolescents referred for delinquency often present with complex mental health needs.
• Identifying high-risk profiles within this population is crucial for effective intervention.
• Routinely collected clinical data holds potential for identifying such profiles.

Purpose of the Study:

• To analyze diagnostic patterns in delinquency-referred adolescents.
• To identify clinically coherent subgroups using data-driven methods.
• To demonstrate a workflow for surfacing high-risk profiles from electronic health records.

Main Methods:

• Analysis of a cohort of 152 delinquency-referred adolescents (ages 11-21).
• Ranking and plotting of the 20 most frequent International Classification of Diseases, 10th Revision (ICD-10) codes.
• Application of a K-Means clustering model, visualized with t-distributed Stochastic Neighbor Embedding (t-SNE), using age and ICD-10 flags.

Main Results:

• Hyperkinetic (F90.x) and family-related psychosocial (Z63.5) diagnoses were the most frequent.
• The K-Means model delineated four clinically coherent subgroups.
• The workflow efficiently identified high-risk profiles.

Conclusions:

• Routinely collected data can be rapidly analyzed to identify adolescent mental health needs.
• Data-driven subgrouping aids in understanding diverse clinical presentations in delinquency-referred youth.
• This approach can inform targeted adolescent mental health interventions.