Machine Learning in Adolescent Mental Health: Advanced Comorbidity Analysis and Text Mining Insights.

Area of Science:

• Adolescent mental health
• Machine learning in clinical psychology
• Justice-involved youth psychology

Background:

• Justice-involved adolescents have high rates of complex mental health disorders.
• Understanding comorbidity patterns is vital for effective interventions in this population.
• Existing research often lacks detailed comorbidity profiles for this demographic.

Purpose of the Study:

• To apply machine learning to identify comorbidity patterns in justice-involved adolescents.
• To reveal actionable profiles for guiding targeted mental health interventions.
• To explore the utility of routine clinical data for understanding complex adolescent presentations.

Main Methods:

• Utilized machine learning techniques including association rule mining, K-Means clustering, and topic modeling.
• Analyzed electronic health records from 124 justice-involved adolescents (ages 11-21).
• Employed t-SNE visualization for cluster analysis and topic modeling for clinical notes.

Main Results:

• Hyperkinetic disorders and family stress were prominent, comprising 45% of diagnoses.
• Clustering improved significantly with the inclusion of Z-codes, revealing a hyperkinetic-family-stress subgroup.
• Association rules highlighted the co-occurrence of learning and attention-deficit disorders (F81 → F90.0).
• Topic modeling identified five key intervention themes aligning with data-driven clusters.

Conclusions:

• Routine clinical data can effectively reveal actionable comorbidity profiles in justice-involved adolescents.
• Machine learning approaches enhance the identification of distinct patient subgroups.
• Findings support the development of tailored interventions for complex adolescent mental health needs.