Automated Risk Assessment for School Violence: a Pilot Study

Area of Science:

• Forensic Psychology
• Child and Adolescent Psychiatry
• Machine Learning in Healthcare

Background:

• School violence is a growing concern, necessitating improved risk assessment methods.
• Current methods for identifying at-risk students may lack standardization and sensitivity.
• Early identification is crucial for implementing timely interventions to prevent school violence.

Purpose of the Study:

• To evaluate the clinical utility of the BRACHA (School Version) and School Safety Scale (SSS) for assessing school violence risk.
• To develop and test a machine learning algorithm for predicting student risk of school violence.
• To identify key factors contributing to school violence risk in adolescents.

Main Methods:

• 103 students (ages 12-18) from a children's hospital were assessed using the BRACHA and SSS, alongside open-ended questions.
• Guardian-provided collateral information was collected prior to student evaluations.
• A machine learning model was trained using transcribed interview content, demographic, and socioeconomic data.

Main Results:

• 55 out of 103 students were identified as moderate to high risk for school violence.
• Both BRACHA and SSS demonstrated high correlation with violence risk (Pearson correlation > 0.82).
• The machine learning algorithm achieved 91.45% accuracy in predicting school violence risk when including demographic and socioeconomic data.

Conclusions:

• The BRACHA and SSS are effective and clinically useful tools for assessing school violence risk.
• Machine learning models can accurately predict school violence risk, offering a valuable tool for prevention.
• Previous violent behavior, impulsivity, school problems, and negative attitudes are significant indicators of violence risk.