Prediction of physical violence in schizophrenia with machine learning algorithms

Area of Science:

• Psychiatry
• Computational Neuroscience
• Clinical Psychology

Background:

• Patients with schizophrenia face a higher risk of physical violence.
• Identifying violent patients is challenging due to the complexity of integrating risk factors.

Purpose of the Study:

• To develop a clinically relevant predictive tool for physical violence in schizophrenia patients.
• To evaluate the efficacy of machine learning algorithms in predicting violent behavior.

Main Methods:

• A cross-sectional study of 275 schizophrenia patients (103 violent, 172 non-violent).
• Identification of demographic, clinical, and sociocultural variables as predictors.
• Application of seven machine learning classification algorithms to predict past physical violence.

Main Results:

• Machine learning algorithms demonstrated significant predictive accuracy compared to random chance.
• The random forest model achieved the highest performance with 62% accuracy and an AUROC of 0.63.
• Algorithms showed varying degrees of predictive power, highlighting potential for clinical application.

Conclusions:

• Machine learning classification algorithms show promise in predicting physical violence in schizophrenia.
• Further optimization of these models is necessary to enhance diagnostic support for early intervention.
• Improved prediction can aid in implementing timely measures to ensure patient and public safety.