This study aimed to develop a predictive model for aggressive behaviors in hospitalized schizophrenia patients using machine learning algorithms. Researchers used cluster sampling to select 2037 patients (611 aggressive, 1426 non-aggressive), collecting data via questionnaires (General Condition Questionnaire, Insight and Treatment Attitude Questionnaire, Family APGAR, Social Support Rating Scale, Family Burden Scale) and the Modified Overt Aggression Scale (MOAS). Multi-layer Perceptron, Lasso, Support Vector Machine, and Random Forest algorithms were employed. Random Forest demonstrated the best predictive effect (AUC = 0.955), significantly outperforming the other models. A nomogram was created for clinical application, highlighting key factors such as Family APGAR, Insight and Treatment Attitude, disease duration, history of attacks, social support, medication adherence, age, and family burden.

Nuo Cheng, Meihao Guo, Fang Yan, Zhengjun Guo, Jun Meng, Kui Ning, Yanping Zhang, Zitian Duan, Yong Han, Changhong Wang, Arjen Noordhof, Ioannis K Gallos, Maryam Ravan