Application of machine learning in predicting aggressive behaviors from hospitalized patients with schizophrenia

Objective: To establish a predictive model of aggressive behaviors from hospitalized patients with schizophrenia through applying multiple machine learning algorithms, to provide a reference for accurately predicting and preventing the occurrence of aggressive behaviors. Methods: Cluster sampling was used to select patients with schizophrenia hospitalized from July 2019 to August 2021, divided into an aggressive behavior group (611 cases) and a non-aggressive behavior group (1,426 cases) according to whether they experienced obvious aggressive behaviors during hospitalization. Surveys included a General Condition Questionnaire, Insight and Treatment Attitude Questionnaire (ITAQ), Family APGAR, Social Support Rating Scale (SSRS), and Family Burden Scale (FBS). Multi-Layer Perceptron (MLP), Lasso, Support Vector Machine (SVM), and Random Forest (RF) algorithms were used to build predictive models; performance was evaluated with ROC AUC. A nomogram was constructed for clinical application. Results: The AUCs were MLP 0.904 (95% CI: 0.877–0.926), Lasso 0.901 (95% CI: 0.874–0.923), SVM 0.902 (95% CI: 0.876–0.924), and RF 0.955 (95% CI: 0.935–0.970). RF had significantly higher AUC than the other three models (p < 0.0001); the other three did not differ (p > 0.5). Conclusion: Machine learning models can fairly predict aggressive behaviors in hospitalized patients with schizophrenia, with Random Forest showing the best predictive performance and clinical application value.