Construction and Validation of a Risk Prediction Model for Violent Aggressive Behavior in Adolescents with Internet Addiction During Withdrawal

doi:10.12104/j.issn.1674-4748.2025.12.011

Abstract：Objective: To analyze the risk factors for violent aggressive behavior in adolescents with internet addiction during withdrawal, and to construct and validate a risk prediction model. Methods: A retrospective study was conducted on 200 adolescents with internet addiction who received treatment at the hospital from March 2021 to March 2023. The subjects were divided into a training set (n=140) and a validation set (n=60) in a 7:3 ratio. They were further classified into an attack group and a non-attack group based on whether they exhibited violent aggressive behavior during withdrawal. Univariate analysis and multivariate logistic regression analysis were used to identify factors associated with violent aggressive behavior during withdrawal. A risk prediction model was constructed, and its predictive performance was validated using the area under the receiver operating characteristic (ROC) curve (AUC) and the Hosmer-Lemeshow (H-L) test. Results: Among the 140 patients in the training set, 26 cases of violent aggressive behavior occurred, with an incidence rate of 18.57%. Multivariate logistic regression analysis showed that being male, having an impulsive personality, experiencing rejection or overprotective parenting styles, poor academic performance, and low self-efficacy were risk factors for violent aggressive behavior in adolescents with internet addiction during withdrawal. The H-L test results for the training set showed χ² = 7.543, P = 0.479. The model's AUC was 0.841 [95% CI (0.761, 0.922), P < 0.05], with a sensitivity of 0.731 and a specificity of 0.868. For the validation set, the H-L test showed χ² = 6.752, P = 0.564, and the model's AUC was 0.810 [95% CI (0.719, 0.901), P < 0.05], with a sensitivity of 0.808 and a specificity of 0.711. Conclusion: The risk prediction model built based on five risk factors—gender, impulsive personality, academic performance, parental education style, and self-efficacy—has good predictive performance and can provide reference for effective clinical intervention.