Nomogram predicting bullying victimization in adolescents

Abstract

Objective: The purpose of this study was to construct a cross-sectional study to predict the risk of bullying victimization among adolescents.

Methods: The study recruited 17,365 Chinese adolescents using stratified random cluster sampling method. The classical regression methods (logistic regression and Lasso regression) and machine learning model were combined to identify the most significant predictors of bullying victimization. Nomogram was built based on multivariable logistic regression model. The discrimination, calibration and generalization of nomogram were evaluated by the receiver operating characteristic curves (ROC), the calibration curve and a high-quality external validation.

Results: Grade, gender, peer violence, family violence, body mass index, family structure, depressive symptoms and Internet addiction, recognized as the best combination, were included in the multivariable regression. The nomogram established based on the non-overfitting multivariable model was verified by internal validation (Area Under Curve: 0.749) and external validation (Area Under Curve: 0.755), showing decent prediction of discrimination, calibration and generalization.

Conclusion: Comprehensive nomogram constructed in this study was a useful and convenient tool to evaluate the risk of bullying victimization of adolescents. It is helpful for health-care professionals to assess the risk of bullying victimization among adolescents, and to identify high-risk groups and take more effective preventive measures.

Keywords: Adolescent; Bullying victimization; Nomogram; Prediction; Validation.