Logistic Regression and Decision Tree Modeling in Factors Influencing Adolescents' Depression, Anxiety, and Stressful Emotions

Purpose: Logistic regression and decision tree modeling were applied to analyze the influencing factors of adolescents' anxiety, depression, and stressful emotions, with the aim of promoting adolescents' mental health and providing information for early preventive interventions of adverse mental health problems. Methods: 3,922 students from 10 middle schools in 5 regions of Xinjiang Uygur Autonomous Region were selected by convenience sampling method, and Logistic regression model and decision tree model were constructed to compare the differences between the results of the two analytical methods and evaluate the prediction effects. Results: The results of both models showed that health status, self-efficacy, myopia, sleep duration, psychological resilience, parent-child relationship, and school adjustment were the influencing factors of adolescents' anxiety, depression, and stressful emotions. The ROC curves of the two models were plotted separately, and the ROC curves of the two models were close to each other, indicating that the classification effects of the two models were similar.The AUC of the logistic regression model was 0.756 (95% CI: 0.741~0.771), with a sensitivity of 74.3% and a specificity of 62.8%. The categorical decision tree model had an AUC of 0.771 (95% CI: 0.757~0.786), a sensitivity of 60.6% and a specificity of 79.1%. Both models have P<0.01, indicating that the classification effects of the two models are of practical significance; and the AUC values of both are >0.7, indicating that the classification prediction results of the two models have a certain degree of accuracy. Comparison of the two models showed that the decision tree model was better than the logistic regression model, and the difference was statistically significant (P<0.01). Conclusion: We should focus on the influence of health status, self-efficacy, myopia, sleep duration, psychological resilience, parent-child relationship and school adaptation on the presence of depression, anxiety and stressful emotions in adolescents, and the combination of Logistic regression model and CRT model can effectively screen the influencing factors of the presence of depression, anxiety and stressful emotions in adolescents, which can help to formulate targeted measures to improve the mental health of adolescents.