The secrets of medical students' psychological resilience: a dual perspective of machine learning and path analysis.

Background: As future healthcare workers, if medical students can enhance their psychological resilience, it will help them better cope with the pressures and challenges of their future work, thereby improving the overall mental health of healthcare personnel. We aim to identify high-risk individuals among medical students with low psychological resilience and to explore the potential mechanisms for improving psychological resilience.

Methods: We developed an Extreme Gradient Boosting (XGBoost) model using basic characteristics and health behavior information of medical students, and compared its performance with that of Logistic Regression (LR), Decision Tree (DT), and Random Forest (RF) models. Additionally, we employed SHapley Additive exPlanations (SHAP) for interpretability analysis of the model. After identifying the key predictive variables, we conducted mediation analysis to explore intervention pathways for enhancing psychological resilience among medical students.

Results: Among the four machine learning models, the XGBoost model demonstrated the best predictive performance, with an accuracy of 0.822 (95%CI: 0.775-0.866), an AUC value of 0.856 (95%CI: 0.799-0.905), sensitivity of 0.609 (95%CI: 0.492-0.722), specificity of 0.902 (95%CI: 0.857-0.943), and an F1 score of 0.700 (95%CI: 0.547-0.739). Smartphone addiction and sleep disturbances are negatively correlated with psychological resilience in medical students, while perceived social support is positively correlated with it. Additionally, smartphone addiction and sleep disturbances can serve as either independent mediators or chain mediators between perceived social support and psychological resilience in medical students.

Limitations: We have only analyzed the dataset from a single institution and have not yet validated our model on an external dataset, which limits the generalizability of the findings. Additionally, some important predictors of Psychological Resilience may have been overlooked, potentially affecting the model's performance.

Conclusions: Perceived social support, smartphone addiction, and sleep disturbance are important factors in predicting medical students' psychological resilience. Additionally, perceived social support can directly influence psychological resilience, as well as indirectly affect it through smartphone addiction and sleep disturbance.