Development and validation of a machine learning model with a web-based calculator for predicting anxiety risk among nursing interns

The clinical internship, a period of profound transition for nursing students, is fraught with stressors that precipitate anxiety, yet current risk assessment tools often lack explainability, limiting their ability to support personalized and proactive psychological interventions. To develop and validate an explainable machine learning model (ML) that predicts anxiety risk in nursing interns and to create a web-based calculator for early detection to facilitate adaptive clinical interventions. A total of 1,233 nursing interns(Class of 2025) were enrolled between January and April 2025 for model development. Anxiety was defined using the Generalized Anxiety Disorder Scale (GAD-7) as the standard assessment tool. Six ML methods, including logistic regression (LR), support vector machine (SVM), gradient boosting machine (GBM), neural networks (NN), random forest (RF), and XGBoost, were implemented to predict the risk of anxiety. External validation was conducted using data from 383 nursing interns(Class of 2026) recruited between July and August 2025 to evaluate the model’s generalizability. The optimal model was selected on the basis of the AUC. This study also applied the SHAP approach to explain the model’s outcomes. Additionally, the best-performing model was used to develop a web-based anxiety risk calculator. Sleep disorder, poor academic performance, high stress level, low social support, and a history of mental illness were identified as important predictors. The GBM model exhibited the best performance, achieving an AUC of 0.894 in internal validation and 0.895 in external validation. This study used explainable ML methods to predict anxiety among Chinese nursing interns and to identify five significant risk factors. The GBM model exhibited the best performance. Coupled with its explainability, this tool holds potential clinical utility for designing personalized, targeted psychological interventions for high-risk individuals. Not applicable.