Developing and evaluating machine learning-based risk models for metabolic syndrome among nurses: a cross-sectional study.

Background: Metabolic syndrome is a group of risk factors that increase the risk of cardiovascular disease. Nurses have a higher risk of metabolic syndrome due to high work pressure and irregular lifestyle. Early recognition of metabolic syndrome is essential to prevent related complications and improve nurse health. Methods: A cross-sectional survey was used to collect valid questionnaire data of 900 nurses. Univariate analysis and multivariate Logistic regression analysis were used to screen the key variables. Logistic regression, random forest, XGBoost, support vector machine and neural network were used to predict and evaluate the performance of the model. SHAP values were also used to explain the importance of model features. In addition, 227 nurses were collected for external validation of the optimal model. Results: The random forest model had the best performance on the test set [AUC = 0.890 (0.855-0.925)], and the AUC of the external validation was 0.80 [0.762-0.838]. Among them, BMI, job burnout, shift work, age and other 9 characteristics were the key factors affecting the prediction. Discussion: The superior performance of the random forest model and identification of BMI and job burnout as core predictors provide evidence-based support for integrating machine learning-based screening into nurse occupational health policies, enabling early identification and targeted intervention for high-risk nurses. Conclusion: This study successfully established the prediction model of metabolic syndrome in nurses. The random forest model has high accuracy. The key factors identified provide theoretical support for early intervention, provide quantitative tools for identifying high-risk nurses, support personalized health intervention, and help to improve the overall health level and work efficiency of nurses.