Machine learning-based prediction of occupational exposure risks among oral healthcare workers

This study aims to identify the key risk factors for occupational exposure among oral healthcare workers and develop a predictive model using machine learning algorithms to lay the foundation for early screening of high-risk populations and the formulation of preemptive intervention plans.

A multicenter cross-sectional study was conducted among 367 oral healthcare workers in 27 hospitals in Tianjin, China, from January 2025 to June 2025. Data were collected via an online questionnaire, encompassing demographic information, Work Preference Inventory, Organizational Climates, resilience, and other relevant factors. Logistic regression, random forest, decision tree, and XGBoost algorithms were employed to construct predictive models. The models were evaluated based on the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, and F1 score.

The incidence rates of occupational exposure in the modeling and validation groups were 15.5% and 16.5%, respectively. Univariate analysis revealed significant differences between the exposed and non-exposed groups in terms of Work Preference Inventory, Organizational Climates, resilience, professional title, hospital level, age, and gender. Multivariate analysis using logistic regression indicated that Work Preference Inventory, resilience, Organizational Climates, professional title, hospital level, and gender were independent risk factors for occupational exposure. The random forest model exhibited the best predictive performance, with an AUC of 0.755, accuracy of 89.2%, sensitivity of 56.3%, specificity of 94.7%, and F1 score of 0.600.

This study successfully identified the key risk factors for occupational exposure among oral healthcare workers and developed a predictive model using the random forest algorithm. These findings can guide the development of targeted interventions to mitigate the risks of occupational exposure. Future research should focus on validating the model with larger and more diverse datasets.