Background: The burden of cancers attributable to human papillomavirus (HPV) infection is increasingly evident among men. In line with the global trend toward gender-neutral prevention, China expanded HPV vaccine eligibility to males aged 9–26 years in 2025. However, male HPV vaccination uptake remains low, particularly in resource-limited regions. Therefore, identifying factors associated with HPV vaccine willingness (HPV-VW) among male university students is crucial for enhancing vaccination willingness and developing targeted interventions.
Objectives: To assess HPV-VW and identify key influencing factors among male university students in resource-limited regions of China, ultimately quantifying the potential benefits of HPV-VW enhancement under different intervention scenarios.
Methods: A cross-sectional survey was conducted among male university students in Gansu Province, collecting demographic and behavioural characteristics and measuring HPV-related knowledge (Cronbach’s α=0.956), attitudes (Cronbach’s α=0.924), behaviours (Cronbach’s α=0.885) (KAP), and HPV vaccine hesitancy (Cronbach’s α=0.934). Data were split into training and test sets (80:20). 9 machine learning models (XGBoost, Gradient Boosting, Random Forest, LightGBM, AdaBoost, Logistic Regression, Extra Trees, Decision Tree, and Naive Bayes) were trained, hyperparameters were tuned using Bayesian optimization with fivefold cross-validation, and performance was evaluated on the test set. Feature-importance rankings were integrated across models to identify the top 5 factors, and Shapley additive explanations (SHAP) was applied to the best-performing model for interpretability and interaction analysis. Then, Structural Equation Model (SEM) (bootstrap=500) was built using the top 5 factors to estimate direct and indirect effects. Finally, combining the optimal machine-learning model and SEM path analysis, population-wide scenario simulations were conducted for the top 5 factors to estimate potential increases in HPV-VW under different intervention options.
Results: A total of 816 male university students were included, with an overall HPV-VW rate of 60.05%. The AUC values of all models exceeded 0.800, with XGBoost demonstrating the best performance (AUC=0.899, sensitivity=0.877, specificity=0.810, accuracy=0.854, F1=0.886). Aggregating the importance rankings results from 9 models, the top 5 factors were attitude score, behavior score, vaccine hesitancy score, knowledge score, and family monthly income. SHAP analysis based on XGBoost showed vaccine hesitancy was negatively associated with HPV-VW, while family monthly income and KAP factors were positively associated with HPV-VW. SEM indicated that vaccine hesitancy had a direct effect on HPV-VW, while family monthly income and KAP factors had both direct and indirect effects (Figure 1). Intervention scenario simulations indicate that  increasing family monthly income by 1, 2, or 3 levels raises HPV-VW rates to 67.03%, 70.96%, and 75.25%, respectively. Increasing knowledge or attitude scores by 60% raised HPV-VW rates to 99.39% and 100.00%, respectively; increasing behavior scores by 70% raised HPV-VW rates to 94.98%; and reducing vaccine hesitancy by 50% raised HPV-VW rates to 66.42%.
Conclusions: HPV-VW is associated with KAP, vaccine hesitancy, and family economic status among male university students in limited-resource regions. Targeted education to improve KAP and reduce hesitancy, together with measures that lower out-of-pocket costs, may increase equitable uptake of male HPV vaccination.

Figure 1. Structural Equation Model of Top 5 Factors and HPV Vaccination Willingness