(A130) Leveraging Machine Learning to Identify Risk Factors for HPV-Related Oropharyngeal Cancer and Sex-Based Differences

Introduction: The increasing incidence of oropharyngeal cancers, particularly those attributed to human papillomavirus (HPV), has become a significant public health concern, disproportionately affecting men. This study aims to utilize machine learning (ML) techniques to identify the key factors driving this trend, with a particular focus on sex-based differences.

Methods: We retrospectively analyzed Massachusetts public health data from 2001 to 2022, incorporating demographic, behavioral, and environmental factors from 352 towns. The analysis was conducted in January 2025. Six machine learning models—Logistic Regression, Random Forest (RF), Decision Tree, Support Vector Machine (SVM), Kernel SVM, and K Nearest Neighbors (KNN)—were trained and evaluated using the area under the receiver operating characteristic curve (AUROC), accuracy, and F1-score, with confidence intervals (CIs) derived from bootstrapping. Dimensionality reduction was applied for feature selection.

Results: Among the 352 towns, the Random Forest model exhibited the best performance for females, with an AUROC of 0.636 (CI: 0.559–0.709), accuracy of 0.636 (CI: 0.554–0.683), and an F1-score of 0.503 (CI: 0.411–0.584). For males, the K Nearest Neighbors model yielded the best performance, with an AUROC of 0.666 (CI: 0.589–0.738), accuracy of 0.663 (CI: 0.596–0.725), and an F1-score of 0.504 (CI: 0.403–0.600). Feature selection for females did not identify significant predictors, whereas for males, increased incidence was significantly associated with low educational attainment, poverty, physical inactivity, limited English proficiency, and being of non-white race or residing in non-family households.

Conclusions: Machine learning models effectively identified key risk factors contributing to the rising incidence of HPV-related oropharyngeal cancer in men. These findings provide valuable insights for refining risk assessment strategies and guiding targeted interventions aimed at reducing the burden of oropharyngeal cancer and improving early detection.