Application of Explainable AI (XAI) in Periodontal Disease Risk Assessment

Periodontal disease is a common, progressive condition that leads to tooth loss and contributes to systemic issues such as diabetes and cardiovascular disease. Although prior studies have linked smoking, obesity, and diabetes to periodontitis, few have leveraged explainable machine learning models to provide transparent, personalized risk predictions. In this study, I used the 2013–2014 National Health and Nutrition Examination Survey (NHANES) dataset (n = 3,720) to develop and compare two classifiers Random Forest and XGBoost on a stratified 60%/20%/20% train–validation–test split repeated across multiple seeds and trials. To make the models’ decisions interpretable, I applied SHAP (SHapley Additive exPlanations) to quantify each feature’s contribution to the prediction of severe periodontitis. SHAP identified age, body-mass index, and systolic blood pressure as the strongest drivers of risk, with additional insights from smoking status, diastolic blood pressure, diabetes status, and gender. A SHAP dependence analysis further revealed that advancing age increases predicted risk more steeply for males than for females. By combining robust model evaluation with patient-level explanations, this approach supports early identification of high-risk individuals and enhances patient provider communication and targeted prevention in dental practice and public health.