Screening for Periodontitis Using Blood Biomarkers and Demographic Data: A Machine Learning Study

Area of Science:

• Biomedical Science
• Public Health
• Preventive Medicine

Background:

• Periodontitis is a prevalent chronic inflammatory condition.
• It is associated with systemic diseases like diabetes and cardiovascular disease.
• Current diagnosis relies on dental exams, often missed by at-risk populations.

Purpose of the Study:

• To evaluate the efficacy of routine blood biomarkers and demographic data for screening moderate-to-severe periodontitis.
• To explore a non-dental screening method for periodontitis.
• To identify key predictors for periodontitis risk.

Main Methods:

• Utilized National Health and Nutrition Examination Survey (NHANES) data (N=3,338).
• Trained an XGBoost classifier on 77 features including demographics, complete blood count, glycemic markers (HbA1c), and heavy metals.
• Assessed model performance using Accuracy, Precision, Recall, and F1 score; employed SHAP for interpretability.

Main Results:

• The model achieved 61.1% Accuracy, 57.4% Precision, 94.6% Recall, and 71.5% F1 score.
• Performance was higher in males (F1=78.2%) than females (F1=63.9%).
• Key predictors identified were age, gender, blood cadmium, blood lead, and HbA1c.

Conclusions:

• Routine blood biomarkers and demographics offer a feasible non-dental screening strategy for periodontitis in primary care.
• The model's high recall minimizes false negatives, identifying at-risk individuals.
• This approach supports integrating oral health into general medical care, especially for underserved populations.