Data-driven prediction model for periodontal disease based on correlational feature analysis and clinical validation

Area of Science:

• Periodontology
• Data Science
• Machine Learning

Background:

• Periodontal disease poses a significant public health challenge.
• Accurate prediction models are crucial for early intervention and management.
• Existing prediction methods may lack comprehensive risk factor integration.

Purpose of the Study:

• To develop and validate data-driven models for predicting periodontal disease.
• To identify key risk factors for periodontitis using correlational analysis.
• To assess the performance of various machine learning algorithms in periodontal prediction.

Main Methods:

• Utilized data from the 7th Korea National Health and Nutrition Examination Survey (n=10,654) for risk factor identification.
• Developed and validated prediction models using logistic regression, SVM, random forest, XGBoost, and neural networks.
• Performed internal validation via 5-fold cross-validation and external validation with clinical data (n=120).

Main Results:

• Identified 16 significant risk factors for periodontitis from over 1000 potential factors.
• The XGBoost model achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.823 (internal) and 0.796 (external).
• Models accurately predicted severe bone loss (AUC=0.813), gingival bleeding (AUC=0.694), and tooth loss (AUC=0.734).

Conclusions:

• Data-driven models incorporating 16 risk factors demonstrate robust performance for predicting periodontal disease.
• These models offer enhanced prediction accuracy and reproducibility across internal and external validation datasets.
• The findings support the use of data-driven approaches for improved periodontal health assessment and patient counseling.