Self-reported periodontitis: A multiethnic community-based validation study

Area of Science:

• Oral epidemiology
• Public health surveillance
• Diagnostic accuracy studies

Background:

• Periodontitis is highly prevalent in Asia, necessitating accessible screening methods.
• Clinical periodontal examinations are impractical for large-scale studies.
• Validation of self-reported measures and sociodemographic factors in Asian populations is limited.

Purpose of the Study:

• To clinically validate self-reported periodontal measures in a multiethnic Singaporean population.
• To develop predictive models combining sociodemographic data and self-reported measures for periodontitis detection.
• To assess the performance of machine learning models in predicting periodontitis.

Main Methods:

• Analysis of cross-sectional data from 426 participants undergoing periodontal examinations and completing the CDC-AAP self-reported questionnaire.
• Definition of periodontitis using the 2012 CDC-AAP case definitions.
• Application of multivariable logistic regression, AUC analyses, and five machine learning models for predictive modeling.

Main Results:

• A model combining self-reported loose teeth, age, and ethnicity showed good discrimination for severe periodontitis (AUC=0.76).
• Machine learning models achieved similar AUCs (0.67-0.76) but exhibited high specificity and lower sensitivity.
• Mild, moderate, and severe periodontitis were present in 16.4%, 42.5%, and 18.1% of participants, respectively.

Conclusions:

• Self-reported questions, particularly regarding loose teeth, are valuable for detecting severe periodontitis in this multiethnic cohort.
• Machine learning models show promise for scalable, data-driven periodontal screening, though require larger datasets for enhanced generalizability.
• Combining self-reported data with demographic variables offers a practical approach to periodontitis surveillance.