Developing Children's Oral Health Assessment Toolkits Using Machine Learning Algorithm

Area of Science:

• Pediatric Dentistry
• Public Health
• Health Informatics

Background:

• Assessing children's oral health status and treatment needs presents significant challenges.
• Existing methods may be resource-intensive or lack predictive accuracy.

Purpose of the Study:

• To develop and validate oral health assessment toolkits for predicting the Children's Oral Health Status Index (COHSI) score.
• To predict referral for treatment needs (RFTN) in pediatric populations using machine learning algorithms.

Main Methods:

• Developed parent and child survey toolkits with short-form items and demographic information.
• Utilized machine learning algorithms (Extreme Gradient Boosting, Naive Bayesian) on data from 545 families.
• Evaluated toolkit performance using metrics like RMSE, correlation, sensitivity, and specificity.

Main Results:

• Toolkits demonstrated strong predictive capabilities for COHSI score (correlation up to 0.91) and RFTN (sensitivity 93%, specificity 49%).
• External validation confirmed the predictive accuracy of the developed toolkits.
• Machine learning models effectively predicted oral health status and treatment needs.

Conclusions:

• Survey responses from children and parents are predictive of clinical oral health outcomes.
• The toolkits are suitable for baseline assessments in school populations and for program evaluation.
• These tools can stratify samples for oral health research based on predicted needs and status.