Developing and testing a prediction model for periodontal disease using machine learning and big electronic dental...

Area of Science:

• Dental Research
• Machine Learning Applications
• Public Health

Background:

• Periodontal disease (PD) affects 42% of the US population, despite advances in treatment.
• Early identification of high-risk patients is crucial for effective PD prevention.
• Existing PD prediction models have suboptimal performance, limiting their clinical utility.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) model for predicting periodontal disease (PD) risk.
• To utilize electronic dental record (EDR) data for large-scale, up-to-date PD risk assessment.
• To identify novel risk factors associated with PD development and progression.

Main Methods:

• A cohort of 27,138 dental patients was analyzed, categorizing PD into healthy, mild, and severe.
• An XGBoost ML model was trained on 74 features extracted from EDR data (80% training, 20% testing).
• A five-fold cross-validation strategy was employed to optimize model hyperparameters.

Main Results:

• The prediction model achieved an average area under the curve of 0.72 in differentiating between healthy, mild, and severe PD cases.
• New risk associations were identified, including patient anxiety, chewing/speaking difficulties, substance use, and various medical conditions (e.g., osteoporosis, cardiovascular diseases).
• The model demonstrated promising performance in predicting PD risk using EDR data.

Conclusions:

• Machine learning models utilizing EDR data show promise for predicting periodontal disease risk.
• The developed model identified novel patient-level and medical condition-related risk factors.
• This approach may aid clinicians in implementing preventive strategies by providing insights into PD risks and progression factors.