Periodontitis Prediction Model Using Linked Electronic Health and Dental Records.

Area of Science:

• Oral Health Research
• Medical Informatics
• Artificial Intelligence in Healthcare

Background:

• Periodontal disease (PD) is linked to systemic health, yet predictive models often exclude medical data.
• Existing models focusing solely on dental records limit comprehensive risk assessment.

Purpose of the Study:

• To enhance periodontal disease (PD) risk prediction by integrating electronic dental records (EDRs) with electronic health records (EHRs) using machine learning (ML).

Main Methods:

• Utilized EDR and EHR data from adult patients, linking dental and medical information.
• Developed automated feature reduction techniques for extensive EHR data preprocessing.
• Trained and evaluated ML models (Gaussian Naive Bayes, Random Forest, LightGBM, XGBoost) using cross-validation.

Main Results:

• The best model, using chi-square selected features, achieved 85% specificity, 67% sensitivity, and 84% AUC.
• Dental factors like oral hygiene and smoking status were dominant predictors.
• Systemic conditions (cardiovascular, endocrine, renal) meaningfully contributed to PD risk prediction.

Conclusions:

• Strong associations between systemic health and oral health were observed, despite dental factors being primary predictors.
• Integrating medical and dental records can guide clinicians in identifying high-risk patients for earlier intervention.
• AI-driven models show potential for personalized care and improved interdisciplinary health management.