Predictability of Oral Disease Progression across the Life Course:A 21-42-Year Longitudinal Study Using Routinely...

Area of Science:

• Dental Public Health
• Epidemiology
• Predictive Modeling

Background:

• Long-term oral health monitoring is crucial for effective public health strategies.
• Predicting oral disease progression aids in proactive dental care and resource allocation.
• Existing prediction models often lack validation with extensive longitudinal data.

Purpose of the Study:

• To evaluate the predictability of long- and medium-term oral disease progression over 21-42 years.
• To identify key clinical and radiographic predictors of oral disease progression in military personnel.
• To assess the utility of routinely collected dental data for risk stratification.

Main Methods:

• Analysis of longitudinal dental records from 198 Dutch military personnel with >20 years of follow-up.
• Development of prediction models using LASSO regression for long-term (baseline data) and medium-term (clinical history) risk assessment.
• Model performance evaluated using ROC curves with AUC values, corrected for optimism via bootstrapping.

Main Results:

• Long-term models demonstrated moderate predictability (AUC 0.705-0.724) for all outcomes; medium-term models showed lower predictability (AUC 0.609-0.676), primarily for extractions and DMFT increment.
• Consistent long-term predictors included year of birth, smoking, and military rank.
• Key medium-term predictors were highest caries score and radiographic bone loss progression.

Conclusions:

• Routinely collected clinical dental data can moderately predict oral disease progression.
• Long-term prediction models showed better performance than medium-term models.
• Prediction is influenced by prior disease experience and lifestyle factors, supporting targeted preventive interventions.