Assess whether machine learning models can predict dental caries presence and severity across age groups using data from Brazil’s 2023 Oral Health Survey.
Approach:
Population: The analysis included 6,698 children aged 12, 8,041 adolescents aged 15–19, and 8,987 adults aged 35–44.
Data and methods: Researchers analyzed demographic, socioeconomic, behavioral, and clinical data and applied several machine learning algorithms. SHAP values were used to assess the contribution of variables to model predictions.
Key Findings:
Model performance varied by age group; discrimination of caries presence was stronger among adults.
Wealth and household density were the leading structural predictors across age groups, and socioeconomic factors outweighed individual clinical markers in the SHAP analysis.
Non-White participants had a higher caries burden across age groups. Private dental insurance and recent dental visits were associated with lower caries prevalence.
Among 12-year-olds, untreated caries was reported in 62.5% of those without piped water access and 35.5% of those with access.
Interpretation:
The authors report that the models can help identify population groups with elevated caries risk and may support targeted public health programs.
Limitations:
The cross-sectional design does not allow causal conclusions.
The study focused on urban populations and did not externally validate the models.
Findings may not translate directly to countries with different dental-care access and socioeconomic conditions.
Conclusion:
The researchers plan to repeat the approach using country-specific factors; a parallel analysis using US National Health and Nutrition Examination Survey data is in preparation.