Developing an Interpretable Machine Learning Framework to Predict and Analyse Early Childhood Caries in Children Aged 2 to 6 Years: A Single-centre Observational Study.

BACKGROUND

Early childhood caries (ECC) is a prevalent global oral health condition, exerting profound adverse effects on children's oral health, general health, and quality of life. Its onset involves demographic, socioeconomic, behavioural, and clinical factors. While many studies report individual risks, systematic analyses to prioritise predictors remain limited.

METHODS

This single-centre observational study included 900 children who visited the hospital between October 2022 and May 2024. Collected variables included age, parental background, feeding practices, general health status, oral hygiene and dietary habits. Univariate and multivariate analyses were conducted to identify factors associated with ECC. Five machine learning (ML) models - logistic regression (LR), random forest (RF), decision tree (DT), k-nearest neighbours (KNN) and support vector machine (SVM) - were developed to predict ECC. ECC presence or absence was used as the clinical outcome. Risk stratification classified children into low-, moderate-, and high-risk groups.

RESULTS

Significant predictors included age, maternal education, vitamin supplementation during pregnancy and for the child, carbohydrate intake, nighttime feeding behaviours, cessation timing, light exposure during sleep and outdoor activity (P < .05). LR achieved the highest accuracy (99.4%), followed by SVM (98.9%), RF (96.7%), DT (95.0%) and KNN (91.1%).

CONCLUSION

The machine learning models effectively identified key determinants of ECC, underscoring the critical roles of maternal education, dietary patterns, and lifestyle-related factors. Risk-based stratification derived from these models may inform targeted early preventive interventions in both clinical and community settings, thereby contributing to a reduction in the overall burden of ECC.