Untreated dental caries in children and adolescents: a cross-sectional analysis of sociodemographic correlates and an explainable machine-learning model

Introduction: Untreated dental caries is a persistent and unequally distributed burden in US children, and machine learning has been proposed to predict it, but the relative contribution of sociodemographic features to an explainable model is not established. Methods: A cross-sectional cohort of 4871 children and adolescents aged 2 to 19 years was assembled from the NHANES 2017-March 2020 pre-pandemic oral health examination. Untreated caries was defined as any tooth with an untreated carious lesion. Crude and adjusted odds ratios were estimated with logistic regression, and a gradient-boosting model with permutation importance quantified the explainable contribution of each feature. Results: Untreated caries was present in 13.6% of participants. Higher income-to-poverty ratio was associated with lower odds of untreated caries (crude OR 0.67 per category; adjusted OR 0.65), and the gradient-boosting model reached an area under the curve of 0.574 with age as the most important feature. Conclusions: Untreated caries remains concentrated among children with lower family income, and a gradient-boosting model explained by age and income offers modest discrimination, which is consistent with the predominance of socioeconomic rather than purely clinical determinants.