Developing an Original Population-Based Dental Caries Risk Prediction Model for Nigerian Children Using Machine Learning

Population-level caries risk prediction enables targeted preventive resource allocation, and developing an original machine learning model for Nigerian children that integrates clinical, dietary, behavioural, and socioeconomic predictors provides a tool with direct public health application. This study developed and validated an original population-based dental caries risk prediction model for Nigerian children aged 6 to 12 years using machine learning methods. A multi-zone cross-sectional dataset was compiled from 3,200 children across six geopolitical zones with 42 predictor variables encompassing clinical measures (baseline DMFT, plaque index), dietary data (sugar frequency, meal patterns), behavioural data (brushing frequency, dental visit history), socioeconomic indices, and environmental variables (water fluoride, rural/urban designation). Caries incidence at 18 months was the prediction target. Five machine learning algorithms were compared: logistic regression, random forest, gradient boosting, support vector machine, and a neural network. The gradient boosting model achieved the highest AUC of 0.84 on the independent validation set (n = 400). Top predictors were baseline caries experience, sugar-sweetened beverage consumption frequency, toothbrushing frequency, and residential area type. The model demonstrated reasonable performance consistency across all six zones. An interpretable risk stratification tool derived from the model assigned children to low, medium, and high caries risk categories with distinct preventive protocol recommendations. Expert review by 16 specialists confirmed the model's methodological rigour. The study recommends integration of the tool into school health screening platforms and state-level preventive resource allocation decisions.

Keywords: dental caries prediction, machine learning, risk prediction model, Nigerian children, public health dentistry