Early childhood caries (ECC) is a major global public health concern, necessitating its early screening. This study aimed to establish a caries risk assessment (CRA) platform for managing caries in community preschool children in underdeveloped regions of Northwest China. We collected clinical examination and questionnaire data of children aged 3 to 5 years in six regions of Gansu Province. Then we selected variables using least absolute shrinkage and selection operator (LASSO) regression and constructed a CRA model utilizing multivariate logistic regression analyses. The predictive performance was assessed by the receiver operating characteristic (ROC), calibration, clinical decision and impact curves. The subgroup application was analyzed on the basis of the residence of children. The CRA model included age, residence, feeding pattern within six months of birth, history of toothache, and history of dental visits as predictors. The Hosmer-Lemeshow test showed that the model fitting was acceptable (χ2 = 7.049, P = 0.531). And the model exhibited an excellent discriminatory performance in external cohort, as evidenced by the ROC curve parameters, with an area under the ROC curve (AUC) of 0.804 (95% CI: 0.765-0.844), sensitivity of 0.807, and specificity of 0.660. The calibration curves showed that the model exhibited good predictive accuracy, and the clinical decision and impact curves showed that the model was useful within reasonable threshold probabilities. Finally, we created an online prediction platform to ensure public use of the CRA model. The presented CRA model offers a novel public platform with the potential to serve as an effective tool for the screening and management of deciduous caries at the community level in underdeveloped regions of Northwest China.
Zhang et al. (Tue,) studied this question.