Development Of A Caries Risk Assessment Tool For School Children

To develop a modified caries-risk assessment tool for school children (SC-MCAT) and use machine learning algorithms to validate this tool among the children in Xixiangtang District, Nanning City, Guangxi, Southern China. A longitudinal study was conducted among 723 children aged 7 years old for one year. Data from the questionnaire and objective examination was entered into the generalized linear model to analyze the relation of the risk level and caries increment. The discrimination ability of SC-MCAT was assessed by ROC analysis, while the calibration was tested by the Hosmer-Lemeshow (HL) test. The feasibility of SC-MCAT was verified through machine learning algorithms. Accuracy, sensitivity, specificity and AUC operating curve were computed. The children were devided by SC-MACT into three caries risk levels, namely low, moderate, and high. Trend Chi-square Test showed that the caries incidence prevalence increased as the risk levels increased (P<0.001). The likelihood of mean DMFS was 5.306 times higher (95% CI=3.506∼8.032, P<0.001) in the high-risk group compared to the low-risk group. HL test indicated good calibration (P=1.00). Random forest showed accuracy of 0.85, sensitivity of 0.84, and specificity of 0.93, and AUC of 0.936. The SC-MCAT could effectively distinguish different caries risk levels, and a higher risk level corresponds to an increased likelihood of future caries increment in permanent teeth among the school children in Guangxi.