Derivation of a Risk Score for High Caries Risk in 3- to 5-year-old Children in Sichuan Province.

Abstract

To explore potential caries risk indicators in 3- to 5-year-old children, and develop a simple risk-score model to screen the children at high risk of caries with decayed, filled, and missing teeth (dmft) > 2. A cross-sectional study involving 2746 children 3 to 5 years of age was conducted in Sichuan province. Children were examined for dmft index, and sociodemographic and behavioural factors were acquried through a questionnaire completed by their caregivers. A prediction model was developed by backward multivariate logistic regression, and its overfitting degree was examined with 5-fold cross-validation. A simple risk-score model was derived to screen the children with dmft > 2 at high risk of caries with the β regression coefficient obtained from the multivariate regression model. A child's oral health status was identified as the highest risk indicator with a β regression coefficient of 1.093. The mean area under curve (AUC) from the 5-fold cross-validation was 0.7408 (95% CI: 72.21%, 75.95%), with a bias of only ca 1%. This result allowed us to eliminate substantial overfitting of the prediction model. The AUC of the risk scoring system was 0.7455 (95% CI: 72.70%, 76.40%), which indicated good screenability. This risk score model has the advantages of simplicity, low cost and relatively high accuracy, and is suitable for use in developing countries, especially for primary screening for high risk of caries. It shows that certain child behaviours and parental attitude play an important role in dental caries among preschool children.