IADR Abstract Archives
Predicting Caries Using Social and Familial Factors: A Nationwide Classification-and-Regression-Tree Analysis
Objectives: The family and social environment are likely of great importance to children’s dental health. However, social and familial factors have never been evaluated as isolated caries predictors at the individual level. This nationwide study examined the discriminant ability of sibling caries and various other social and family-level factors in predicting caries risk.
Methods: This study included all 15-year-olds in 2003 (index-siblings) and their biological siblings (co-siblings) born within ±3 years. For each individual, data on the outcome and risk predictors were compiled after linking the national dental, social, and population registers. The outcome was caries experience in co-siblings, measured by the DMFS index. The predictors included index-sibling caries, socioeconomic position (parental education, occupation, and income), gender, co-sibling birth order, ethnicity, and household type. The discriminant ability of the predictors was assessed using classification and regression tree (CART) analyses. Using CART, both fully-saturated and the simplest clinically-relevant decision trees that retained useful predictive power were generated. The predictive power of the models was evaluated using the Area under the Receiver Operating Characteristic curve (AUROC) statistic (AUROC: ≥0.8, excellent; 0.7–0.79, useful).
Results: There were 23,847 sibling pairs (n=47,694) in the study. The prevalence of caries experience (DMFS>0) in the study population was 73.6%. The overall predictive power of the CART models ranged from useful to excellent (AUROC: fully-saturated trees, 0.8–0.82; clinically-relevant trees, 0.7). Index-sibling caries yielded the greatest influence in predicting co-sibling caries (~67% higher than parental education, the next best surrogate predictor). The simplest clinically-relevant tree contained only index-sibling caries and a socioeconomic position indicator as predictors. This model demonstrated perfect sensitivity (but poor specificity). Per this model, caries could be expected in ≥84% of co-siblings of adolescents with ≥3 caries-affected tooth surfaces (DMFS ≥2.94).
Conclusions: Caries in a sibling might suggest preventive family-based approaches targeting co-siblings.
Methods: This study included all 15-year-olds in 2003 (index-siblings) and their biological siblings (co-siblings) born within ±3 years. For each individual, data on the outcome and risk predictors were compiled after linking the national dental, social, and population registers. The outcome was caries experience in co-siblings, measured by the DMFS index. The predictors included index-sibling caries, socioeconomic position (parental education, occupation, and income), gender, co-sibling birth order, ethnicity, and household type. The discriminant ability of the predictors was assessed using classification and regression tree (CART) analyses. Using CART, both fully-saturated and the simplest clinically-relevant decision trees that retained useful predictive power were generated. The predictive power of the models was evaluated using the Area under the Receiver Operating Characteristic curve (AUROC) statistic (AUROC: ≥0.8, excellent; 0.7–0.79, useful).
Results: There were 23,847 sibling pairs (n=47,694) in the study. The prevalence of caries experience (DMFS>0) in the study population was 73.6%. The overall predictive power of the CART models ranged from useful to excellent (AUROC: fully-saturated trees, 0.8–0.82; clinically-relevant trees, 0.7). Index-sibling caries yielded the greatest influence in predicting co-sibling caries (~67% higher than parental education, the next best surrogate predictor). The simplest clinically-relevant tree contained only index-sibling caries and a socioeconomic position indicator as predictors. This model demonstrated perfect sensitivity (but poor specificity). Per this model, caries could be expected in ≥84% of co-siblings of adolescents with ≥3 caries-affected tooth surfaces (DMFS ≥2.94).
Conclusions: Caries in a sibling might suggest preventive family-based approaches targeting co-siblings.