Multivariate features selection from demographic and dietary descriptors as caries risk determinants in oral health diagnosis: Data from NHANES 2013–2014

The main pathology that make up the epidemiological profile of oral health status is dental caries. This pathology is considered a challenge because its high prevalence, besides being a chronic but preventable disease, which can be caused by a series of different demographic, dietary and laboratory data, among others. Therefore, in this research two multivariate models were proposed for the classification between control patients (‘0’), presence of caries (‘1’) and presence of restorations (‘2’). Analyzing 189 demographic and dietary features from NHANES 2013–2014, by FBS method based on the P-value they were reduced to 24 features, which were separated in two sets according to their demographic or dietary information. Both datasets were subjected to a statistical analysis based on linear regression, obtaining their residuals, AUC, ROC and OR values, in order to validate their classification accuracy. For each model were obtained significant statistical results and none of them presented problems related to outliers, according to their residuals. Both models shown P-values > 0.05 for most features and small OR confidence intervals. Finally, demographic data model obtained an AUC = 0.664 and dietary data model obtained an AUC = 0.633, averaging the AUC values for each outcome, proving to be statistically significant. By these results it's concluded that these specific demographic and dietary features are significant determinants for estimating the oral health status patients, based on their likelihood of developing caries.