Dental caries remains the most prevalent chronic infectious disease of childhood and a major public health concern in both low- and high-income nations, even though it is largely preventable. The present study assessed the feasibility of identifying dental caries in children through a machine learning model based on parental perceptions of their child’s oral health collected via questionnaire. Data were gathered from 182 parents or caregivers and their children aged 2–7 years residing in Los Angeles County. The random forest algorithm was applied to determine which survey questions predicted active caries or overall caries experience. Three-fold cross-validation was implemented, and the cutoff point was defined by maximizing sensitivity and specificity with a minimum sensitivity of 70%. The predictive contribution of each survey item to the classification of active caries and caries experience was quantified using mean decreased Gini (MDG) and mean decreased accuracy (MDA). Strong predictors of active caries included parent’s age (MDG = 0.84; MDA = 1.97), unmet healthcare needs (MDG = 0.71; MDA = 2.06), and being African American (MDG = 0.38; MDA = 1.92). For overall caries experience, key predictors were parent’s age (MDG = 2.97; MDA = 4.74), reports of oral health problems within the past year (MDG = 2.20; MDA = 4.04), and the child having experienced dental pain (MDG = 1.65; MDA = 3.84). Findings support the viability of using parent-completed surveys analyzed through machine learning for caries risk screening among young children.
