A Preliminary Look at Predicting Caries Lesion Patterns/Timing in Children

Objectives: Prediction of caries lesion patterns, timing and trajectories can aid in targeting cost-effective preventive interventions. Objective: To predict patterns and timing of development of cavitated caries lesions in children using a classification tree approach
Methods: 1325 primary caregiver (PCG)-child pairs were recruited and followed over time. At child’s age 1, 2.5, 4 and 6.5 years, children received caries examinations using the International Caries Detection and Assessment System criteria (ICDAS) and PCGs completed a 52-item risk questionnaire. Additional clinical variables were also assessed (i.e., spacing/crowding of incisors and constricted tongue frenum). Bitewing radiographs were taken at age 6.5. Participants were categorized into 4 groups by their prevalence of cavitated caries lesions (d₃mft+D₃MFT>0; d/D=ICDAS≥3) based on combinations of location (anterior/posterior, mandibular/maxillary, smooth/interproximal/occlusal), and age/timing when detected. A classification tree to assess the relationships between risk factors and caries patterns/timing groups was modeled using a recursive partitioning algorithm.
Results: The groups were: 1=(No lesions)=20%; 2=(Lesions anywhere by age 2.5)=9%; 3=(Anterior lesions starting at age 4 and/or 6.5, with or without posterior lesions)=18%; and 4=(Only posterior lesions, starting at age 4 and/or 6.5)=53%. The classification tree correctly classified 65% of the participants’ caries pattern/timing groups. Risk factors selected by the algorithm were: participation in a public assistance program, lower brushing frequency, higher child’s caries risk score at age 1, child having ≥3 closed mandibular anterior contacts, higher frequency of PCG consuming sugary drinks, higher number of children in the household, higher frequencies of PCG and child eating or drinking anything other than water before bed, higher frequency of PCG’s gums bleeding, PCG sharing food with the child using the same plates/glasses/utensils, and PCG having cavities within the previous 2 years.
Conclusions: Preliminary analyses suggest that a classification tree to predict caries patterns and timing of cavitated caries lesions development has moderate accuracy.