Developing caries inequalities risk prediction tools for children under the age of six

Abstract

Introduction:
Dental caries in deciduous teeth is the 10th-most prevalent condition, affecting 621 million children worldwide (Kassebaum et al., 2015). Although overall prevalence of Early Childhood Caries (ECC) is declining, but there is no improvement in the unequal distribution of caries across social gradients. It is therefore important to investigate and demonstrate the relationships between the factors responsible for caries inequalities in children under the age of six.
Methods
This thesis used systematic review methods and the Cochrane-validated tool (PROGRESS-PLUS) to explore factors implicated in ECC inequalities. Existing conceptual ECC frameworks were then evaluated (mapped) against factors identified from the review. The relationship between identified factors were established using Directed Acyclic Graphs and path analyses to create new conceptual models, and their predictability assessed using traditional and machine learning statistical techniques.
Results
Sixty-seven publications were eligible for this review and the result showed that there are 24 ECC-related risk factors. The mapping of existing caries frameworks also revealed that none of the current caries conceptual model had more than 48% relevant Social Determinants of Health (SDH) in their frameworks. Two separate ECC conceptual frameworks (temporal and hierarchical) were developed based on systematic review evidence, and the predictability of the models also showed an Area under Receiver Curve (AUC) of 76% and 74% for the child-level and area-level prevalence prediction respectively.
Conclusion
This thesis found that existing conceptual frameworks lacked in-depth consideration of the full wider determinants that are responsible for health inequalities, which may be the cause of the current low ECC prevalence, but widening or stagnant caries inequalities pattern seen globally. The model developed in this thesis included more SDH content (64%) than any existing models and therefore increases the chances that preventive measures, built on the new SDH models, will be able to address both prevalence and inequalities. This model also demonstrated good predictive capabilities.