AI-driven predictive modelling of orthodontic relapse using retainer compliance and patient factors

Orthodontic relapse remains a critical concern, often compromising long-term treatment success and patient satisfaction. Therefore, it is of interest to develop and validate an AI-driven predictive model using SMART microsensor-based retainer compliance data and patient-specific variables. Among 156 monitored patients over 24 months, the Random Forest algorithm achieved the highest accuracy (92.3%), sensitivity (89.7%) and specificity (94.2%). Key predictors included daily retainer wear duration, treatment complexity, age at completion and initial malocclusion severity. The model supports personalized retention strategies and early intervention to enhance post-treatment stability.