Developing an Original Computational Model for Predicting Orthodontic Treatment Duration in Nigerian Populations

Accurately predicting orthodontic treatment duration at case initiation improves patient informed consent, clinical scheduling, and treatment planning, yet no prediction models have been developed or validated for Nigerian orthodontic patient populations, whose morphological characteristics and compliance patterns differ from populations in which existing models were developed. This study developed and validated an original computational prediction model for orthodontic treatment duration in Nigerian orthodontic patients. A two-phase study was conducted using retrospective and prospective data from orthodontic clinics at University of Ibadan Teaching Hospital, Nnamdi Azikiwe University Teaching Hospital Nnewi, and University of Maiduguri Teaching Hospital. Phase 1 developed the model using retrospective data from 380 completed fixed orthodontic cases. Features included initial IOTN-DHC score, arch length discrepancy, ANB angle, overbite, overjet, extraction pattern, patient age, and appointment compliance rate. Random forest regression and multiple linear regression were compared. Phase 2 prospectively validated the model in 85 new cases. The random forest model achieved R-squared of 0.74 on the prospective validation set, significantly outperforming linear regression (R-squared 0.52). Mean absolute error in duration prediction was 3.2 months. Appointment compliance rate was the single strongest predictor. Extraction cases required significantly longer predicted durations. The original model provides a practical clinical tool for Nigerian orthodontic practice and recommendations include its integration as a digital planning aid at orthodontic clinics, further validation in North East zone populations, and publication as an open-access calculator.

Keywords: orthodontic treatment duration, prediction model, random forest, Nigeria, orthodontic planning