Non-clinical determinants predicting implant longevity: Model development and validation

Abstract

Accurate prediction of dental implant longevity remains difficult using readily available non-clinical patient factors alone. Therefore, it is of interest to evaluate the internally validated 5-year implant longevity models using demographic, systemic and behavioural and maintenance variables from 864 implants in 326 patients. Logistic Regression, Random Forest and Gradient Boosting Machine models were assessed using AUC, Brier score, calibration and decision-curve analysis. The 5-year failure rate was 18.2% and Gradient Boosting Machine showed the best performance with an AUC of 0.64 and Brier score of 0.179. Thus, data shows the non-clinical variables provide moderate predictive accuracy and may support early risk stratification and future multimodal prediction models.