Dataset for Machine Learning Guided Mortality Risk of Tooth Loss and Glycemic Control

This repository contains supplementary materials, underlying statistical code, and summary findings supporting the manuscript titled "Machine Learning Guided Mortality Risk of Tooth Loss and Glycemic Control." This study investigates the combined mortality risk of substantial tooth loss and elevated HbA1c, exploring their synergistic impact as a high-risk oral-metabolic phenotype. Utilizing data from 33,748 adults in the National Health and Nutrition Examination Survey (NHANES; 1999–2014) linked to 2019 National Death Index mortality records, the analysis employs a robust temporal split methodology. Explainable Boosting Machines (EBM) were applied to a discovery cohort to uncover data-adaptive, nonlinear prognostic severity gradients—identifying critical risk accelerations at HbA1c ≥ 11.36% and ≥ 17.5 missing teeth. Subsequently, survey-weighted, cause-specific Cox, and Fine-Gray regression models in a validation cohort demonstrated that the combined phenotype carries a near-twofold all-cause mortality hazard (HR 1.97) and a heavily pronounced heart-disease mortality burden (csHR 2.50). The findings confirm that substantial tooth loss serves as a powerful marker for heart-disease vulnerability when co-occurring with diabetes-range HbA1c, highlighting the necessity for integrated cardiometabolic risk reviews and bidirectional medical-dental referrals.