Predicting plaque-gingivitis risk in schoolchildren using an interpretable machine learning model: a cross-sectional study

Comprehensive Summary

This research, presented by Wu et. al examined the application of preventive dentistry through an interpretable machine learning model and SHAP method that determined plaque induced, gingivitis risk of pediatric patients. A cross sectional survey was administered to children aged 6-12 (n = 1755) in Lanzhou, China that consisted of 22 questions processed by select ML algorithms namely, LightGBM, RF, LR, XGBoost, Decision Tree DT, and KNN. Results indicated 51.3% of patients clinically diagnosed with plaque‑induced gingivitis accounting for 901 individuals of the 1,755 participants. Random Forest (RF) and LightGBM machine learning models achieved excellent discrimination scores (AUC) above 0.9 in training and testing (0.991 and 0.909; 0.970 and 0.904) respectively. SHapley Additive exPlanations (SHAP) further interpreted significant contributing factors including age, yearly income, brushing duration and frequency, regularity of dental check-ups, and incidence of bleeding gums upon brushing. Wu et. al proposed future directions applying interpretable ML across the pediatric public health sector in an effort to increase the generalizability of the study given that the ML models were only tested on one dataset.

Outcomes and Implications

By providing accessible preventative dentistry to vulnerable populations, the effort to close the health gap is furthered while plaque-induced gingivitis risk can simultaneously be minimized. The implementation of such risk assessments presents a new screening tool that could effectively lower high risk plaque-induced gingivitis in children. Not only that, but the ability to implement early interventional medicine could significantly reduce the incidence of plaque-induced gingivitis in various socioeconomic demographics.