Utilizing Electronic Dental Records to Predict Neuro-Degenerative Diseases in a Dental Setting: A Pilot Study.

Abstract

Limited research demonstrates the possible correlations between dental diseases and neurodegenerative diseases like Alzheimer's disease (AD) and Parkinson's disease (PD). Nevertheless, dental diseases are often overlooked while assessing the risk of AD and PD in clinical settings. It is unknown whether AD/PD risk can be predicted using electronic dental record (EDR) data collected in a routine dental setting. This pilot study determined the feasibility of predicting AD/PD using 84 features routinely captured in the EDR. We utilized the Temple University School of Dentistry clinic data of 27,138 patients. Using a natural language processing (NLP) approach (accuracy=97%), we identified patients with AD/PD and their matched controls (matched by age and gender). XGBoost machine learning model with 10-fold cross-validation was applied for prediction. With 77% accuracy, we found 53 features significantly associated with AD/PD that could be utilized to predict the risk of AD/PD. Further studies are warned to confirm these findings.