SDoH Impact on Periodontal Disease Using Machine Learning and Dental Records

Abstract

The impact of social determinants of health (SDoH) on periodontal disease (PD) is critical to study, as a deeper understanding of SDoH offers significant potential to inform policy and help clinicians provide holistic patient care. The use of machine learning (ML) to analyze the association of SDoH with PD provides significant advantages over traditional statistical methods. While statistical approaches are effective for identifying trends, they often struggle with the complexity and unstructured nature of data from dental electronic health records (DEHRs). The objective of this study was to determine the association between PD and SDoH using big data through linked DEHR and census data using ML. We used the records of 89,937 unique patients (754,414 longitudinal records) from the Temple University School of Dentistry who received at least 1 treatment between 2007 and 2023. Patient PD outcomes were categorized based on progression, improvement, or no change, using longitudinal data spanning up to 16 y. We applied ML models, including logistic regression, Gaussian naive Bayes, random forest, and XGBoost, to identify SDoH predictors and their associations with PD. XGBoost demonstrated the best performance with 94% accuracy and high precision, recall, and F1 scores. SHapley Additive exPlanations (SHAP) values were used to provide explainable ML analysis. The leading predictors for PD progression were higher social vulnerability index, poverty, population density, fewer dental of…