Periodontitis Prediction Model Using Linked Electronic Health and Dental Records

Comprehensive Summary

This retrospective predictive modeling study sought to create and test a machine learning model for periodontitis risk prediction by combining the use of linked electronic dental records (EDRs) and electronic health records (EHRs). With data from more than 20,000 adult patients treated at an academic dental institution, researchers were able to link dental data with medical diagnoses, medications, procedures, demographics, and social determinants of health. After significant preprocessing and feature reduction to account for the high dimensionality of EHR data, machine learning models were developed and compared. The model that performed best showed excellent discriminatory power, with an area under the receiver operating characteristic curve of about 0.84, reflecting good performance in distinguishing patients with periodontitis. While dental factors such as oral hygiene variables, smoking status, and previous periodontal treatment were the most important predictors, several medical conditions, including cardiovascular, metabolic, respiratory, renal, and hematologic diseases, also made significant contributions to periodontitis risk prediction. This study ends by asserting that the combination of medical and dental health information using machine learning techniques allows for effective periodontitis prediction and that the relationship between oral and systemic health is closely intertwined.

Outcomes and Implications

This research supports the integration of medical and dental electronic health data to enhance risk-based screening and preventive management of periodontitis. By utilizing machine learning algorithms in current health record systems, it may be possible to earlier recognize high-risk patients and design preventive strategies for them. The proven value of systemic diseases further validates the two-way relationship between oral inflammation and systemic diseases, and this further supports an interprofessional approach to patient management. The application of such predictive algorithms in health systems may help in population-level monitoring and resource allocation, and this may further advance personalized dentistry. Finally, the integration of medical and dental data analysis may help in improving systemic health outcomes by facilitating the early recognition and management of inflammatory and chronic disease processes.