Leveraging machine learning to create user-friendly models to mitigate appointment failure at dental school clinics

Area of Science:

• Machine learning applications in healthcare
• Predictive modeling for patient no-shows
• Dental informatics and electronic health records

Background:

• Patient no-show appointments pose significant challenges in dental school clinics (DSCs).
• Accurate prediction of no-shows is crucial for optimizing clinic operations and educational outcomes.
• Electronic health records (EHRs) contain valuable data for developing predictive models.

Purpose of the Study:

• To develop an efficient machine learning (ML) model using DSC EHR data to identify patients with a high likelihood of no-show appointments.
• To create a user-friendly system providing a risk score for clinicians and administrators.
• To identify key factors contributing to patient no-show appointments through ML modeling.

Main Methods:

• Eight ML algorithms were evaluated using de-identified DSC EHR data.
• Model performance was assessed using metrics like F1 score, AUC, precision, and recall.
• Key factors influencing no-shows were identified through ML analysis.

Main Results:

• The Bagging classifier model demonstrated the best performance with an F1 score of 0.41 and AUC of 0.76 at a probability threshold of 0.20-0.25.
• Strong correlations were found between no-shows and appointment confirmation, patient age, prior visits, failed appointments, lead time, and medical alerts.
• The developed ML model effectively identified high-risk patients.

Conclusions:

• Implementation of the user-friendly ML model can enhance DSC workflow efficiency.
• The model supports improved dental student learning outcomes.
• Optimized patient care and resource allocation are potential benefits.