Publications

Objective: To develop a machine learning prediction model for ambulatory appointment non-arrivals that can be deployed across multiple medical specialties. Methods: We analyzed 4.3 million ambulatory appointments from 1.2 million adult patients using the XGBoost machine learning algorithm. The model incorporated patient demographics, appointment history, provider information, weather data, and lead time. Results: The XGBoost model achieved the highest predictive performance (AUC 0.768). The most important features included rescheduled appointments, lead time, appointment provider, days since last appointment, and prior appointment status. The model calibrated well across all departments, especially for the operationally relevant 0-40% non-arrival probability range. Clinical Application: The model can be integrated into electronic health systems or dashboards to identify high-risk patients and reduce no-shows.