Data-Driven Strategies to Combat Missed Appointments:Machine Learning Prediction and Optimization of Healthcare Resources

Abstract

Missed medical appointments cause significant loss of time and resources of medical centers. Current systems mainly depend on automated or manual reminders to ensure attendance. In this scenario, introducing a machine learning model to predict no show of patients can significantly reduce missed appointments, increase revenue as well as allow for optimized scheduling for patients who need timely assessment. The main objective of this study was to predict patient no-shows using machine learning techniques for the hospitals to take data driven strategies. A dataset containing patient ID, appointment schedule, appointment date, disease type, and attendance status (show/no-show) was analyzed. We implemented and compared three modeling pipelines: a CatBoost‑based resampled ensemble, a stacked LightGBM ensemble, and an AutoML pipeline using AutoGluon. Our best-performing approach, AutoGluon, achieved an accuracy of 0.808423053 and a ROC AUC of 0.763069012, producing well‑calibrated probability estimates and outperforming traditional and ensemble baselines. Feature‑importance analysis revealed that age, SMS reminder receipt, and scheduling‑time features were among the strongest predictors of no‑show behavior. These findings demonstrate that AutoGluon can effectively identify patients at high risk of missing appointments, offering a practical tool for healthcare providers to proactively intervene and optimize scheduling, potentially reducing missed visits, improving resource allocation, and enhancing patient care continuity.

Ähnliche Arbeiten