Mining and Predicting No-Show Medical Appointments: Using Hybrid Sampling Technique

Alaidah, A.; Alamoudi, E.; Shalabi, D.; AlQahtani, M.; Alnamshan, H.; Abubacker, N. F.

Clinics use scheduling systems for patients’ appointments. However, no-shows are frequent in both general medical practices and specialties, and they can be quite costly and disruptive. This problem has become more severe because of COVID-19. The primary purpose of this study is to develop machine learning algorithms to predict if patients will keep their next appointment, which would help with rescheduling appointments. The main objective in addressing the no-show problem is to reduce the false negative rate (i.e., Type II error). That occurs when the model incorrectly predicts that the patients will show up for an appointment, but they do not. Moreover, the dataset encounters an imbalance issue, and this paper addresses that issue with a new and effective hybrid sampling method: ALL K-NN and adaptive synthetic (ADASYN) yield a 0% false negative rate through machine learning models. This paper also investigates the leading factors that affect the no-show rates for different specialties. The SHapley Additive exPlanation (SHAP) method reveals several patterns to identify the target feature (patient no-shows). It determined that a patient’s history of missed appointments was one of the leading indicators. It was also found that greater lead times between booking the appointment and the appointment date were associated with more no-show behavior. © 2021, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.