Predicting Patient Arrival Rates in a Multi-Specialty Outpatient Department

Abstract

Although the medical interventions provided by Outpatient Departments (OPDs) are scheduled, the process of patient arrival is often unpredictable. This can lead to a variety of issues such as overcrowding, long waiting times, and patient dissatisfaction. Therefore, predicting the patient arrival rate is necessary to allow healthcare managers to be more proactive and make more efficient and realistic decisions. In this context, this real-life study aims to predict the hourly rate of patient arrival in a multi-specialty outpatient department using Machine Learning (ML) algorithms. The considered OPD regroups three specialties: pulmonology, allergology, and cardiology. One-year patient arrival data for all specialties were extracted and preprocessed. Thereafter, the exhaustive feature selection was applied to identify the best subset of features. Next, seven ML algorithms were investigated and assessed to determine the best predictive model. The conducted experimental study revealed that support vector regression shows the best performance for the prediction of pulmonary patient admissions, while random forest was identified as the best predictive model for the other specialties.