Predicting Hospital No-Shows Using Machine Learning

All over the globe, significant amounts of patients miss their appointments without cancelling in time or even cancelling at all, resulting in billions of dollars wasted yearly due to increased idle time, overtime and waiting time that the other patients and hospitals face. Hospitals are actively trying to implement methods to try to reduce the idle time caused by patient no-shows by using overbooking and reminder systems. However, these two methods can be very costly. Overbooking can lead to patient dissatisfaction and constant personalized reminders, such as phone calls, to every patient can be annoying and costly in terms of manpower. This paper focuses on offering a solution which mitigates the global phenomenon of medical no-shows by creating a machine learning model using existing patient datasets to discover patterns and relationships between multiple patient variables and their tendency to miss appointments. Therefore, the likelihood of a patient showing up, given their information, may be predicted. The machine learning model used to form the solution predictive model is based on the decision tree classification algorithm. Furthermore, a scheduling system was implemented such that the overall model detects whether a patient has a risk of missing an appointment with a 95% accuracy, upon which it automatically enables the risky patient's schedule slot for overbooking and notifies medical staff or administration to contact them accordingly.