A multi-appointment patient scheduling system with machine learning and optimization

Appointment scheduling is critical to increasing resource utilization and operational performance in various industry domains, especially healthcare. Costs to care for several serious diseases are projected to grow due to the aging population and rising drug prices. Thus, there is an urgent need for efficient operational planning and scheduling to reduce expenses. This research explores ways to effectively schedule outpatient chemotherapy visits with multiple appointments requiring different resources. The study aims to assess the impact of patient no-shows and individual stochastic appointment durations in scheduling performance and determine if overbooking is viable to mitigate the adverse effects of patient no-shows. The study first applies artificial neural networks (ANN) to calculate patient no-show probabilities and individualized appointment durations. Then, it builds several optimization models that use the ANN models' outcomes to schedule outpatient chemotherapy visits. The performance of patient schedules obtained from these optimization models is assessed using simulation analysis to identify the effectiveness of overbooking to combat patient no-shows and determine if individual stochastic appointment durations produce better key performance indicators.