Utilizing Greedy Algorithm and Deep Reinforcement Learning for Optimizing Surgical Schedules in Adaptive Scheduling

Operating room scheduling faces persistent challenges in healthcare facilities worldwide, with inefficiencies leading to resource wastage, extended patient waiting times, and staff burnout. This study addresses these challenges through three methodologies: greedy algorithm, deep reinforcement learning (DRL), and a novel hybrid model. Analysis of 35,000 surgical procedures revealed significant inefficiencies in current practices, including OR overutilization (463.87%), substantial waiting times (170.07 minutes), and frequent delays (58.39% of procedures). The hybrid model demonstrated superior performance, achieving a 34.2% reduction in OR utilization, 55.9% reduction in waiting times, and 87.5% improvement in on-time procedures compared to baseline. These improvements translated into significant clinical benefits, including reduced staff overtime (57.1%) and enhanced emergency case accommodation (17.6%). The hybrid model's resilience to operational disruptions and balanced performance across multiple dimensions provides compelling evidence for implementing adaptive scheduling methodologies in clinical practice, offering a comprehensive solution that balances efficiency, adaptability, and patient-centered care.