Self-optimized learning algorithm for multi-specialty multi-stage elective surgery scheduling

As populations age and living standards improve, the demand for elective surgery has risen significantly. Yet, economic and policy constraints slow the expansion of medical resources, requiring hospitals to manage multi-specialty resources collectively. This makes modelling and scheduling elective surgeries complex. This study proposes an integrated scheduling model for multi-specialty and multi-stage elective surgeries. We use novel evaluation indicators to measure medical workload, waiting times and service continuity from hospital and patient perspectives, ensuring dimension consistency in the objective function. To address scheduling complexity, we design a self-optimized learning algorithm combining reinforcement learning (RL), genetic algorithm (GA), and heuristic rules, that is, a dual-encoding hybrid genetic algorithm based on Q-learning and scheduling rules (DQGA). Our results show that the hybrid algorithm achieves a 64.99% improvement in solution accuracy compared to GA and 66.07% to Q-learning-based genetic algorithm (QGA). Experimental findings highlight that our scheduling model not only enhances resource utilization efficiency but also provides practical, scalable solutions for real-world hospital scheduling problems, where traditional methods often struggle with larger problems.