A reinforcement learning-based optimal control approach for managing an elective surgery backlog after pandemic...

Area of Science:

• Operations Research
• Healthcare Management
• Artificial Intelligence

Background:

• Pandemics like COVID-19 lead to significant delays in non-emergent elective surgeries, creating substantial backlogs.
• Hospitals face challenges in restoring surgical services due to limited resources and the need for patient safety.

Purpose of the Study:

• To develop an operational solution for clearing elective surgery backlogs post-pandemic.
• To provide timely surgical care for patients using limited healthcare resources.

Main Methods:

• Modeling the surgery backlog recovery as a discrete-time queueing network system using a Markov decision process.
• Proposing a scheduling optimization algorithm based on piecewise decaying epsilon-greedy reinforcement learning for dynamic daily scheduling.
• Considering factors like new patient arrivals, waiting times, and clinical urgency in the scheduling algorithm.

Main Results:

• The proposed method effectively and rapidly clears pandemic-induced surgery backlogs.
• Ensures all patients receive timely surgical care, outperforming current policies.
• Demonstrated efficacy through simulations and implementation in a Chinese hospital.

Conclusions:

• The developed stochastic control process-based method offers an effective strategy for managing surgery backlogs.
• Encourages wider adoption for surgery scheduling during public health crises.
• Facilitates safe restoration of surgical activity and timely patient care.