Adaptive queue management in healthcare using supervised Q-learning with time-varying reward and cost structures.

Area of Science:

• Operations Research
• Artificial Intelligence in Healthcare
• Queueing Theory

Background:

• Efficient hospital patient management is crucial.
• Time-varying demand and dynamic service environments pose challenges.
• Current models often lack adaptivity for complex healthcare systems.

Purpose of the Study:

• To develop a heterogeneous medical patient queueing model.
• To minimize overall patient waiting time using adaptive decision-making.
• To integrate reinforcement learning with stochastic queue dynamics.

Main Methods:

• A heterogeneous queueing model distinguishing first-time and returning patient service providers.
• Non-homogeneous Poisson process for patient arrivals.
• Q-learning framework with a supervised ε-greedy policy for operational actions.
• Separate Q-tables for arrival and departure events.

Main Results:

• Significant reduction in total patient waiting time.
• Substantial decrease in overall system costs compared to homogeneous models.
• Demonstrated effectiveness of adaptive, data-driven queue management.

Conclusions:

• The proposed model offers an efficient, data-driven approach to dynamic hospital queue management.
• Reinforcement learning effectively optimizes resource allocation under varying conditions.
• The framework is extensible for broader healthcare resource optimization.