Dynamic multistage scheduling for patient-centered care plans.

Area of Science:

• Operations Research
• Health Systems Engineering
• Applied Mathematics

Background:

• Outpatient health programs face complex scheduling challenges due to patient variability.
• Customized care plans require dynamic scheduling to optimize resource allocation.
• Existing scheduling methods may not adequately address patient no-shows, reschedules, or ineligibility.

Purpose of the Study:

• To develop and evaluate an optimized scheduling model for multistage outpatient health programs.
• To assess the impact of customized care plans on system performance and clinic operations.
• To compare the proposed scheduling approach against human-led and deep neural network policies.

Main Methods:

• Formulation of the scheduling problem as a Markov decision process (MDP).
• Introduction of a linear approximation to the value function to manage large state spaces.
• Development of an approximate dynamic program (ADP) with dual variable aggregation for efficiency.
• Implementation of the ADP to identify optimal scheduling actions.

Main Results:

• The scheduling model demonstrates superior system performance for customized care plans compared to non-customized approaches.
• The ADP approach significantly improves clinic profitability and patient throughput.
• The model reduces practitioner idleness and shows robustness against errors in patient care plan assignment.
• Performance surpasses policies mimicking human schedulers and deep neural network-derived policies.

Conclusions:

• The proposed approximate dynamic programming approach offers an effective solution for complex outpatient scheduling.
• Customized care plans, when optimally scheduled, enhance operational efficiency and patient outcomes.
• This methodology provides a scalable and robust tool for improving healthcare delivery in dynamic environments.