PICUリソース管理の最適化：キャパシティとフローモデリングのためのデータ駆動型離散イベントシミュレーションアプローチ

There is a critical need for advanced modeling tools that can provide data-driven insights into optimizing resource utilization and patient flow in hospitals. We developed and evaluated a flexible, data-driven discrete event simulation (DES) model to optimize capacity utilization and patient flow through a multiunit hospital system, particularly focusing on the PICU and its downstream units.

Retrospective discrete-event simulation modeling and validation study.

Quaternary referral hospital.

All patients admitted to Boston Children's Hospital between January 2012 and February 2025.

None.

The model was validated against a real-world PICU expansion (from 30 to 48 beds) using historical data (83,315 encounters pre-expansion, 4,108 post-expansion). Simulating PICU expansion, without considering limitations in capacity of downstream units to which patients were transferred, resulted in a predicted average PICU length of stay (LOS) of 4.56 days and capacity utilization of 65.6%. These figures significantly differed from the actual observed post-expansion LOS of 5.82 (p = 0.004; 95% predictive interval, 4.38-4.81 d) and utilization of 82.6% (p = 0.004; 95% predictive interval, 62.7-69.6%). However, when the model incorporated actual downstream unit capacities, the predicted post-expansion PICU LOS was 5.27 days (p = 0.083; 95% predictive interval, 4.85-5.90 d) and utilization was 75.6% (p = 0.124; 95% predictive interval, 69.4-84.5% d), closely aligning with observed outcomes and highlighting the critical impact of downstream bottlenecks. Using synthetic data simulations, we have further demonstrated the model's utility for capacity planning and optimizing new service line scheduling.

The proposed open-source DES model effectively simulates patient flow across multiple hospital units, offering a powerful tool to administrators for optimizing hospital operations and resource allocation. The model handles complexity, is flexible, and considers interdependencies between units, enhancing decision-making capabilities. Our model is readily transferrable to other healthcare systems and is easily adaptable to a variety of scenarios.