BedreFlyt BedreFlyt: A Digital Twin for Resource Planning in Healthcare

Duration
01.05.2024–31.12.2025

The proper handling of resources at a hospital is crucial to efficient operations, e.g., to determine how trained staff, bed availability in the hospital ward, and necessary rooms and equipment match the needs of different activities at the hospital, such as the treatment of patients. The dynamic allocation of these resources is necessary to efficiently manage the workflow and adjust it to avoid bottlenecks in operations, and to improve the prioritization and utilization of available resources.  By connecting simulation models to live data, the digital twin can ensure that the simulations more accurately reflect the actual resource allocation problems of the hospital. This way, a digital twin becomes a meeting point between static planning and dynamic optimization, allowing a better and more dynamic management of the workflow and its associated resources.

Image may contain: Font, Diagram, Screenshot.

 

The BedreFlyt digital twin aims to aid hospital staff with resource planning in a ward by solving the problem of room allocations for an incoming stream of patients. The complexity of this problem arises from the unknown, new patients arriving at every time step, creating a dynamic scheduling problem.

The BedreFlyt digital twin integrates several formal techniques into a tool chain for prescriptive analysis. Our digital twin combines formalized domain knowledge (using knowledge graphs) about patient treatments and hospital wards, an actor-based executable formal model (with ABS) to explore strategies for streams of incoming patients with associated treatments, and an optimizer (via Z3) to perform the actual bed bay allocation. The orchestration language SMOL connects the different components. The four key components of the architecture are: 

  1. a digital twin orchestrator, 
  2. a knowledge base and its interface, 
  3. a patient flow simulator, and
  4. an optimizer

Participants

Published Apr. 1, 2025 12:18 PM - Last modified Apr. 1, 2025 12:33 PM