Overview: Hospital digital twins in the healthcare sector
Digital twins are rapidly being deployed across industries, and healthcare is no exception. They can be applied either on a micro level, for example, to help with patient treatment precision, or on a macro level to improve hospital operations.
A hospital digital twin mirrors the facility as a safe environment to test changes dynamically. In other words, you can test what-if scenarios risk-free and immediately see the impact on the hospital’s performance. It can inform your operational strategy, resource capacities, staffing, and care delivery.
Problem: How to improve hospital operations, patient experience, and resource allocation
Decision Lab is an expert in mathematical modeling, optimization, simulation, data science, and artificial intelligence. It uses these technologies to solve business problems. For one of the projects in the healthcare industry, the company was tasked to develop a simulation-based digital twin of two hospitals for the NHS Foundation Trust.
The NHS Foundation Trust is a unit within the National Health Service of England and Wales. It generally serves either a geographical area or a specialized function.
The goal of this joint project was to develop a digital twin that could help investigate what-if scenarios for potential improvement of hospitals’ operations, patient experience, and resource allocation. The key task was to simulate the whole journey of both non-elective (emergency) and elective (planned) patients, from their arrival to exit.
Solution: Building a simulation model as the backbone of a hospital digital twin
AnyLogic simulation software provides limitless customization because it supports the three main simulation modeling approaches individually as well as their combinations, which allows the developers to model real-world systems of any complexity and level of detail – from a patient-level to a hospital-level.
In this project, a discrete-event simulation approach was used to model processes in and between the facilities, and an agent-based method was applied to model patient behavior.
A simulation model Decision Lab created in AnyLogic served as a foundation for the future digital twin. It encompassed the operations of two hospitals – Cheltenham General Hospital (CGH) and Gloucestershire Royal Hospital (GRH), with a total of 75 units and buildings (locations).
In these hospitals, elective and non-elective patients have different behavior pathways. On entering a hospital, elective patients either go to their planned surgeries or sometimes have their surgeries cancelled. It can happen because there are no available beds, in which case the patients leave the facility and return on a different day. If surgery is performed, then a patient lands either in a specialty ward or first in a department of critical care and later in the specialty ward for recovery. Eventually, they also leave the hospital.
The pathway logic for non-elective patients is more complicated. A standard route for most patients is to go through nurse triage, where a nurse assesses the seriousness of a patient’s condition. Based on this assessment, the patient proceeds to other units to get the appropriate treatment, lands on a specialty ward until recovery and/or then leaves the hospital.
Additionally, the hospital model was highly customized and had several key attributes:
- Different medical issues and the corresponding treatment are grouped into specialties, such as cardiology, dermatology, etc., and each ward has a certain range of specialties and medical issues it can treat.
- If a patient is supposed to be placed in a ward that treats issues of a certain specialty but all beds there are occupied, they could be placed in another ward with a different specialty. For example, a cardiology patient could end up in a dermatology ward.
It takes longer to treat such a patient because of the lack of available specialists in another ward that could treat their medical issue. Minimizing the number of such patients would improve the overall patient experience. - Hospital beds have a queuing system – each patient has a score that helps with ranking. Patients that arrive from the Critical Care unit have the highest priority. The goal here is to reduce the patient queue time, which also impacts their experience.
The model’s interactive UI and statistics
The engineers from Decision Lab used the visualization capabilities of AnyLogic and gave the final simulation-based digital twin of the hospitals a user-friendly UI. Any manager could easily use it and gain insights about the hospitals’ performance.
In the model, there are overall results as well as the patients’ and the locations’ views. The management could see the performance results for each hospital, different departments, resources, and a whole queuing report as it was one of the key metrics to track.
In the locations’ view, the users could select four out of 75 possible hospital units and buildings, including acute care units, cardiac wards, and so on, to analyze detailed statistics for each of them. If a user wants to analyze an individual patient’s behavior, they can switch to the patients’ view tab.
Result: Hospital digital twin applications and future plans
The outputs of the simulation model were divided into three categories:
- Patients – amount of time each patient spent in every hospital location and as an outlier (staying in a ward of a different specialty).
- Wards – bed utilization and the number of such outlier patients.
- Emergency department – time for triage and trolley utilization in a certain area.
As they were exported as CSV files, the NHS Analytics team could visualize them for further analysis in other tools.
Now that the hospital digital twin was ready, the NHS management could use it for planning to reduce the queues and the length of stay, improve other patient experience metrics, or plan for building a new ward. They could also stress-test the hospitals’ capacity during the demand peak in winter, identify the impact of resource utilization on the overall performance, and more.
In the future, the development team can further extend the simulation model by adding more detailed staff statistics to optimize workforce allocation, cost statistics to identify the impact of managerial decisions, scenario comparison charts, and more.
The case study was presented by Peter Riley, of Decision Lab, at the AnyLogic Conference 2022.
The slides are available as a PDF.
