Santeon project AI in the ICU enters new phase

Wednesday, June 5, 2024
News

The Santeon hospitals have received the green light from health insurers Zilveren Kruis and CZ to initiate the next phase of the project 'AI in the ICU'. This innovative project focuses on the use of artificial intelligence (AI) in intensive care (ICU) to make better use of scarce ICU capacity. After a successful first phase, the project is now being expanded to other Santeon hospitals. Work is also underway to expand the functionality of the system.

Santeon and the health insurers have been working with software supplier Pacmed for several years, which developed the Pacmed Critical program. This is decision-support software that should ensure that patients leave the ICU earlier and healthier. The predictive computer model helps Santeon intensivists make discharge decisions, after it had previously been used in research setting at the Amsterdam UMC. This not only improves ICU care, but also makes it more efficient.

In the coming years, both the demand for care and the staff shortage will only increase. It is therefore important to look for innovative solutions to keep healthcare available and functioning. AI can play an important role in this. Project leader Roald van Leeuwen of Santeon: “This project allows us to learn in practice how we can use AI in healthcare in a responsible and structural way. The health insurers are making IZA(*) transformation funds available to follow up on the experiences we gained in the first phase of this project at two Santeon hospitals.”

Predicting risk

Pacmed Critical helps ICU doctors determine whether a patient can leave the ICU. Based on machine learning, the program predicts for each patient how likely it is that the patient will be readmitted to the ICU within seven days or will die unexpectedly if the patient were to go to a nursing ward at that time. To this end, the model was fed with data from all patients who were previously in that ICU. Based on the factors and patterns that previously contributed to the chance of readmission, the model then predicts in real time what the chance of readmission or death is for the patients who are currently in the ICU.

Corstiaan den Uil, intensivist at the Maasstad Hospital, explains that the data that the system is fed with is combined with the real-time data of the patient: “Based on that data, the program makes a prediction for each patient. A human can never oversee all that data and the relationships between them. The system is therefore a valuable additional source of information for intensivists.” Intensivists combine the prediction with the medical knowledge they have and decide based on that. That decision always remains the responsibility of the doctor.

Expansion to other Santeon hospitals

Pacmed Critical has been used by the Maasstad Hospital and OLVG since last year. In this new phase of the project, AI in the ICU will be further scaled up and expanded within other Santeon hospitals. The Catharina Hospital and MST are now also going to use Pacmed Critical in the ICUs. They expect to be able to start working with the system in the fall of this year (2024).

In the meantime, the software at the Maasstad Hospital has already been updated based on the initial experiences. This update will be implemented at the OLVG this summer. Work is also underway to expand the software to support ICU care providers in other decisions.

More AI in the ICU

Pacmed Critical is not the only form of AI being used in the ICU. Gelre hospitals, SAS and KPMG have been working together since last year on better planning in intensive care using AI. This model makes it possible to predict the inflow of patients to the ICU, which enables better staffing. The smart AI tool was developed by the three parties during a global SAS Hackathon.

* The IZA, Integraal Zorg Akkoord (Integrated Care Agreeement) is an agreement between the Ministry of Health, Welfare and Sport and many parties in healthcare. The aim of the IZA is to keep healthcare good, accessible and affordable in the future.