Dutch OLVG hospital starts using AI in the ICU

23 May 2023
The smart software that works with artificial intelligence has been used for a while in a long trial period. Due tot proven success during the trial, OLVG decided for permanent use of the AI-driven discharge software. The Maasstad hospital, with whom there has been intensive collaboration for this innovation, will also implement the smart AI-software in its ICU in a few weeks. The five other Santeon hospitals will follow later.

Decision support software

Artificial intelligence is increasingly being used in hospitals for diagnosis, research, treatment and now also to assist with discharge decisions in the ICU. Digitization and data-driven technology can make an essential contribution to more effective care and are therefore a spearhead in Dutch policy through the Integral Care Agreement. Specialists agree that AI, if used smartly, can help to reduce the the ever-increasing workload that healthcare is confronted with. Examples include supporting radiologists in diagnosing cancer based on images from scans. Or smart triage software that can help doctors with intake choices. The use of artificial intelligence in the discharge procedure in an ICU is a new example of how AI can support doctors with the process of decision making. In practice, this means making a decision whether or not a patient can be safely transferred from the ICU to a nursing ward.

World First in ICU

Doctors from Amsterdam UMC had a world first by using this software in a research setting. They are supported in their discharge decision in the ICU by artificial intelligence, based on large amounts of information from the electronic patient file. This allows intensive care treatments to be used as optimally as possible. The research phase is now a thing of the past and the OLVG and the Maasstad Hospital in Rotterdam are the first of seven Santeon hospitals who are licenced to use the AI-driven discharge software in practice.

Discard software for ICU

The software is an assistant that uses a smart model to predict the likelihood that a patient will have to be readmitted to the ICU or, worse, pass away after discharge. For this purpose, the model was fed with data from all patients who were previously admitted tot that ICU. The software then looks for factors and patterns that contribute 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. According to co-founder Wouter Kroese of Pacmed, software supplier, doctors should be aware that the AI tool does not provide a conclusive answer to the question of whether or not someone can be discharged from the ICU. “That decision still lies with the doctor. Our model only predicts the likelyhood that a patient has to be readmitted, or dies, shortly after discharge. But there's also data that our model doesn't know that doctors do see. For example, the color in the patient's cheeks” he says.