Amsterdam hospital to use AI to guide ICU discharges in world’s first
The Amsterdam UMC launched an experiment in which artificial intelligence will help intensivists decide when a patient can be discharged from the intensive care unit to a regular ward. A world first, according to the hospital. The goal is to use available ICU capacity as optimally as possible, especially in view of future Covid-19 waves.
Moving a patient out of ICU is a complicated decision intensivists face every day. If they move the patient too soon and their condition deteriorates to the point they need intensive care again, the risk of death is higher. It is also not good for a patient to stay in the ICU too long, intensivist Paul Elbers of Amsterdam UMC explained.
“It is annoying for the patient. But it is also very expensive, and the bed cannot be used for another patient who may need it much more,” Elbers said. “All very undesirable with the relatively low number of ICU beds in the Netherlands, a lack of ICU nurses, and high absenteeism since corona.”
Elbers and fellow intensivist Patric Thoral developed the AI with Amsterdam software developer Pacmed. The AI, called Pacmed Critical, compares the data of ICU patients with the pseudonymized data of nearly 25,000 patients recently treated in the ICU. Based on that, the AI can say in real time what the chances are that the patient will have to return to the ICU within a week if transferred to a nursing ward.
The discharge software also shows how that prediction changes over time and the list of factors on which the forecast is based. The intensivist makes the final decision and is, therefore, ultimately responsible. “We see it as a kind of second opinion. And by investigating how the software works in practice, we want to further increase the value for patients and healthcare providers,” intensivist Thoral said.
The goal is to reduce the number of readmissions and avoid treating patients in ICU for too long. “We estimate that gain can be made particularly with unplanned admissions, such as for blood poisoning, pneumonia, and a Covid-19 infection,” Thoral said. “With these types of admissions, the chance of readmission is generally greater than with ICU admissions for plannable care.”
Pacmed Critical is a world first, according to the intensivists. “As far as I know, it is the only medical decision support based on artificial intelligence that is used in this way in a hospital. I mean decision support at the bedside of the ICU patient, whereby algorithms and developed models can be used directly and in real-time with data from the electronic patient file,” Thoral said.
