Share this story!
Researchers at the University of Gothenburg have developed AI-based decision support that will help doctors determine which treatment is most effective for each patient. The researchers used data from tens of thousands of patient cases to teach the AI how different factors influenced past outcomes.
The researchers have developed two different kinds of decision support, one of which indicates whether a patient's case is similar to other cases where the patient survived or died 30 days after their cardiac arrest.
"The decision support is one of several puzzle pieces in the big puzzle that is the doctor's overall assessment of a patient. We have a variety of factors to consider when deciding whether to proceed with cardiopulmonary resuscitation. It is a very tough treatment that we only will give to patients who will benefit from it and who have the opportunity to live a subjectively valuable life after their time in the hospital", says Fredrik Hessulf, researcher and doctor and the one who developed the decision support.
The second decision support will identify which patients are at risk of another cardiac arrest and show which factors are important for long-term survival.
"The accuracy of this decision support is reasonably good. It can predict with approximately 70 percent certainty whether or not the patient died or had another cardiac arrest within a year. Like Fredrik's decision support, this one also has a strength in that a few factors can predict the outcome almost as well as the model with several hundred variables," says research doctor Gustaf Hellsén at Sahlgrenska University Hospital, who developed this decision support.
So far, the decision support is at the research stage, but the researchers continue to develop the AI and doctors who want to test the tool can do so via the project's website.
"My colleagues and I who have tested the tool see great potential in its use in our clinical everyday life. The answer from the decision support can strengthen the doctor's opinion because a calculation is made based on the knowledge of tens of thousands of patients. It can potentially help us not subject patients to painful care that will not benefit the patient, while at the same time saving healthcare resources," says Araz Rawshani, researcher ST physician and the one leading the research group.