October 29, 2025

There are situations in which patients are no longer able to express how they would like to be treated - in the case of dementia or cardiac arrest. How can the patient's wishes still be taken into account as much as possible; which therapies would you agree to and which not?

Ideally, a living will should be available in such cases. If this is not the case, relatives or legal representatives must determine the patient's presumed wishes. And if this is not possible in an acute emergency, it is up to the doctors treating the patient to act in a way that is in the patient's best interests.

A team from Switzerland and Germany has now investigated whether machine learning, i.e. artificial intelligence, could also help to draw conclusions about the patient's wishes in certain situations. To this end, the algorithm was trained with the treatment preferences of almost 2,000 people over 50 in the event of cardiac arrest. Among other things, the assessments of the relatives of the study participants were then compared with the results of the AI. The result: the AI performed just as well or better than the actual relatives in this test setup.

In an interview in the „Sprechstunde“ section of Deutschlandfunk radio, Dr. Dr. Georg Starke provides insights into the study and an outlook on its practical application in everyday clinical practice; listen to it here here.