Heart failure is a devastating disease. Because of it, patients are too tired to do things they used to enjoy. Their health is always getting worse, requiring more and more hospital stays. Cardiac Resynchronization Therapy (CRT) can help some patients with advanced heart failure live better and longer. Patients must receive an implantable CRT device. Having a tool that alerts doctors to patients with CRT devices at increased risk of dying would be really helpful. In that way, aggressive therapy can be offered to patients that need it most. A team from Hungary may have just come up with such a tool, using machine learning: the SEMMELWEIS-CRT score.
The team behind the SEMMELWEIS-CRT Score
Machine learning researchers from the Hungarian Semmelweis University founded the Argus Cognitive. This spin-off company offers “Analytics-as-a-Service” solutions to medical institutions that want to use data to improve the quality and efficiency of treatment. While their previous work focused on autism diagnosis, their current flagship is the SEMMELWEIS-CRT Score.
The SEMMELWEIS-CRT Score
The Semmelweis team used the data from 1500 patients with fitted CRTs. In the end, they decided upon 33 parameters that appeared to have potential. Those most strongly linked to death were the age at CRT implantation, kidney function and certain features found on the ECG. In these 1500 patients, the SEMMELWEIS-CRT score predicted death better than any of the other currently available tools.
The future of the Score
The creators of the score are confident that it can be easily embedded in electronic health record solutions of cardiology wards. In this way, it can automatically alert doctors about heart failure patients that are at risk of dying, earlier than ever before. Of course, the Argus Cognitive has only tested the SEMMELWEIS-CRT in patients from a single center, so it is not entirely certain that it could work equally well in populations from other hospitals. Nevertheless, it seems that AI plays an ever important role in medical prognosis. As a result, conventional statistics may soon become obsolete.