A new artificial intelligence algorithm can predict critical diseases, including the need for ventilator treatment, among hospitalised patients as early as several days before the condition develops. The technology could come to play a crucial role in beginning the correct treatment at an earlier stage and ensuring efficient utilisation of the hospital's resources.

2020.04.15 | Mette Vestergaard Rasmussen

Artificial intelligence can spot potentially critically ill patients at the time of admission. This is shown in a research project from Horsens Regional Hospital. Photo: Rune Borre-Jensen, Central Denmark Region.

According to the Danish Health Authority's risk assessment, approximately 60,000 Danes will become so poorly due to COVID-19, that they will contact the Danish healthcare service for help. Further patients will be admitted to hospital with other critical diseases during the same period.

But which of the hospitalised patients are at risk of becoming so poorly that they require intensive treatment – perhaps with the help of a ventilator?

The new artificial intelligence algorithm +Prio is ready to provide the answer to this question, and the technology can therefore be an effective tool for Denmark’s emergency and medical departments in the early identification and prioritisation of particularly vulnerable patient groups.

"The algorithm is designed to support the doctor treating the patient by providing vital information about the individual patient's risk of developing critical diseases. With the technology, we can take much earlier and better targeted action to ensure that the patient gets the right treatment up to several days before the condition actually develops," says Jeppe Lange, consultant, associate professor and academic coordinator at the Department of Clinical Medicine at Aarhus University and Horsens Regional Hospital.

"The algorithm will even be able to tell the doctor who is responsible for treating the patient why it’s making its 'prognosis' for the individual hospitalisation. This makes it possible to direct treatment and care specifically at the cause," says Jeppe Lange.

The algorithm is trained on health data

The group behind the new technology comprises researchers from the Department of Clinical Medicine at Aarhus University and Horsens Regional Hospital, project employees from MTIC, and IT engineers from Enversion A/S. The project is part of an industrial PhD supported by the Innovation Fund Denmark.
The project group has already trained the algorithm on a large volume of historical health data from the Horsens cluster during the research project TVÆRSPOR. This consists of patient data from municipalities, general practitioners and hospitals, with all data anonymised.

At the moment, the project group is working to test the algorithm in clinical contexts at the hospitals. 

"We’re very much looking forward to transferring from research to real world at some point, especially now that technological solutions are needed more than ever to support the important work of the healthcare sector," says Jacob Høy Berthelsen, Healthcare Director at Enversion.

The preliminary research results are under publication and are expected to be finally published in a peer-reviewed medical journal during 2020.

Read more about the research project in Nature Communications.

Contact

Academic Coordinator, Associate Professor, Consultant Jeppe Lange

Department of Clinical Medicine, Horsens Regional Hospital

Mobile: (+45) 2032 3986.