Scientists from the College of Copenhagen have developed an AI device that may predict who’ll die from COVID-19 with as much as 90% accuracy.
The researchers fed the system well being knowledge from nearly 4,000 COVID-19 sufferers in Denmark to coach it to seek out patterns of their medical histories.
The device decided whether or not an uninfected one who later catches the virus will die from the illness with as much as 90% certainty.
[Read: How much does it cost to buy, own, and run an EV? It’s not as much as you think]
It additionally predicted whether or not somebody who’s admitted to hospital with COVID-19 will want a respirator with 80% accuracy.
Professor Mads Nielsen of the College of Copenhagen said the device might assist establish who to prioritize for vaccines and what number of respirators a hospital will want:
We’re working in direction of a purpose that we must always have the ability to predict the necessity for respirators 5 days forward by giving the pc entry to well being knowledge on all COVID positives within the area. The pc won’t ever have the ability to change a health care provider’s evaluation, however it will possibly assist docs and hospitals see many COVID-19 contaminated sufferers directly and set ongoing priorities.
The system additionally analyzed which well being components would doubtless lead a COVID-19 affected person to die or want a respirator.
Unsurprisingly, BMI and age have been essentially the most decisive indicators. However the examine additionally confirmed that males and folks with hypertension or a neurological illness had an elevated danger.
The subsequent most influential well being components have been having continual obstructive pulmonary illness (COPD), bronchial asthma, diabetes, and coronary heart illness.
“For these affected by a number of of those parameters, now we have discovered that it might make sense to maneuver them up within the vaccine queue, to keep away from any danger of them changing into contaminated and finally ending up on a respirator,” mentioned Nielsen.
You may learn the examine paper within the journal Nature.
Printed February 9, 2021 — 13:04 UTC
