Artificial intelligence recognizes who can die from COVID-19

Artificial intelligence recognizes who can die from COVID-19

Artificial intelligence has a 90% probability of determining who will die from SARS-CoV2 infection if they already become infected - and predicting who will need a respirator. According to the researchers, these results should be included in the scheduling of immunization queues.

Using medical data, almost 4,000 Scientists from the University of Copenhagen have learned an artificial intelligence system to quantify the risk of severity and death from infection responsible for the coronavirus pandemic to Danish COVID-19 patients.

In the case of people who have not even infected themselves yet, the program can with 90% with certainty to estimate the risk of death if the infection did occur. After admitting the patient to the hospital, he can also with 80% accurately predict who will need the respirator.

"We were working on models supporting hospitals when during the first wave there were concerns that ventilators would not be enough for patients in intensive care units" - says Prof. Mads Nielsen, co-author of a study published in the Scientific Reports magazine.

"Our results can also be used to precisely identify people in need of vaccination," the researcher said.

The program identified a number of factors that increased the risk of death and severe disease. In order of importance, these are body mass index (BMI), age, high blood pressure, male gender, neurological diseases, chronic obstructive pulmonary disease, asthma, diabetes, heart disease.

“Our results confirmed that age and BMI are the most important risk parameters for severe COVID-19. However, the likelihood of having to connect to a ventilator also increases when the patient is male, has high blood pressure or neurological diseases, ”says Prof. Nielsen.

According to the researchers, these results should be considered in determining the priority of vaccination. “We've found that when someone is affected by one or more of these parameters, it may make sense to move that person closer to the top of the line. This will prevent it from becoming infected and being treated under a respirator "- emphasizes the researcher.

Scientists hope that the program will soon work in Danish hospitals. They are also improving their system by using the medical data of another group of patients from the Danish Capital Region.

“We are working on anticipating the need for a respirator five days in advance. We do this by giving the computer access to the medical data of all people with COVID-19 in the region "- says Prof. Nielsen.

"The computer will never be able to replace medical expertise, but it will help doctors and hospitals take into account all infected at once and set appropriate priorities" - noted the specialist.

More information at  https://www.nature.com/articles/s41598-021-81844-x  (PAP)

Marek Matacz