Researchers from Denmark developed artificial intelligence that can determine whether a person will die from COVID-19 with up to 90% accuracy, according to a new study, with body mass index, gender and blood pressure status contributing most to its nearly spot on forecasts.
The trained computer could also predict the number of coronavirus patients who will eventually need to be hospitalized, admitted into an intensive care unit or put on a ventilator with about 80% accuracy.
The researchers say artificial intelligence could help hospitals by continuously predicting the need for medical resources in real time; the more health data inserted into the system, the better the computer gets at making estimates, the study published Friday in the journal Scientific Reports says.
In the midst of its predictions, the computer could determine who should receive priority for vaccination against the disease, with the goal of preventing infections in high-risk groups. The team said it is also working to get the artificial intelligence to predict the need for ventilators five days ahead.
“The computer will never be able to replace a doctor’s assessment, but it can help doctors and hospitals see many COVID-19 infected patients at once and set ongoing priorities,” study co-author Mads Nielsen, a computer science professor at the University of Copenhagen in Denmark, said in a news release.
The team trained the computer by feeding it health data from 3,944 Danish coronavirus patients. Over time, it learned which patterns and correlations within medical histories and COVID-19 turnouts increased a person’s chance of dying from the disease.
Among the patients included in the study, 1,359 required hospitalization, 181 were admitted to an ICU and 324 died.
The likelihood of being severely affected by COVID-19, according to the artificial intelligence, increases the most with age and body mass index (BMI). However the chances of dying or ending up on a ventilator also spike if someone is male, has high blood pressure or a neurological disease, which correlates with what other studies have found on COVID-19 mortality trends.
Patients admitted to a hospital were more likely to have high blood pressure, diabetes, stroke, chronic obstructive pulmonary disease, asthma, heart disease, cancer or chronic kidney disease, among other comorbidities, than patients who didn’t need a hospital visit, according to the study.
They were also more likely to be smokers, male and older in age.
Knowing this, the computer could automatically tally who should be next in line for a COVID-19 vaccine, the researchers said.
“For those affected by one or more of these parameters, we have found that it may make sense to move them up in the vaccine queue, to avoid any risk of them becoming infected and eventually ending up on a respirator,” Nielsen said.
Katie Camero is a McClatchy National Real-Time reporter based in Miami focusing on science. She’s an alumna of Boston University and has reported for the Wall Street Journal, Science, and The Boston Globe.
