Since the start of the pandemic, various research groups have been working on prediction models using patient history and public health data to assess how badly someone could be affected by COVID-19 and to try and eliminate risk. Various risk factors have been identified that increase the chances of someone dying of COVID-19.  

Now new research from the University of Copenhagen has shown that artificial intelligence (AI) can help predict with 90 percent accuracy whether someone will die from COVID-19 before or after they get infected by assessing some of these risk factors.

Furthermore, the findings, published in the journal Nature, could help predict how many people may end up in hospitals and how many might need respirators, something which could help alleviate pressures on healthcare systems.

“We began working on the models to assist hospitals, as during the first wave, they feared that they did not have enough respirators for intensive care patients," said Professor Mads Nielsen of the University of Copenhagen’s Department of Computer Science in a statement. "Our new findings could also be used to carefully identify who needs a vaccine.”

The machine learning (ML) model developed in the study is based on health data from 3,944 Danish COVID-19 patients collected from the United Kingdom Biobank. The model took various risk factors into account and the computer AI then used the data to identify patterns and correlations with prior illness and the patients' bout with COVID-19, which was then extrapolated.  

The findings suggested that it was possible to predict hospital and Intensive care Unit (ICU) admissions using only a limited number of variables, age, gender, and body mass index (BMI). From these, the ML model could predict mortality from COVID-19 with a 90.2 percent accuracy. 

“Our results demonstrate, unsurprisingly, that age and BMI are the most decisive parameters for how severely a person will be affected by COVID-19. But the likelihood of dying or ending up on a respirator is also heightened if you are male, have high blood pressure or a neurological disease,” said Professor Nielsen. “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.” 

It is worth noting that the study did have several limitations. Firstly, there was only a limited number of patients analyzed. A larger sample size may have produced different results, especially the limited number of ICU patients that they had assessed.

Secondly, the researchers also selected a subset of variables to assess in the model. If they had included other variables the results might have been different. Lastly, the researchers also described in their paper that the changing criteria for SARS-CoV-2 testing may have impacted their results. 

Nevertheless, even with some of the limitations of the study, the model could still be used to help and identify patients that are most at risk and may serve as a potential tool in clinical settings in the future.  

“We are working towards a goal that we should be able to predict the need for respirators five days ahead by giving the computer access to health data on all COVID positives in the region,” said Prof. Nielsen. “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.”

A larger and preferably multinational cohort should be used in future ML prediction models, the researchers conclude.