AI tool predicts those most at terminal risk from Covid-19

Since the first wave of the coronavirus pandemic, researchers have been working to develop models that can predict how severely people will be affected by Covid-19, based on disease history and other health data. Accurate predictions about mortality risk could be used to help decide who should be prioritised for vaccination and how hospital resources should be distributed.

The university’s Department of Computer Science developed their model using patient data from the Capital Region of Denmark and Region Zealand, comprising health data from around 4,000 Danish Covid-19 patients. These data were used to train the model to recognise patterns in patients’ prior illnesses and in their experiences with Covid-19.

Factors known to increase risk of mortality from Covid-19 - such as body mass index, gender, age and high blood pressure - are heavily weighted in the model. In descending order of priority, other factors associated with higher mortality are: neurological diseases, COPD, asthma, diabetes and heart disease.

“Our results demonstrate, unsurprisingly, that age and BMI are the most decisive parameters for how severely a person will be affected by Covid-19,” explained Professor Mads Nielsen. “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.

“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.”

A Scientific Reports study suggested that the AI tool was capable of predicting with up to 90 per cent certainty whether an uninfected person would die of Covid-19 if they became infected. It can also predict with 80 per cent accuracy whether the person will need a respirator.

The model is 90 per cent accurate while the Covid-19 case fatality ratio is approximately 2-3 per cent worldwide (figures vary wildly between countries, depending on many variables). This means that there is a chance of producing false positives using the tool. However, this does not diminish the usefulness of the tool in helping to make clinical decisions, such as about which patients may need to be intubated.

“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 Nielsen. “Our new findings could also be used to carefully identify who needs a vaccine.”

The researchers are working with the Capital Region of Denmark to take advantage of their work and hope that the model could soon be used to help hospitals predict need for respirators in advance.

“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,” Nielsen added. “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.”