UCD research may predict severity of Covid-19 from blood tests

22 Jul 2021

Image: © James Thew/Stock.adobe.com

By using machine learning to analyse routine blood tests, researchers were able to predict how ill a patient with Covid-19 is likely to become.

New research from University College Dublin (UCD) could help practitioners predict the severity of a patient’s Covid-19 illness and better allocate medical resources where they’re needed.

In a paper, published in Frontiers in Medicine, researchers described using blood tests and a machine learning algorithm to predict which Covid-19 patients were likely to suffer severe symptoms.

The work was a collaboration with scientists and clinicians from UCD and Dublin’s Mater Misericordiae hospital, as well as with data scientists from AI and analytics company SAS. The research was financed by Science Foundation Ireland’s Covid-19 Rapid Response Funding allocated to the UCD Conway Sphere research team.

“Rather than adopting a ‘wait-and-see’ approach, the information would allow clinicians to start earlier with more appropriate interventions for patients whose blood tests indicate they will develop severe infection,” said Dr Paulina Szklanna, manager of the UCD AI Healthcare Hub and lead author of the paper.

“Knowing upon admission whether a patient may or may not need the ICU makes a big difference in terms of patient management and allocation of hospital resources.”

‘The beauty of this study is that it uses very simple blood tests which are routinely done for patients across the globe’

There were three elements of interest in the blood of Covid-19 patients for researchers. These were: how many platelets were in the blood, how long it would take for the blood to clot, and the ratio between different immune cells in the body.

These could be indicators of severe Covid-19 infection.

Examining the interaction of these factors isn’t straight-forward, however. This is where the machine learning came in.

By using a cloud computing algorithm, researchers could input a blood sample into a computer and give a risk score back to a clinician through an app on their smart device.

This proof-of-concept study was carried out during the first wave of Covid-19 in 2020 and involved the bloodwork of 54 patients at the Mater hospital. Of these, 34 of the patients needed critical care support while 20 did not.

Because of the low numbers, the researchers highlighted the importance of validating their findings in a larger sample.

“The beauty of this study is that it uses very simple blood tests which are routinely done for patients across the globe. Individual centres and hospitals in different countries could take this data and basically validate it for their own cohort of patients,” said Szklanna.

Roderick Crawford, vice-president and country manager at SAS UK and Ireland, added: “We’re delighted that the team at UCD has been able to use our AI and advanced analytics technology via the cloud to generate these insights into the severity of Covid-19 for each patient.

“The ability to generate fast, accurate predictions in this way is vital to prioritise high-risk patients and resources they will require, especially where new variants mean the rate and type of infection is constantly changing.”

Sam Cox was a journalist at Silicon Republic covering sci-tech news