Are you at risk of developing long COVID? This computer program may be able to tell

Using machine learning, researchers sifted through hundreds of characteristics to predict who may develop long COVID.

Chris Arnold 4 minute read May 20, 2022
Personnel evaluation by artificial intelligence,blue background,vector illustration

Long COVID, or post COVID-19 condition, is estimated to affect one in ten patients who previously came down with the virus. GETTY

Scientists may have found a way to identify who is susceptible to long COVID, thanks to machine learning and artificial intelligence. 

Long COVID, or post COVID-19 condition, is estimated to affect one in ten patients who previously came down with the virus. The most common symptoms include fatigue, memory problems, sleep disturbances, shortness of breath, anxiety and depression in the weeks and months after an illness, according to the Government of Canada, and can be severe enough to impact a patient’s ability to work.

Now, scientists at the University of North Carolina, say they have trained three machine-learning models to identify the most important features and risk factors for long COVID and identify a cohort of potential long-COVID patients. 

“These results speak to the powerful impact of real-world clinical data … to help better understand and find solutions for significant public health problems such as long COVID,” Joni Rutter, acting chair of NCATS, said. 

Shortness of breath, difficulty breathing, pre-existing diabetes, chronic kidney disease common traits of long-COVID patients

Researchers used machine learning to identify the common characteristics of COVID-19 patients who later reported to long-COVID clinics. 

“Characterizing, diagnosing, treating and caring for long-COVID patients has proven to be a challenge due to the list of characteristic symptoms continuously evolving over time,” Emily Pfaff, first author of the study, said in a statement. “We needed to gain a better understanding of the complexities of long-COVID, and for that it made sense to take advantage of modern data analysis tools and a unique big data resource like N3C, where many features of long COVID are represented.” 

Patient data like demographics, health-care visits, medical conditions and prescription drug orders from patients before and after they contracted COVID-19 was collected from the National COVID Cohort Collaborative (N3C). The N3C is a database containing information of more than 13 million Americans from 72 areas across the U.S., with nearly 5 million of those people having tested positive for COVID-19 since the pandemic began in March 2020.

From the N3C findings, researchers made a list of almost 600 patients who had been identified to have long COVID. The models were then tested on a cohort of 97,995 adults who had COVID-19 at least 90 days prior and who had visited a health-care site with a long-COVID treatment centre (indicating they had access to long-COVID care.) Researchers tracked to see who from this second list attended a long-COVID clinic.

The most important characteristics of long COVID clinic attendees were: shortness of breath, difficulty breathing, have pre-existing diabetes or chronic kidney disease. Being on or having taken Albuterol (a breathing medication eg. for asthma,) Metoprolol (medication for angina and high blood pressure,) and/or melatonin were also common characteristics of patients. Having the COVID vaccine or taken Dexamethasone seemed to influence against requiring treatment for long-COVID.

Researchers noted that their data disproportionally represents those patients who are more inclined to use (and perhaps trust) health-care clinics, hospital in-patients, as well as those who had more severe initial symptoms. 

More work on the computer models are required

The team at N3C plans to continue to refine the AI learning process with more data, meaning that more people down the line will be identified as potentially long COVID patients. 

Additional types of long COVID, different effects on certain people, will also be researched, as well as potential treatment options. 

“Depending on where the research leads, we may find that patients with different presentations of long COVID are different enough to warrant different treatments entirely,” Pfaff said. “So, it’s important for us to determine if long COVID is one disease, or a constellation of related conditions that are also related to having had acute COVID-19.”

The study was sponsored by the National Center for Advancing Translational Sciences, an organization that transforms scientific breakthroughs and discoveries into new treatments or cures. 

The researchers add that thanks to their efforts of identifying large swaths of patients quickly using AI, better methods of research will soon be available to study more COVID patients, and more treatment options. 

 

Chris Arnold is a Toronto-based writer.
Thank you for your support. If you liked this story, please send it to a friend. Every share counts.