• AI reveals long COVID symptoms through Twitter analysis of nearly half a million first-hand tweets 

• Over 100 symptoms identified. The most frequent symptoms include fatigue, pain, brain fog, anxiety, and headache. Patterns of co-occurrences also emerged

• Results can help inform and expedite researchers’ and clinicians’ long-hauler prognosis and treatment 

• Framework has the potential to help with other first-hand disease symptom gathering on social media

Toronto, February 11, 2022 – Today, the Vector Institute, along with industry collaborators Roche Canada, Deloitte, and TELUS released a study that can help shed some light on the frequency of symptoms and subsequent treatment of people who experience long COVID. Knowing the crucial role social media can play in public health monitoring, Vector, working with its industry collaborators, created a framework to analyze over 460,000 Twitter posts using machine learning. The framework helped to identify the most notable symptoms expressed by individuals exhibiting long COVID such as fatigue, brain fog, headache, pain, and anxiety. It also recognized patterns of co-occurrences of multiple symptoms.  

Long COVID isn’t well-understood. Approximately half of those who’ve had COVID-19 have reported at least one lingering symptom three months after infection. There is also no single way to diagnose the condition nor is there a single treatment. Vector hopes to bring researchers and clinicians one step closer to helping treat long COVID by providing missing information on its symptoms.

Research has shown that social media is a commonly used outlet to express experience with illness. Users rely on social networks for information as well as emotional and social support. This makes social media a rich resource for researchers, but only if they can reliably filter the sea of daily posts and pinpoint those with relevant language.   

Working with Roche Canada, Deloitte, and TELUS Vector applied natural language processing (NLP) techniques to over 460,000 Twitter posts by people who had self-reported long COVID to see if any patterns or signals would arise amidst the social media noise.

The hope was that these patterns could reveal clues from daily life outside of the hospital setting about when and how frequently symptoms arise and where clusters of the condition occur. Insights could be shared with clinicians and researchers to hone their questions, identify trends early, or inform treatment strategies. 

After conducting a series of experiments, preliminary results showed that patterns related to symptom frequencies, co-occurrence, and distribution over time could be successfully detected and visualized using the framework. Social media analysis confirmed a number of recurring symptoms and patterns of co-occurrence. The next step is to work with researchers and clinicians. 

“This study is a tangible example of how AI can help improve lives. It provides the medical community with early signals and first-hand insight into long COVID symptoms that can help inform further investigation and treatment.” said Cameron Schuler, Vector’s Chief Commercialization Officer & Vice President, Industry Innovation. “It also demonstrates how AI specialists, corporate and health sectors can work together and share their expertise to help address some of the most worrisome outcomes of COVID-19.” 

“Long COVID can manifest itself in many different ways. We are still learning about the heterogeneity of this condition and how to treat symptoms and there is no one size fits all approach. The findings from the Vector’s social media study help strengthen our understanding of long COVID, especially the more bothersome symptoms. This proof of concept study can be applied to many areas in medicine and can be used to inform health care providers in many ways – from figuring out early signals in a pandemic to early signals of side effects of new drug therapy,” said Dr. Angela Cheung, Senior physician-scientist, TGHRI, Schroeder Arthritis Institute, at University Health Network, Professor of Medicine at University of Toronto and the co-lead of CANCOV, the Canadian COVID-19 Prospective Cohort Study. 

Indeed, the value of this machine learning framework doesn’t start and end with COVID. Researchers now have a tool and process that can make social media a key resource for understanding other population-level health events, such as new emerging infectious diseases, rare diseases, or the effects of booster shots on infection. 

The collaborative Covid Long Haul project, run by Vector, is ongoing, and a short technical paper has been accepted for presentation at the 6th International Workshop on Health Intelligence at the AAAI 2022 Conference. A full project report with outcomes will be released in early 2022. 

Vector recognizes the valuable machine learning and clinical expertise contributions of sponsor partners Roche Canada, Deloitte, and TELUS including original project ideation; review of clinical literature; data collection modelling; engineering of visualizations; and interpretation of results.

About the Vector Institute

The Vector Institute is an independent, not-for-profit corporation dedicated to advancing artificial intelligence, excelling in machine learning and deep learning. Our vision is to drive excellence and leadership in Canada’s knowledge, creation, and use of AI to foster economic growth and improve the lives of Canadians. The Vector Institute is funded by the Province of Ontario, the Government of Canada through CIFAR Pan-Canadian AI Strategy, and industry sponsors across Canada.

For more information please contact:

media@vectorinstitute.ai 

Orli.Namian@Vectorinstitute.ai