“More research is needed before we can use AI tools in the clinic to predict bleeding risk, but this proof-of-concept study shows they hold great promise,” - Dr. Phil WellsTreating venous thrombosis, the third-deadliest cardiovascular disease, is a balancing act. Many patients take blood thinners for months or years to keep potentially deadly clots from forming in their legs, but these medications come with a risk of major bleeding. 

Current tools to estimate a person’s risk of bleeding can only consider information from their first visit, even though medication changes or weight gain reported in follow-up visits can have a major impact on their bleeding risk. That’s where AI can help. Researchers at The Ottawa Hospital and uOttawa trained four machine learning algorithms on eight years of visit data from 2,542 patients with venous thrombosis, 118 of whom had major bleeding. 

As reported in the Journal of Thrombosis and Haemostasis, they found the AI models that could use information from all the patient’s visits better predicted an individual’s bleeding risk than the six existing clinical models.

“More research is needed before we can use AI tools in the clinic to predict bleeding risk, but this proof-of-concept study shows they hold great promise,” said Dr. Phil Wells, senior scientist at The Ottawa Hospital and professor at the University of Ottawa.

Funding: This research was enabled in part by the Digital Research Alliance of Canada.

Authors: Soroush Shahryari Fard, Theodore J Perkins, Philip S Wells 

Core resources: The Ottawa Bioinformatics Core Facility

The Ottawa Hospital is a leading academic health, research and learning hospital proudly affiliated with the University of Ottawa and supported by The Ottawa Hospital Foundation.