By Ian Gormely
October 5, 2021
Applying machine learning (ML) techniques to data from the Ontario Health Data Platform (OHDP), Vector Faculty Member Bo Wang helped determine a set of mortality predictors for long-term care (LTC) residents with COVID-19.
Wang and his team at UHN’s Peter Munk Cardiac Centre collaborated with researchers at ICES (formerly the Institute for Clinical Evaluative Sciences) to look at de-identified health data for more than 60,000 Ontario LTC residents who had been given COVID tests during the first and second waves of the pandemic (January to August of 2020). “We wanted to predict important risk factors for mortalities,” says Wang, who notes that this is the first instance in Canada of ML practices being applied to assess patient risk factors for COVID at the population level.
Their findings, outlined in the paper “Predictors of Mortality Among Long-Term Care Residents with SARS-CoV-2 Infection,” which was published in the Journal of the American Geriatrics Society, confirmed commonly reported factors including comorbidities and age.
But they also uncovered the role of functional status, a medical term for a patient’s level of physical, mental, and physiological activity which can be quickly assessed through a series of questions. “We identified that functional status was very important in outcome in long-term care homes after a positive COVID-19 test,” said Dr. Douglas Lee, co-principal investigator on this project. “These findings came to light because of the partnership between Dr. Wang’s group and our data analytics group at ICES”.
Wang was able to access the de-identified LTC patient data through ICES’ site on OHDP which was established by the Province of Ontario last year to give approved health researchers better access to data to better detect, plan, and respond to COVID-19. The platform, for which Vector provided strategic and research user input, was created in collaboration with a number of key health stakeholders including Compute Ontario, Ontario Health/Cancer Care Ontario, Schwartz Reisman Institute for Technology & Society, and Queen’s University. “This was one of the first papers to use ML techniques on OHDP,” says Wang. “The research would not have been possible without it.”
Long-term care residents in Ontario were hit especially hard by COVID-19, resulting in a disproportionate number of outbreaks and deaths. Wang’s work is among the first to look at the sector from a research perspective. “I think that COVID shed light on the importance of care for LTC residents,” says Wang. “This is a societal issue that calls for more attention and care in Ontario and beyond.”