Predicting COVID-19 impacts on Canada’s health-care system as provinces tackle restarting economies without spurring a resurgence of coronavirus cases is the goal of a federally funded artificial intelligence (AI) project launching at the University of Guelph.
From gauging needs for ventilators and other hospital equipment to predicting any likely new wave of coronavirus cases, the project aims to help decision-makers in government, business and health-care authorities plan for next steps.
Engineering professor Ed McBean said the project is unique in Canada for both its use of AI and its focus on predicting future impacts of the novel coronavirus on the country’s health-care system.
Currently, experts lack the ability to foresee potential changes in case loads and their effects on health care resources, he said.
Referring to existing coronavirus modelling studies based on current case numbers, McBean said, “A lot of that work is happening on ‘how are we doing today’ as opposed to predicting. Other modellers are mostly observing what happens as opposed to what may happen.”
The team aims to use various kinds of information in the models to develop forecasts of the likely course of infection or recovery from the virus under different scenarios, he said.
“We need forecasting models to improve understanding of how the environment and people’s behavioural and age patterns influence the escalation and control of case loads.”
Led by McBean and Prof. Andrew Gadsden, School of Engineering, the team has received a one-year, $50,000 award under the Alliance COVID-19 grant program of the Natural Sciences and Engineering Research Council. They will work with Adastra Corp., an information management and data science firm based in Markham, Ont.
The team plans to develop new AI models to collect and analyze data on regional COVID-19 infection levels as well as demographic and behavioural information such as age, lifestyle, health and travel history. Researchers will then use those models to predict likely impacts on the health-care system under both current quarantining measures and various provincial plans to ease lockdown measures in the coming weeks and months.
Under the project, undergraduate and graduate students will use AI to scour information from public health websites across Canada and create models for analyzing the data.
McBean said he expects the models will likely be available for use beginning in May or June.
He said health authorities might use the model to gauge whether practices in one province may be applied or adapted elsewhere, or to predict effects of any subsequent infection wave and attendant demand for infrastructure from ventilators to hospital beds.
“We want to see if there are distinguishable differences between regions and how governments in those regions could be better informed,” said Gadsden.
Governments may use the research to project health outcomes based on public compliance with physical distancing, mask-wearing and other preventive measures.
By predicting likely infection rates under different scenarios, the project may also help authorities plan for reopening businesses, said McBean. “This might help to say whether it’s a good idea. If we see an upturn in cases of infection, we can model how bad it is going to be.”
He said the accuracy of the team’s AI models will depend on the quality of data obtained from various health authorities.
“There are gaps in the data,” he said, noting authorities need to balance information access and patient confidentiality. “For AI modellers, the more data you give us, the better we can use the data and make predictions.”
Prof. Ed McBean