Online pandemic tracker launched to flag surges in COVID-19 cases

Ontario now has an "early warning system" for surges in COVID-19 hospital admissions, an online pandemic tracker that uses machine learning and real-time data to try to flag where and when the province's health-care system is becoming overwhelmed.

Using natural language processing, a type of machine learning, the tool scans province-wide hospital admission data for words and phrases likely to be used when physicians admit a patient with COVID-19: not just "COVID" but also language related to pneumonia, sepsis and asthma, syndromes the disease either creates or worsens.

The pandemic tracker does not show confirmed diagnoses. That's actually why it's useful: rather than relying on positive laboratory tests - which currently lag by several days, and take 24 hours in ideal circumstances - the tool offers an immediate look at whether a COVID-19 surge is starting, one Premier Doug Ford warned would come soon.

In fact, the pandemic tracker already shows that hospital admissions for COVID-19-related syndromes are above historical averages in Ontario.

"Today it's slightly higher than what we've expected, and to be honest, I am concerned," says Dr. Kieran Moore, the medical officer of health for Kingston, Frontenac and Lennox & Addington Public Health, the local health unit that developed the tool.

"We are going to be watching now to see if this number continues to take off."

Erie-St. Clair, the region that includes Windsor and Chatham-Kent, is "lit up," Moore notes. That region is seeing an even more significant spike than the province as a whole, across all seven hospitals that report data to the tracker.

While it's not clear why Erie-St. Clair might be seeing a COVID-19 surge - and it's important to remember the tool only suggests surges - it could be the impact of the border with Michigan, Moore says, which is still open for crossings for thousands of health-care workers who travel back and forth to the United States.

Moore and others in public health will be monitoring the pandemic tool, but the public can access it too.

"I've heard complaints that public health or the health system is not being transparent with their data. So we decided with our partners to make this accessible," he says.

"We think it will be a signal that will help everyone understand the impact of COVID-19."

The pandemic tracker relies on a system called Acute Care Enhanced Surveillance (ACES), which was developed after the 2003 SARS crisis to provide an early warning system for a pandemic. During SARS, breakdowns in communication between hospital emergency departments and public health units were one of the many information silos that fuelled the spread of the disease, Moore says.

ACES was created to send up a signal for an unusual cluster of hospital admissions, whether that was from another infectious disease outbreak, a bioterrorism event, or another public health crisis. Since then, ACES has been pressed into service for influenza surveillance, and is used every year to monitor flu flare-ups across the province.

In mid-February, the ACES flu mapping tool started to show a surge of patients going to emergency departments with influenza-like symptoms: a possible COVID-19 surge, since the illnesses share symptoms. But then the province's dedicated COVID-19 assessment centres began opening, and the visits to emergency for flu-like symptoms dropped.

Neither was a reliable signal, Moore says: emergency departments were awash with people worried they had COVID-19 and wanted a lab test, making it hard to spot a true regional spike, since hospitals everywhere were seeing the same influx. When assessment centres opened, it siphoned away the non-COVID patients as well as the true ones.

So Kingston public health unit team rebuilt the new pandemic tracker, which monitors all admissions - not just emergency department visits - and scans for a group of syndromes Moore believes is most predictive of actual COVID-19 patients, based on what hospitals in other countries hit by the pandemic have seen.

"We think it will correlate well with the surge and requirement of people needing to be admitted for COVID-19," says Moore.

"It should give everyone in the community, not just health system partners but directly to the public, an idea if there's increased activity in their community and the impacts of that increased disease activity on the health system."

Kate Allen is a Toronto-based reporter covering science and technology. Follow her on Twitter: @katecallen