How artificial intelligence data mining can help us fight COVID-19
While we focus on vaccines, anti-virals and respirators in the fight against COVID-19, there’s another type of technology that gets less attention, but may be even more important in lessening the impact of the pandemic—information technology.
Given that we’ve been hearing more and more about the importance of widespread testing, we’re probably less surprised at this than we might have been a few weeks ago. It’s becoming increasingly clear that knowledge is actually one of the most important tools we have. The good news is that we have people working all the angles on this, from artificial intelligence data mining to genetic sequencing.
Why is this so important? It can get a little bit abstract, but the all-important testing strategy demonstrates exactly why knowledge is power.
Current evidence suggests that a certain number of people will get the novel coronavirus and be entirely asymptomatic. Some will never even know they had it. Meanwhile, others will get COVID-19 and have mild symptoms—not severe enough to go to the hospital. And, after they recover, they won’t know for sure what they had. Maybe it was just the flu.
Working under the assumption that people develop an immunity after they get it (likely, but not proven yet), these people, after the contagion period ended, would no longer be at risk of either getting or spreading it. That would leave them free to go back to work or, better yet, volunteer at a hospital.
Without testing, though, we have no way of knowing if they ever had it. And since, right now, we’re mainly limiting testing to people who have symptoms, we’re nowhere close to finding out who doesn’t have it at this point—a number that’s just as valuable as the number of active cases.
“Public health works best when we can get as precise as possible,” explains Steven Hoffman, director of the global strategy lab and professor of global health law and political science at York University. “So, as long as we’re in the current situation where we don’t have the full testing capacity, we’re stuck with more blunt tools like closing schools and asking everyone to remain at home except for essential trips outside. Those are effective for slowing the spread of the virus and buying us time but not, ultimately, for getting society back to normal.”
By now, most people understand that “buying time” through social distancing is important so that hospitals aren’t overwhelmed but, as Hoffman explains, it’s also about getting us to a point where we’ve beefed up our public health agencies so they’re ready to do mass-testing. The cuts to the Ontario Public Health budget under the Ford government means we started out at a disadvantage. And without mass “surveillance testing,” the official numbers we see reported are always out of date. It takes five days for symptoms to show up, then several more days for test results, so the confirmed case numbers we’re looking at could be 10 days post-transmission.
Fortunately, in other areas, we’re ahead of the curve, thanks to academic research into artificial intelligence that can mine data from social media and other sources to track outbreaks and hotspots before we get official confirmation of a positive test. Researchers at University of Guelph developed a system for doing this with avian flu and it’s now being applied to COVID-19. The idea, roughly, is to look for patterns and early signals, which includes everything from trade routes linked to disease transmission to people describing or complaining about symptoms on social media.
“So, we’re working with AI algorithms, the robots, and, so far, we have collected three million tweets related to COVID-19,” explains Rozita Dara, a professor in computer science at the University of Guelph. “Only two per cent of them have geo-location data but two per cent of three million is still a lot. So we are very excited to look at that and see if we can see patterns.”
In the United Kingdom, where mass surveillance testing hasn’t begun yet either, a team of doctors and scientists at King’s College London, Guys and St. Thomas hospitals have developed the COVID Symptom Tracker , an app that allows people to report their symptoms daily, whether they’re sick or not. From there, the researchers can aggregate and analyze data to identify clusters and researchers can better understand symptomology. For example, it’s still unclear if diarrhea and/or losing your sense of smell are early warning signs, but both are definitely on the radar. Apps like this might help determine how relevant they are.