Computer scientists tracking the deadly coronavirus epidemic have been working diligently to predict the virus’s next moves. The novel virus, which causes a respiratory illness dubbed COVID-19, has taken the lives of more than 2,100 people. It first emerged in December in the Chinese city of Wuhan, and has since infected more than 75,000 people, mostly in China. The numbers of new cases have begun to drop in China, but concern is growing over expanding outbreaks of COVID-19 in Singapore, Japan, South Korea, Hong Kong, and Thailand.
Alessandro Vespignani, a computer scientist at Northeastern University in Boston who has developed predictive models of the epidemic, spoke with IEEE Spectrum about computational efforts to thwart a global pandemic. His team has developed a tool, called EpiRisk, that estimates the probability that infected individuals will spread the disease to other areas of the world via travel. The tool also tracks the effectiveness of travel bans.
IEEE Spectrum: What’s the status of the COVID-19 epidemic now?
Alessandro Vespignani: In the last few days, we have seen some good signals in terms of reduced case reports from the ground in China. The fear now is the possibility of chains of transmission in other countries near China that we are not detecting. Places such as Hong Kong have instituted strong interventions like restricting transportation and closing schools. But other regions are not doing that. If these countries start seeing cases that were not imported from China, that could signal that the epidemic is spreading in other places. And it’s possible that these regions could become new epicenters.
Spectrum: We’re almost two months into the COVID-19 epidemic—what are disease modelers focusing on now?
Vespignani: Everything. We need an all-hands-on-deck approach. We need to understand what’s happening in China and the effect of the interventions there. We need to understand if there are signals that the epidemic is spreading elsewhere. We need to understand if there are new epicenters. There is no piece of this puzzle that is not important at this stage. That’s why we’re all stretched at the moment.
Spectrum: Tell us about EpiRisk and what it does.
Vespignani: Our modeling approach is to use all the possible data sources. At the moment, we’re focusing on the surveillance data coming from China and nearby countries. Social media and news sources are also on the table. First, we model the epidemic to achieve situational awareness, particularly outside China. Since we know how many people are traveling from regions hit by the epidemic in China, it’s possible to then infer the size of the epidemic elsewhere from the cases of COVID-19 detected internationally. We’re able to make these predictions ahead of on-the-ground reporting systems, since those often rely on official confirmation. Then we have the models look at how interventions like travel restrictions affect the transmission of the disease. Around Wuhan, airports are shut down, long-range transportation is shut down, schools are closed. We’re trying to understand the likelihood that these measures can contain the epidemic in China, and the likelihood of seeing cases outside China.
Spectrum: Does your model suggest the travel restrictions are working?
Vespignani: When the travel ban was issued in Wuhan, it was too late and it had little effect in China. The disease had already seeded in many other provinces. But the ban that the U.S. issued on travel to and from China is having a large effect on reducing the number of imported cases from China. That buys us some time to prepare while hoping that China will be able to contain the disease there. Unfortunately a travel restriction of this level cannot be kept indefinitely because of its huge economic cost. And if the epidemic spills into other countries, at that point travel restrictions have little impact, because you can’t freeze the world.
Spectrum: You’ve been working long hours. What is it like to be a computational disease modeler during a deadly epidemic?
Vespignani: In our field there are two different kinds of work: peace time research when there are no health emergencies or threats, and then there is what we call war time, and that’s what we’re in now. Unfortunately, when there are emergencies like this COVID-19 epidemic, we have to work with limited data, a constantly changing landscape, and a lot of assumptions. Perhaps what you produced the day before has to be completely revised because a new piece of information has arrived. So it’s quite hectic. But the fighters of the battle are the doctors, nurses, and public health people in the field. They are the ones risking their lives. What we do as computer scientists and computational epidemiologists is provide them with intelligence to anticipate the move of the enemy.
Spectrum: How many teams of scientists are working on modeling COVID-19?
Vespignani: We do global conference calls with research teams and agencies, and I’ve seen at least 80 to 100 teams on some of these calls. These teams bring all kinds of expertise. There are people doing phylogenetic analysis and mobility modeling, and there are people forecasting, now-casting, and doing long-term projection analysis. We have much more collaboration and communication between teams and agencies than in previous epidemics.
Spectrum: Should some groups start consolidating their work?
Vespignani: No. Think about weather or hurricane forecasting. If all the models point in the same direction and results are consistent, you can trust the outcome. It’s the same with disease modeling. No model is exactly the same as another model. Each could use a different population map, or different assumptions on the relative susceptibility of certain age brackets. There are so many questions and ways of tackling those questions that you want a have a portfolio of models.
Spectrum: National and global agencies like the World Health Organization seem to be relying more and more on computational disease modeling. How did those relationships come about?
Vespignani: For instance, in the U.S. there was an initiative started by the CDC [U.S. Centers for Disease Control and Prevention] a few years ago called FluSight, among others, and they started including the modeling research teams to help forecast seasonal flu. These kinds of initiatives created a community, and the community has stayed in contact with each other and the agency. Other agencies around the world have been a part of similar initiatives and have built communities that way.
Spectrum: How can Spectrum readers and computer scientists in other fields help?
Vespignani: There may be issues such as computing and algorithm development—things that computer scientists can do without reinventing themselves as computational epidemiologists. Start by teaming up with people already in the field of infectious disease modeling. That will help you avoid those initial common mistakes and bring you to the center of the fight right away.
Emily Waltz is a contributing editor at Spectrum covering the intersection of technology and the human body. Her favorite topics include electrical stimulation of the nervous system, wearable sensors, and tiny medical robots that dive deep into the human body. She has been writing for Spectrum since 2012, and for the Nature journals since 2005. Emily has a master's degree from Columbia University Graduate School of Journalism and an undergraduate degree from Vanderbilt University. She aims to say something true and useful in every story she writes. Contact her via @EmWaltz on Twitter or through her website.