How Many People Will Get Sick From the Coronavirus? Epidemiologists Model Answers
Scientists quietly see a best-case scenario of tens of thousands of deaths; 10 million over two years is the worst case
The coronavirus has so far infected more than 100,000 people around the world and killed almost 3,500 as of Friday. The question that scientists are scrambling to figure out is how far and fast it will spread and how deadly it could become.
Hundreds of teams of epidemiologists, mathematicians and statisticians are scouring for data, dialing into conference calls and communicating on Slack channels to share information about the disease. They are pumping reams of data into computer models to refine predictions.
The answers they generate will enable governments to better formulate policies to slow it down and let hospitals prepare for who might be coming through their doors—and when.
Most forecasters are reluctant to predict—at least publicly—how this will play out over months or even years. Amesh Adalja, a senior scholar at Johns Hopkins Center for Health Security, said the outbreak is still in the first inning. It isn’t yet clear how many people have the disease, how quickly it is spreading or even how deadly it is.
Ashleigh Tuite, an epidemiologist at University of Toronto, has been working on a model of the spread of the disease in China that stretches out through mid-March.
She said there were too many moving pieces to make predictions beyond that. “Forecasting is a really tricky game, I don’t want to put a number on it,” she said.
But quietly, forecasters are refining their predictions about the possible spectrum of what could be coming, and several agreed on a range. The coronavirus could kill as few as tens of thousands of people, said Dr. Narges Dorratoltaj, senior scientist at AIR, which provides forecasting for governments, corporations and the insurance industry.
Or, as Jeffrey Shaman sees it, the virus could kill as many as five million to 10 million in two years. The professor of environmental health sciences at Columbia University and director of the Climate and Health Program called that a worst-case scenario if all of the efforts to slow the disease failed.
The spread of the disease, as well as the fatality rate, are challenging to calculate, because there are so many variables. It isn’t clear how many people who contract the virus are actually seeking treatment. The World Health Organization has stated the death rate is 3.4% globally. U.S. officials have said it is likely much lower than that.
“The risk in the United States as a whole is still low,” said Anthony Fauci, head of the National Institute of Allergy and Infectious Diseases.
Forecasters have to wrestle with under-reporting, much as with the flu season. If you come down with the flu but don’t go to the doctor, you will likely not show up in an official tally as infected.
To get a better estimate of who is sick, scientists look for alternative signals like finding out how many people are Googling “flu,” or opening the Wikipedia page for flu, or buying medication for fever or calling in sick to work.
“All these behaviors leave a pattern,” said Roni Rosenfeld, head of the Machine Learning Department at Carnegie Mellon University who has focused on forecasting epidemics.
The next step is to estimate how rapidly the virus is spreading and then to plug that information into models that already exist. At the University of Texas, Austin, Dr. Lauren Ancel Meyers is retrofitting a model that was already under development for the federal Centers for Disease Control and Prevention for pandemic flu preparedness.
The model contains detailed demographic breakdowns of populations in cities across the country, as well as behavioral information such as commuting patterns. That helps scientists better understand how a disease might move around.
One Toronto-based health surveillance company uses airline flight information to predict disease vectors.
Eventually, scientists will try to measure how changes in behavior will impact the spread of the virus. Will people stop shaking hands? Will schools close? Will governments impose a quarantine? Will airlines cancel flights?
These responses can flatten the rate at which the virus spreads, giving hospitals more time to prepare and limits overcrowding.
As the models become populated with data, scientists run simulations of how the disease could spread through the entire U.S. population, using high-performance computers.
“Because there are so many uncertainties and variables, I’m running thousands of millions of variables across different possible parameters,” Dr. Meyers said.