The best graphs and data for tracking the coronavirus pandemic

The novel coronavirus has now reached nearly every corner of the world, leaving billions under lockdown. The respiratory illness it causes, COVID-19, has a wide range of symptoms and a long incubation period; by the time most people realize they’re sick, they may have been spreading the virus for days. The virus can breed paranoia nearly as much as disease — and that’s left many people, including Verge staffers, anxiously looking for facts.

News outlets are good sources for specific growth trends and day-by-day increases, particularly when there’s local context for how a region is responding. But if you want to track the raw mathematical progress of the pandemic, you may need something more specific. There are plenty of raw figures to look at — tests executed, confirmed cases, hospitalizations, deaths — and each one can be tracked over time or against total population numbers. If you look at the numbers right, you can get a sense of how well a particular region is doing at containing an outbreak. But that takes the right graph and the right perspective on exactly what the numbers mean.

To that end, we’ve put together some of the most helpful public resources — and some factors to keep in mind when you check them.

General guidelines

As many experts have cautioned, even up-to-the-minute maps are effectively operating with a delay. Infected people can take up to 14 days to develop symptoms, and they might wait even longer for a test — if they’re tested at all. So it takes a while to see the effects of social distancing and quarantine measures.

“There is this literal lag in the data,” Virginia Tech biostatistics expert Ronald Fricker tells The Verge. “You always have to be thinking: what I’m observing today is a result of what we chose to do a couple of weeks ago. That’s a critical piece that I think a lot of people don’t quite get.” Cases will climb even after lockdowns have started taking effect, and deaths can spike well after a serious outbreak because patients may spend weeks in treatment.

Likewise, there are many undetected cases, particularly in the United States, where testing was catastrophically delayed. Early research suggests that a large portion of infected people show few or no symptoms, and few countries are testing comprehensively enough to find them. But with that in mind, the sites below can illuminate what we know about the novel coronavirus’s spread.

Johns Hopkins University’s Center for Systems Science and Engineering (CSSE) has assembled one of the simplest ways to track the virus worldwide. The map aggregates data from 17 sources, including the World Health Organization, the European Centre for Disease Prevention and Control, and several individual governments. The site tallies the total cases by country or hotspot, the number of deaths, and more optimistically, the number of people who have recovered.

If you want a general sense of the pandemic’s reach, this is a strong place to start. Just note that places sort data in varying ways, which can affect the visualization. America’s granular reports form a huge red mass on a zoomed-out map, for instance, while China also has many (albeit fewer) cases but features cleaner-looking single dots for each province.

Case numbers don’t tell the whole story, though. It’s also helpful to know how many people are being tested. Inside the United States, that’s where the COVID Tracking Project is helpful. The project is a volunteer effort led by The Atlantic’s Alexis Madrigal, presenting a straightforward state-by-state tally of tests. It collects data from the most trustworthy known sources (mostly state public health authorities), then it reports how many tests have come back positive or negative, how many people are hospitalized (if that data is available), and how many have died in each state.

Scattered and inadequate testing has hampered the US’s response, and many states are still bracing for COVID-19’s impact. So this tracker helps convey not just how many residents of a state are infected, but what percentage of residents are testing positive and how robust the testing program is. The site even grades how comprehensively states are reporting results. “I thought this was a nice simple way to try to get an idea of whether or not you could believe the numbers — that they’ve got them fully reported,” says Fricker.

You’ve probably heard the phrase “flatten the curve.” Basically, if we can slow the virus’s spread enough, we can keep infections at a manageable level so hospitals won’t be overwhelmed. How do we measure that curve? Fricker recommends the visualizations on 91-DIVOC, created by University of Illinois associate computer science professor Wade Fagen-Ulmschneider. The site pulls data from Stanford’s dashboard and plots it on a chart comparing different countries or US states, with each location’s timeline starting on the day it reported 100 cases. It marks not just the absolute growth of cases, but the rate at which cases are doubling, helping readers gauge whether an outbreak is slowing down.

91-DIVOC also helps demonstrate the value of logarithmic charts. Right now, a simple linear graph may show a seemingly exponential jump in COVID-19 cases — which accurately reflects how the virus is affecting people. Logarithmic scales represent an exponential curve as a straight line, meanwhile, so it’s easier to pick out subtler changes in the rate of infections. A small downturn might feel like cold comfort if you’re in a hotspot. But it’s also a sign that the virus isn’t unstoppable — especially if companies and governments help keep people safe and (whenever possible) at home.

Fricker also recommends the country-by-country breakdowns of Worldometer, a general reference site that aggregates case numbers. (It draws data from government agencies and reliable media reports, and it’s one of the sources for Johns Hopkins’ map.) Worldometer was apparently hacked in late March, and some faked Pakistan statistics caused temporary panic. The numbers were reverted, but it’s a good reminder to double-check any seriously alarming trends with other dashboards or news outlets.

For a more concrete look at why flattening the curve matters, the University of Washington’s Institute for Health Metrics and Evaluation (IHME) has broken down how COVID-19 may affect hospitals across the US.

IHME’s site models the growth of the virus across the next few weeks and months, projecting how many hospital beds and ventilators each state will need, compared to how many they actually have. Deborah Birx, the White House coronavirus response coordinator, has cited its data. The models are updated periodically, with explanations posted here.

Fricker points out that the IHME helpfully includes projection ranges as colored bands, not just a single line, emphasizing that there’s a lot of uncertainty in these models. He says its numbers seem “a bit optimistic.” For example, the site predicted around 80,000 American deaths when we spoke, while estimates more commonly predict between 100,000 and 200,000 deaths. “Maybe it’ll turn out to be true, but I think it’s on the very low end of what an epidemiologist or an expert in this area would say,” he says.

COVID Act Now presents a less rosy model. The site was created by a team including ex-Google employee Max Henderson, Alaska state legislator Jonathan Kreiss-Tomkins, and Stanford University medical scholar Nirav Shah. It models the growth of cases under several different scenarios, including a “shelter in place”-style scenario with poor compliance, a well-enforced “shelter-in-place” order, and a limited “social distancing” order. The state-by-state breakdown also notes which measures states are currently taking.

The site is updated every four days and isn’t supposed to be a detailed prediction. Fricker believes it also may be overestimating COVID-19’s spread, at least in the epicenter of New York. For planning purposes, though, he says it’s not a bad model of what states are facing: “If you’re Governor Cuomo, or you’re someone who’s trying to plan for this, I would err on the side of overestimation, not underestimation.”

For everyone else, it’s worth remembering that these charts need to account for feedback loops — ideally, officials will use them to decide on policies like shelter-in-place orders, which (again, ideally) change the infection rate and make the old model obsolete. Combined with the delay in detecting cases, that’s one more reason to treat models as guidelines rather than hard rules. “It’s so cool that these things are being disseminated, but some of them are very sophisticated and very complicated, and the average person is going to have a hard time being able to distinguish whether they should believe them or not,” says Fricker.