Data on the infection rate is difficult to find. This post will present some data sources and show some visualizations of the infections. It’ll also provide some tips on how to avoid coronavirus.
This post will be updated frequently (hopefully daily).
First, let’s visualize the fraction of people that are infected in each country. Where are different countries in the coronavirus infection cycle?1
This is frightening on 2020-03-14.2 Not for me – I’d probably be asymptomatic if I contracted COVID-19 because I’m young and healthy. My primary concern is with the load on the US health care system. Case in point: what happens when grocery stores or FedEx distribution centers can’t be staffed because workers refuse to work because they’re fearful of getting sick? Food insecurity and supply chain/economic collapse.
It looks the health care system will be overrun on in early April on 2020-03-14. Hopefully the infection rate slows before that can happen.3 How can we aid the infection slowing down?
Slowing the rate of infection
Right now (2020-03-14) it seems that the only hope is social distancing. I think the timeline for a vaccine is too long (around 18 months), and cures/therapeutic drugs aren’t here yet (though they’re coming). Here are some resources to help with social distancing:
- CDC fact sheet on social distancing
- Forbes “What Is Social Distancing? Here Are 10 Ways To Keep The Coronavirus Away”
In short, avoid social contact as much as possible. The biggest tips are to work from home and avoid large crowds. Social distancing is why so many sporting events have been canceled and so many universities are canceling class.
The positive effect of immediate and extreme social distancing is best illustrated with the 1918 flu epidemic/the Spanish flu. The hardest hit city was Philadelphia. The infection hit the city very suddenly, meaning the health care system was quickly saturated and overrun. One effect of this: they had to stack bodies on the sidewalk because the morgues were overflowing (source).
However, St. Louis practiced immediate and extreme social distancing when the flu hit the city. Just 2 days after the flu hit the civilian population, Dr. Max Starkloff quickly ordered the closer of schools, movie theaters, sporting events and public gatherings.4
This meant the flu spread much more slowly throughout St. Louis than Philadelphia. Here’s a graph of the death rate for the two cities4
Functionally, the time taken to start social distancing is the only difference between these two curves. St. Louis responded about 14 days quicker than Philadelphia. That meant St. Louis had half the total death rate of Philadelphia, and they didn’t have to pile bodies on the streets because their morgues were too full.
The difference between these two cities is discussed more in these two articles:
- “St. Louis saw the deadly 1918 Spanish flu epidemic coming. Shutting down the city saved countless lives”.
- “This chart of the 1918 Spanish flu shows why social distancing works” (published on March 11th).
How fast is the virus spreading?
The direct visualization of this is with the number of new cases each day:
This perhaps isn’t the best visualization; it’s pretty noisy, and there’s no information about the retired population or the quality/size of the health care system. Here’s a different visualization:
It appears the number of confirmed cases is doubling every 2.4 days before the infection gets under control5 (as with South Korea and Italy). Details on the back-of-the-envelope calculation are in the appendix; briefly, I take the number of cases required to Italy’s health care system and scale by the population that’s 65 years or older and number of hospital beds.
This looks to be rather scary for the U.K. The Prime Minister Boris Johnson has largely kept the UK open. Schools and universities have been kept open far too long, and many businesses and citizens didn’t take coronavirus seriously (source). For example, on March 12th the government said it didn’t want to “disrupt daily life before it was absolutely necessary”.6 It looks like the UK will have more deaths and a higher economic cost.7
How fast is the virus spreading in the US?
How many infections does the US have? When will the number of people who need hospitalization overwhelm the number of hospital beds?
If social distancing does not work and no medicine is developed, it looks like this virus will overwhelm the health care system sometime in late March.
How fast is the US spreading in each state?
In most states, it appears the number of confirmed coronavirus cases is growing more slowly on 2020-03-31!
These tests are confounded by testing. On 2020-03-22, it looks like New York is rapidly expanding testing, or the virus is spreading faster. I believe they’re expanding testing – they recently opened a drive-thru testing facility (source), and NY has about 8× the number of cases as WA but only 1.2× the number of deaths (on 2020-03-22).
I interpret the increase in testing to be finding cases earlier in the infection cycle and/or discovering the infection in unconcerned populations (e.g., college students), especially because I doubt the virus can spread faster than doubling once every 2.4 days. The increased testing is likely masking the effect I’m concerned about, the load on the health care system.
A direct measure of the load on the health care system is the number of new people entering the hospital each day:
This is looks great on 2020-04-28. The number of hospitalizations has decreased by a 4× even though the number of cases has been (relatively) steady.
For Minnesota, let’s look at the number of people currently in the hospital and if they’re in the ICU or not:
How serious are people about social distancing?
Let’s study a weak proxy for this question, the number of restaurant reservations bookings through OpenTable:
Here’s a visualization of the subway usage in New York City:
The result of this: it looks like the virus is decreasing rapidly! Kinsa Health makes smart thermometers, and they’ve done some analysis to see the level of total infection in the US (not just COVID-19). For a better description, see “Can Smart Thermometers Track the Spread of the Coronavirus?”.
Kinsa’s fever dashboard looks promising:
This blog post is written as a projection for the very initial stages of the pandemic. The visualizations are geared towards the question “can the US health care system handle COVID-19?” The following links give some more information on other aspects, including questions like “how quickly will the pandemic be over?” and “how many people will the Wuhan virus infect?”
- “COVID-19 from a data scientists perspective.” This post looks at Iceland (which did some pretty heavy testing) and extrapolates to the entire world. Main conclusions: the Wuhan virus will infect 96% of the world and have a 0.06% mortality rate.
- “COVID Trends Dashboard.” This shows some metrics on how fast the virus is spreading, specifically the total number of cases/deaths versus the cases/deaths in the last week.
On March 12th, the U.K. goverment said that their approach…
include[d] asking people with even [a] very mild cough and flu symptoms to stay home, but stopped short of the measures seen in most other European countries and increasingly in North America, including closing schools and colleges.
The government said Thursday [March 12th] that it may have to impose these more stringent measures at some point in the future, but that doing so now would be premature, disrupting daily life before it was absolutely necessary, and risking the public would grow tired of complying with the restrictions and begin to ignore them just when the peak number of infections might be expected ….
—”Even while canceling mass gatherings, the U.K. is still aiming for deliberate ‘herd immunity’”. Fortune, March 14th.
They waited 6 days until March 18th before they “disrupted daily life”. It looks like they waited too long – 6 days means 5.6× more cases because the infection doubles every 2.4 days. That’d be bad even if they had a quality health care system; however, I estimate the U.K.’s health care system can handle 5× fewer patients than the US health care system.
My visualizations are geared towards to the US because I live there. If you’d like to customize these plots, visit the repository at stsievert/covid-19.
If you’re interested, I’ve published a page with all the images in this post without any text.
Visit the stsievert/covid-19 for all the data sources. Briefly, some of my main data sources include the following:
- Worldometers for country level infections
- nytimes/covid-19-data for state level infections
- OpenTable’s dataset for restaurant reservations
- The MTA’s public turnstile data for their data on NYC subway rides through Todd Schneider’s wonderful New York City Subway Usage Dashboard
- Wikipedia for misc. stats (populations, hospital beds, etc)
For full detail, please the repository at stsievert/covid-19.
(edit) It looks like it started slowing down around late March! Sometime between 2020-03-25 and 2020-03-29. ↩
“Public health interventions and epidemic intensity during the 1918 influenza pandemic” by Richard J. Hatchett, Carter E. Mecher, and Marc Lipsitch. 2007. https://doi.org/10.1073/pnas.0610941104 (the chart I show is a modification of fig. 1). ↩ ↩2
I can’t estimate the amount (on 2020-03-21). ↩