Evidence that Sunetra Gupta and her Oxford team were right: R0 for coronavirus may be 5.7

“High Contagiousness and Rapid Spread of Severe Acute Respiratory Syndrome Coronavirus 2” (preprint on cdc.gov):

Results show that the doubling time early in the epidemic in Wuhan was 2.3–3.3 days. Assuming a serial interval of 6–9 days, we calculated a median R0 value of 5.7 (95% CI 3.8–8.9).

In “Full range of coronaplague opinions” (March 25), I first noted that a team at Oxford led by Sunetra Gupta conjectured that as many as half of Britons were already infected with coronavirus due to the high R0. If the conjecture were true, it would invalidate the recommendations of the Imperial College study that led to a Western rich-country shutdown (except for Sweden). (No point in shutting down a society/economy to reduce transmission of a disease if the disease has already reached most people.)

When the Imperial College study leader upped his estimate of R0 to “just over three”, the scale of the deaths in the UK dropped by a factor of 10 (“Update on the coronapanic fueled by Imperial College”).

Can we believe this spectacular R0 number? It is lower than measles (12-18 without vaccinations). But such a high value seems inconsistent with the success of various Asian countries that have suppressed COVID-19.

Note that the authors of the new paper are huge fans of government and public health interventions:

Our results suggest that a combination of control measures, including early and active surveillance, quarantine, and especially strong social distancing efforts, are needed to slow down or stop the spread of the virus. If these measures are not implemented early and strongly, the virus has the potential to spread rapidly and infect a large fraction of the population, overwhelming healthcare systems.

But if they are right about R0, it is unclear how these could work in the U.S. once the economy and society are reopened. Consider a single infected traveler to an uninfected region of the U.S. A week after a visit to a business convention or a theatrical performance there could be more than 300 cases (45 infected at the first event, as with the choir practice in Washington; doubling to 90 after 2.5 days; doubling to 180 after 5 days; doubling to 360 after 7.5 days).

Also, if they’re right about R0, it is way too late for Sweden to do anything to change its trajectory. Our media are desperate for Swedes to die (today: “Sweden’s Relaxed Approach to the Coronavirus Could Already Be Backfiring” (TIME)), but since they didn’t shut down weeks ago the virus has to be pretty much everywhere by now. There wouldn’t be any point to Sweden falling in line with the sober non-science-denying governments.

[Note that Sweden’s new case rate is lower than Massachusetts and the death toll is similar on a population-adjusted basis (10 million for Sweden; 7 million for Massachusetts). The current University of Washington deadpool forecast is 5,625 for MA and just 4,182 for Sweden (i.e., shut-down MA will see substantially more deaths per 1,000 residents than open-for-school-and-work Sweden). Somehow the deaths of thousands of Massachusetts residents after dramatic government school and business closures is not interesting. (I’m keeping the numbers up to date in this older post.)]

Why can’t we do a test of the general population? “Coronavirus may have reached Colorado as early as January” (Colorado Sun) If R0 is, in fact, 5.7, shouldn’t everyone in Colorado who is going to develop symptoms already have developed them? (The U Washington deadpool says that Colorado peaked yesterday for hospital utilization.)

Doing a super crude calculation, in which the doubling time stays constant, we take the base 2 log of Colorado’s population of 5.6 million. From a single infected person showing up, it takes just over 22 doublings for everyone to be infected. If the doubling time is 2.5 days, that’s 55 days (i.e., if the first infected person arrived in mid-January and the doubling time lengthened gradually (due to a lot of people already being infected), the 100-percent infected time would be right around now).

Or maybe a single R0 number is not that useful. R0 is high in China and U.S. urban mass transit environments such as New York and Boston, but much lower in sprawling car-dependent environments (e.g., Denver)?

20 thoughts on “Evidence that Sunetra Gupta and her Oxford team were right: R0 for coronavirus may be 5.7

  1. “Why can’t we do a test of the general population?” Could you imagine the horror if they tested everyone and we all were already over it? How would they push their fear and tyranny?
    Side note, something that is never going away- One way aisles/habitrail lines, in supermarkets and stores. Making every shopper go down every aisle is a dream come true.

  2. High R0 means:

    1) isolation and social distancing are USELESS. (They only reduce reproduction rate by some percentage… maybe 50%. For low transmissibility infections that could be enough to reduce R below 1, thus extinguishing the epidemic; otherwise it’s simply delaying the acquisition of herd immunity and inflicting absolutely unnecessary economic damage).

    2) it implies that a very large part of population already had it, asympomatically. Which makes actual case fatality rate to be rather low, way below the common flu. What exactly is this scare about, then? Imaginary hobgoblins?

  3. Colorado School of Public Health predicts this summer for a hospitalization peak.
    http://www.ucdenver.edu/academics/colleges/PublicHealth/coronavirus/Pages/coronavirus.aspx (see “Colorado COVID-19 Modeling Report | April 6, 2020”)

    Pretty big difference from the UW models.

    I like your thinking on R0 and it not being just a single number. Looking at the state maps (https://canyon23.net/c19/) you can see in many Western states the cases/capita is pretty small since there are less people and they are spread out over huge areas.

    • Thanks, Paul. Do we reject “science” by believing one over the other? U. Washington has a nicer web site for its prophecy. If the native Colorado forecast is correct, everyone who lives there should move to New York in May. Lots of herd immunity in the western world’s filthiest city and the NY hospitals will be empty by then.

  4. Coronavirus Study Finds Twice as Many Infections in Austria Than Earlier Thought

    “More twice as many people have been infected by the new coronavirus in Austria than official figures showed, according to a new survey, with a fatality rate of 0.77%.”

    “The study, conducted by polling firm SORA in cooperation with the government the Red Cross, tested a random, representative sample of 1,544 people aged 0 to 94 from across the country in their homes or in drive-in testing stations. It indicated that 28,500 people, or around 0.33% of Austria’s 8.9 million population, were infected with the virus by April 6, sharply higher than the 12,467 infections recorded by that date, with 220 people dying of Covid-19, the disease the virus causes.

    The findings suggest that while the death rate implied from the study, 0.77%, is lower than the World Health Organization’s estimate for reported cases, which is over 3%, it would still mean that the virus could kill many millions of people before a vaccine is available.”

    • Only 0.33% of Austrians are infected?!? That’s tough to square with the high reported R0 numbers, Austria being right next to plague-ridden Italy (confirmed community spread by the third week of February), and Austria not shutting down until March 15 (even assuming that the Swedes are wrong and these “soft shutdowns” have a huge effect (Austrians were still going to the grocery stores together, for example)). The only way that 0.33% makes sense to me (when combined with R0 of 3-6) is if a ton of Austrians already caught coronavirus, were asymptomatic, and recovered.

  5. As many of the above comments have suggested, R0 is not a fixed parameter of the virus, but strongly dependent on human behaviour. Many states have introduced contact tracing and social distancing, and then lockdown, which has dramatically reduced R0 over the duration of the epidemic in those locations.

    Similarly: Herd-immunity as a fraction of the population is not a fixed number, but equal to (1-1/R0). A high value of R0 requires a very high fraction for herd-immunity, while lower values of R0 will fizzle out at lower fractions of herd immunity.

    Trying to work backwards from the current number of deaths, to estimate the underlying infection rate, is a fools errand IMO. The few studies that have done better, used population sampling (such as Iceland) to determine the underlying infection rate. It seems clear that the true mortality rate amongst fit young adults is very low (< 0.5-1%).

    In theory the best strategy would be to isolate all of the vulnerable (elderly and already sick) while the epidemic rages, then gradually release them when the peak has passed. This presents massive logistical and ethical issues, and is unlikely to be tried outside of authoritarian states.

    • Did a lot of states actually start contact tracing early enough to make a difference? https://patch.com/massachusetts/boston/ma-launch-first-coronavirus-contact-tracing-program-u-s says that Massachusetts is the first and that, as of April 3, it was still in the future. The prophets of doom from U. Washington say that peak demand on hospitals will be April 11 (tomorrow). https://covid19.healthdata.org/united-states-of-america

      Isolating the vulnerable and letting the epidemic rage… that’s exactly what Sweden is doing!

    • When I said ‘states’, I meant countries (outside the US). Singapore was big on surveillance on contact tracing, though that no longer works in migrant worker dormitories (12 to a room!)

      The country where I live (South Africa) is fortunate to have no land borders with infection hot-zones (like Australia and New Zealand). Its very easy to stop and screen every person who came on a flight from Europe (before the lockdown), and do contact tracing and self-isolation from there.

      This tactic worked very well here, such that R0 was very low (below 2), and all new cases were ‘imported’ (at first). Since lockdown, and the spread of local transmission, that tactic is less effective. However the lockdown has kept numbers very low so far (less than 50 cases and 1 death per million population).

      What happens next is very hard to predict, since the general (poorer) population is not adhering to the lockdown regulations, and we have the worlds highest incidence of active TB and HIV immunocompromised population.

  6. Fauci said an antigen test is coming in about a week. That will quickly give us infection rates by locales, and should settle the debate (half will look like fools).

    So Phil, no need to keep trolling with ideological arguments disguised as amateur epidemiology.

    • Mr. Monkeycrap: While I wouldn’t want to be guilty of doubting Dr. Fauci, I also wouldn’t want to doubt the New York Times: “In the U.K., for example, the tests are plagued with false negatives (not picking up antibodies when they’re present) and with false positives (indicating antibodies when there are none). Some of the tests may not be specific enough to the new coronavirus; they may pick up a signal from antibodies made in response to infections with the coronaviruses that cause common colds.”


    • The cheaper antibody tests have a high rate of false results (both positive and negative). However, they do offer a first level sample screening of the general population, so that not every test has to be an expensive PCR one. Simple modelling should allow for correction of statistical data, though not for individual diafnostic purposes.

    • Gordon Richardson, according to this antibody testing versus PCR testing is not a matter of cost; they have different purposes (a past infection has been cleared versus someone currently hosting the virus).

  7. 50% infected in the UK is very far from either the current consensus or the experimental data.

    The tests have been completed for one city that was part of the epicenter of cases in Germany; they found 2% currently sick and 14% previously infected. There is no reason to expect similar or higher numbers in parts of the globe that are not considered to be COVID hotspots.


    Feel free to provide better data!

    • ilya: Thanks for this. The Yale doctor quoted in the article is famous for peripheral involvement in the Great Ivy League Halloween Costume Debate!

      How can we square this 16% number with the high R0 numbers that researchers have been using and with the fact that Germany didn’t shut down for quite a number of weeks after the “February carnival celebration” mentioned in the article?

    • @PhilG Can’t recall where I saw it but the woman from Oxford and the man from Imperial have a sordid history. It ain’t always science, sometimes it is personal.

Comments are closed.