Let’s ask again: Number of new COVID-19 cases worldwide is declining now?

On March 26, I asked “Number of new COVID-19 cases worldwide is declining now?”

I’ve been updating that post daily with numbers from WHO and Massachusetts. Neither sequence seems consistent with frightening exponential growth (maybe the worldwide data shows an exponent larger than 1, but not all that much larger). Massachusetts does seem to be experiencing a flat-to-down trend in new cases, though without the strong upward trend in testing that I would have expected.

Maybe this is because all of the world’s governments had the foresight to put everyone into “lockdown”? (Exchanging germs only when at essential jobs, when at essential shops (such as liquor and marijuana stores), when at the laundromat, when on a Tinder date, etc.) The WHO data don’t show a dramatic difference between otherwise comparable countries that have radically different rules (Sweden has fewer cases per capita than Denmark, for example, despite Sweden being open for business while Denmark is purportedly locked down. Norway has twice as many cases per capita than Sweden, albeit fewer deaths so maybe they are just testing more of the mild cases) . More importantly, the WHO data don’t seem to show coronavirus burning exponentially through any country, regardless of what the policy might be (even Brazil, which has chosen “immunity via sewage”).

(This is not to suggest that a declining number of cases will cheer anyone up. On March 17, we were looking at the potential for 2.2 million Americans to die from coronaplague (Business Insider), based on Imperial College forecasts. Now a forecast of 100,000 to 200,000 deaths is “grim”, “stark”, and “dire” (Associated Press). Nobody will say “That’s fewer than are killed by medical errors each year and we don’t bother taking any action to reduce those.” Nobody will say “We’ve killed way more than that with taxpayer-funded Medicaid-reimbursed opioids.”)

Readers: What do you think? Given the increased testing capability that countries have been building and the numbers from the WHO, it is possible that the number of new daily coronavirus cases is actually on the decline?


32 thoughts on “Let’s ask again: Number of new COVID-19 cases worldwide is declining now?

    • js: Interesting link. Thanks! The part I hadn’t considered was the following:

      As an outbreak progresses and countries have limited testing capacity, they tend to enact policies that aim to save test kits on “almost certain cases based on symptomatic evaluations” and to reserve them for testing, say, contacts of infected cases. France and Spain already officially enacted such policies.

      While this makes sense, the problem is that those “almost certain cases” are diagnosed and treated as if they had contracted the COVID-19 virus, but they are not included in the count of confirmed cases.

    • js: At the same time, the number of deaths should be reasonably accurate, right? (with the exception that some countries may be more or less likely to list an underlying condition as the cause rather than COVID-19) And those will lag by at least 2-3 weeks from infection. The March 28 WHO report listed 3,158 deaths. Today’s March 31 report lists 3,301. That’s not consistent with dramatic exponential growth a few weeks ago.

    • > At the same time, the number of deaths should be reasonably accurate

      According to John Lee they are not, at least in the UK; death counts are of people with the virus and not necessarily of people who died from the virus.

      I.e. when it comes to fatality rates, in addition to the denominator being understated (lack of random testing), the numerator is overstated (other causes of death conflated).

  1. Can’t believe anything China says. It was amusing that the world’s testing capacity was stagnant until the virus hit US. Only once in US, did innovation in testing begin. US was the only country which improved the speed & quantity of testing, leading to US showing the most cases.

  2. I don’t believe any of it. The fear porn of the media or the governments official statements. I don’t know the truth, I just know what isn’t the truth. On the other hand the social ramifications keep coming up roses, moms home with their kids, whole families out walking together, and my most recent observation- stores changing to old school hours. Supermarkets, etc. closing at 6 pm just like 40 years ago. As an agent of conservative change there hasn’t been anything close to corona chan in hundreds of years.

  3. The LA Times reports on a new Lancet article which makes the newest estimate of infection fatality rate and case fatality rate (i.e., fatality rate if you show symptoms, apparently).

    They also claim they had some way to account for Chicom lying (“censoring”) their death rates.


    Average death rate for infected person is 0.66%, average death rate for symptomatic person is 1.4%

    For the “its just like the flu , bro” crowd, in order to get the same death rate as a bad seasonal flu (80k) we’d need to limit infection to less than 12 million Americans.

    But actual death rate is skewed heavily towards 50 and older
    Fatality rate for those infected | those with COVID-19 symptoms
    80 and above: 7.80% | 13.40%
    70-79 4.28% | 8.61%
    60-69 1.93% | 3.99%
    50-59 0.60% | 1.25%
    40-49 0.16% | 0.30%
    30-39 0.08% | 0.15%
    20-29 0.03% | 0.06%
    10-19 0.01% | 0.01%
    0-9 0.00% | 0.00%

  4. “following the advice of Professor of Epidemiology Donald Trump?”

    AFP is reporting that Trump is finally taking the virus more seriously.

    President Donald Trump said Tuesday that the risk from coronavirus is emphatically worse than regular flu, reversing his previous statements.

    Trump told a White House press conference that “a lot of people” had previously suggested the country should simply let the coronavirus take its course, just like the seasonal flu.

    “Ride it out, don’t do anything, just ride it out and think of it as the flu,” they said, according to Trump, who said: “But it’s not the flu. It is vicious.”

    Better late than never!

    “the WHO data don’t seem to show coronavirus burning exponentially through any country, regardless of what the policy might be”

    Really? I’m glad Massachusetts is no longer showing exponential growth in confirmed cases, but if you look at new death counts for the US plotted on a logarithmic graph (Financial Times), it sure looks exponential.

    A week ago (March 25), when there were about 1000 US deaths so far, with only piecemeal social distancing measures, I predicted that US deaths would continue doubling every three days for three weeks, reaching 120,000 deaths by April 15. Three days later, it doubled to 2000 deaths – but three days after that, it’s still only at 3100 deaths, suggesting that social distancing measures in place around March 15 or 16 have been having an effect.

    How do we know that this is the result of social distancing, and not some other factor, like the Oxford scenario where a large proportion of the population have already been infected (and thus are now immune) without showing symptoms? The random testing in Iceland appears to rule out the Oxford scenario: only 6% of those tested are infected.

    Data from Kinsa (which sells Internet-connected thermometers and collects the data, allowing it to detect unusual numbers of high fevers a week in advance of hospital-based data) provides enough detail to suggest the effectiveness of particular social distancing measures. For example, in Florida:

    On March 12, a state of emergency had been declared [in Florida], but according to Kinsa’s data, fevers were continuing to rise. Closing local schools on March 16 had little effect.

    But on March 18, Miami’s bars and restaurants were closed, and within two days reports of fevers started to drop sharply, according to Kinsa’s data.

    So how can you handle a pandemic without shutting down the economy? South Korea demonstrates that it’s possible: South Korea keeps COVID-19 at bay without a total lockdown. Lockdown is a extremely blunt instrument. So far South Korea has been able to avoid it by testing early and often.

    That life in South Korea has not ground to a complete halt is mostly owing to an early and aggressive response to the COVID-19 pandemic. The country’s disease-control authority approved the first test kit for the virus in early February, less than a week after the application was filed. By the time the number of cases began to rise a couple of weeks later, it had the capacity to test thousands of people a day and get results within a few hours from a network of labs across the country.

    This was invaluable when an enormous cluster of cases emerged in the city of Daegu, centred around a messianic cult.

    • Russil: All interesting, but as noted above in my response to Henry, if the goal of “flattening the curve”/”social distancing” is to have the same number of infections spread out over a longer period of time, what practical difference does it make? Roughly the same number of people will die. Unlimited ICU/ventilator access does not guarantee that a COVID-19 patient will live. In fact, 86% of those on ventilators ultimately died (see https://www.thelancet.com/journals/lanres/article/PIIS2213-2600(20)30110-7/fulltext ). Maybe it would be a more peaceful death in a less crowded hospital, but the result is the same.

    • (I think that I have noticed a general trend. The people who were saying two or three weeks ago that flattening the curve was essential to avoid overwhelming hospitals are now saying that somehow social distancing will actually SAVE millions of lives. This is presumably because it will cut the percentage of humanity that gets infected, but nobody articulates a mechanism by which a temporary shutdown of society does more than slightly delay the inevitable spread.)

    • Phil,
      To answer your question, there are several ways that flattening the curve will improve outcomes:
      First and most important, there is an expectation that there will be medical interventions that are developed that will improve outcomes. Flattening the curve does not just slightly delay cases; it drastically delays cases (if flattening measures are maintained). There is a high likelihood that anti-viral, antibody seeding, or other treatments will be developed in the near term that will greatly improve mortality rates.
      Even without medical interventions, we can work to improve testing and contact tracing capacity.
      In addition, will have time to prepare medical resources including PPE for medical workers.

      I am suspect of your conclusion that ventilators don’t offer meaningful improvements of outcomes based on the article that you linked.

    • Marcus: I would love to see some sort of effective therapy! But it has been roughly three months, right? And we have nothing so far? If shutting down society for two months delays half the cases by two months, do we realistically expect significant breakthroughs to occur within the next month or two?

      Separately, if you don’t accept “86% will die” (Lancet, based on Wuhan), how about 66% of those who got “invasive ventilation, BPAP or CPAP via endotracheal tube, or tracheostomy, or extracorporeal respiratory support” (UK data; https://www.physiciansweekly.com/mortality-rate-of-covid-19-patients-on-ventilators/ )?

      “An article in The Guardian said this about the ICNARC study, ‘The high death rate raises questions about how effective critical care will be in saving the lives of people struck down by the disease.'”

    • Russell –

      1) actually that graph does not look exponential at all. On a log linear graph (which is what it is) an exponential function would be a straight line, those graphs look more like a linear function. The game of sliding the time scale around adds to the confusion. (lower left graph https://textimgs.s3.amazonaws.com/boundless-algebra/logarithmic-scales.svg#fixme )

      2) The random testing in Iceland does not provide any evidence for or against social distancing as causation.

    • @Alan That plot has the y axis on a log scale, just look at the numbers. So it does in fact show a line, which as you correctly note, is an exponential growth. For lots of cool visualisations and updated graphs (in Norwegian only), have a look here: https://www.vg.no/spesial/2020/corona/#verden . If you scroll down the number of cases to a section called “Vekst i dødsfall”, it shows the number of deaths per country as a function of number of days since the outbreak. On the right hand side there is a switch for log or linear scale, so you can experiment. In the log version they also show the lines which slope means a doubling of cases every number or days (hver dag = every day, hver 2. dag = every second day, and so on). You can also select to see individual countries, and further down the plots you have the deaths per country as a function of time. The US is, unfortunately, looking like a very nice exponential.

    • @Jarle –
      1) It is log/linear because one axis is log scaled and the other is linear scaled.

      2) Perhaps you have your own definition of a line. For me those labeled like “hver 3.dag” are lines. The others are curves with decreasing slope over time.

    • @Alan Yes, it is log linear because only the y axis is on a log scale. All the lines of doubling are in fact lines, the only difference is the slope. Perhaps you are confusing lines with the special line with unit slope (y=x)? Note that to be exponential, a function doesn’t have to be of the form f(x) = exp(x), or f(x) = 10^x, it can have a multiplication factor such as f(x) = exp(m * x) or f(x) = 10^(m * x), so that on a log linear plot it would look like a line of slope m (respectively for a natural logarithm or a base-10 logarithm).

      In the US case, you can see on the log plot that the increasing number of deaths looks very close to the linear slope of doubling every 3rd day. If you click on “lin” you will see a plot that looks like an exponential. If you choose China (Kina) you will see a curve that on the log plot looks like a quarter circle (seems to be reaching a limiting number), which is not compatible with exponential growth. Plugging into linear, you see that the curve as an S shape: characterised by a exponential growth in the beginning and a converging growth in the second phase.

    • Philip,
      I do think that medical interventions will be developed within the next several months. There are probably some very useful medical interventions that have already been developed and deployed; it just takes time for them to be tested, proven, recognized, and disseminated. As time goes on treatments and interventions will further be honed and improved.

      I think you are making several mistakes in your analysis:
      You are using cases instead of deaths to understand data about infection rates. Data on cases is entirely inconsistent between countries, states, timing, etc. In many places, the rate of infection might actually grow at a rate larger than testing capacity. You will get a much better understanding if you look at death rates and build in estimates for changes. In my own estimates, I assume death rates represent infections from 2-3 weeks ago and then assume that infection rates will be lower (10% or so) based on strong social distancing and warmer weather. If you look at death rates without any social distancing, you will almost always find consistent exponential growth.

      You seem to imply that we should accept the 100,000-200,000 deaths and just go about our business, but this estimate is based on strong and disruptive social distancing that continues on an ongoing basis. In the case that we eliminate social distancing measures, all expectations seem to be that the number of deaths will be much higher.

      Regarding your response on ventilators, I am reasonably convinced that ventilators don’t have a huge impact on the number of deaths (you have changed my mind); even making a marginal impact, I feel that they are likely worth the effort and expense. But ventilators are not the only reason to flatten the curve. For every patient that goes on a ventilator, there are far more than are admitted to ICU. My guess is that a large percentage of these are life threatening. If you ignore the number of ventilators, you would still run into capacity issues with ICU beds and associated medical personnel.

      If I were a doctor or nurse and knew that the public did not make an effort to flatten the curve in an effort to maximize my safety, I would definitely be inclined to avoid the risk and avoid working altogether. Especially with no hazard pay and substandard access to protective equipment.

      The basic things that society needs to survive are a very small percentage of GDP. Why not just sit this out for a while so that we can improve resources and better cope with the real problem.

    • Marcus: I don’t know enough to suggest a course of action, though I do think the Swedes are probably going to be shown to have done it smarter than the U.S. Isolate the old/at-risk, ban mass gatherings, keep everything else running (to what the Swedes have done, I would add: have people wear a mask when in crowded public spaces; cut tables/seating in restaurants by at least half). My neighbors went to the (Waltham, Massachusetts) supermarket at 6:15 am for the special “senior-only” time. The parking lot for the entire strip mall was full. The supermarket was packed. Nobody was wearing a mask. This is “flatten the curve” theater.

      Look at the list of what’s still open in New York, the plague center of the entire world. It includes DAY CARE!


      They brag like heroes on Facebook for staying in their apartments much of the time, but I don’t think the virus is shy about spreading in the 100+ kinds of venues that are still up and running to the unmasked. Imagine a day care where a nurse is a parent. The nurse gets it at the plague hospital and gives it to the child after coming home. The child gives it to 10 children and the daycare workers. The children give it to their respective parents. That could all happen in a few days. This is nothing like the lockdowns that the Chinese implemented that apparently actually were effective.

      There is a precedent for this. The Flagellants during the Black Death. https://blackdeathfacts.com/flagellants/ They had a religious belief that what they were doing was helpful. If you’d shown them statistics from a country that didn’t have Flagellants on the same trajectory of plague cases, it wouldn’t have shaken their faith that they were helping “flatten the curve” or whatever else it was that people said in 1348.

    • “They had a religious belief that what they were doing was helpful.”

      This is why I thought the Kinsa data was particularly interesting: it provides detailed geographic data, and because individual states and even counties are all doing different things at different times, you can see pretty quickly what measures are actually effective. (I’m glad to see that Florida finally has a stay-at-home order.)

      Because Sweden is using mitigation rather than suppression, that will also provide some evidence of whether mitigation would have been sufficient.

      “This is ‘flatten the curve’ theater.”

      Social distancing measures in the West may not be as draconian as in China, which moved all COVID-19 patients and their household members to quarantine wards, but they don’t have to be perfect to slow down the virus. Short of keeping health workers in some kind of dormitories isolated from their families, child care for the children of health workers seems unavoidable. https://www.washingtonpost.com/graphics/2020/world/corona-simulator/

    • Russil: I still haven’t seen any quantification of how much the curve has been flattened.

      https://www.nytimes.com/2020/03/30/health/coronavirus-restrictions-fevers.html says “The turning point began on March 16, the day schools were closed. Bars and restaurants were closed the next day, and a stay-at-home order took effect on March 20. By March 23, new fevers in Manhattan were below their March 1 levels.”

      So basically the epidemic is over, right? But the article itself says “By Tuesday, the hospitalization rate took four days to double. This is roughly what the fever readings predicted, said Nita Nehru, a company spokeswoman.”

      If new fevers in Manhattan have fallen so much, why are the hospitals overwhelmed? https://abc7ny.com/nyc-hospital-queens-coronavirus-news/6070475/ says things are “dire” and getting worse.

      Could it be that these thermometers are sold to rich white collar people who can easily work from home and self-isolate? If not, how do we explain that almost nobody is getting a new fever and yet New Yorkers continue to stream into the EDs?

      (Separately, the article continues to conflate “flattening the curve” (spreading out disease/death) with “reducing the infection rate”. It says “making it clear that social distancing is saving lives”. Is there any evidence that hospitals are helping a significant number of patients? There is no treatment for COVID-19, not matter how advanced the hospital.)

    • Russil: from today… https://www.cnn.com/2020/04/02/health/aerosol-coronavirus-spread-white-house-letter/index.html

      A prestigious scientific panel told the White House Wednesday night that research shows coronavirus can be spread not just by sneezes or coughs, but also just by talking, or possibly even just breathing.


      So I stand by my confusion about how American-style “lockdowns” actually slow down the disease by more than a couple of weeks. People are mixing enough to spread a disease that can be spread by breathing.

    • “If new fevers in Manhattan have fallen so much, why are the hospitals overwhelmed?”

      Again, median time from infection to first symptoms (when they show up in the Kinsa data) is five days. Median time from first symptoms to hospitalization is seven days. So if fevers in Manhattan are less frequent (around March 23), it takes seven days before new cases showing up at local hospitals will start to decline (should be around now). During that time, the number of new cases reflects what was happening two weeks previously (around March 18), when infections were growing exponentially.

      “So I stand by my confusion about how American-style ‘lockdowns’ actually slow down the disease by more than a couple of weeks. People are mixing enough to spread a disease that can be spread by breathing.”

      The less they mix, the longer it takes for the virus to spread. https://www.washingtonpost.com/graphics/2020/world/corona-simulator/

      One anecdote supporting aerosol transmission. There was a choir practice on March 10 in Washington state where nobody had symptoms, and people practiced social distancing and used hand sanitizer – but 45 out of the 60 people who attended have come down with COVID-19, and two have died.

  5. Most of the kills scored to Corona are from comorbidity and an excessive application of modern medicine. That more modern medicine will save us would be laughable if it wasn’t sad and dangerous. I suggest a public campaign of flatten the curve by staying home to die if you are sick. More people would live, and the healthcare system could get back to doing more profitable things like elective surgeries.

    • “Most of the kills scored to Corona are from comorbidity and an excessive application of modern medicine.”

      Interesting theory about modern medicine failure. How do we know if it’s true or not? Preferably with some numbers.

    • 250,000 deaths per year is about 700/day. COVID-19 deaths in the US just passed 1000/day. (#1 is heart disease, 1800/day, and #2 is cancer, 1700/day. In terms of daily deaths, COVID-19 should overtake them in a few days.)

    • Russil: Those are some ugly numbers. I have to admit that I didn’t think it would get this bad. After 3.5 years of hysterical doom-predicting headlines from the media I have learned to tune out hysterical doom-predicting news. Trump was going to turn the U.S. into Nazi Germany. The economy and stock market would collapse if Trump were elected (Krugman). Due to his stupidity and lack of sophistication, Trump was going to get us into a nuclear war with Iran/Russia/China/North Korea. Trump’s election was going to cause sea level to rise enough in the near term to wipe out all of our coastal cities.


      is on this subject. I think I’ll try to do a separate post on this in a couple of weeks when, I hope, the epidemic will have peaked.

    • “I have to admit that I didn’t think it would get this bad. After 3.5 years of hysterical doom-predicting headlines from the media I have learned to tune out hysterical doom-predicting news.”

      If you’re looking for a reasonably reliable, business/economics-focused news source, I’d recommend the Economist. The current issue has an article talking about cost-benefit analysis in the fight against coronavirus, as well as articles talking about what’s happening in the US, Sweden, and India. https://www.economist.com/leaders/2020/04/01/covid-19-presents-stark-choices-between-life-death-and-the-economy

  6. The time, effort, energy and quality put into reporting and researching this epidemic is insane. Even worse, it is focusing on the wrong area.

    We get 24/7 live update about the “numbers” and the “curve” but hardly *any* coverage or reports about educate the public how to not get sick: wash your hand often and good, stay home when you are sick, take care of your health, etc., etc..

    If the media and so called experts use 1/10 of their time to cover discussions and education on preventive measures, aired over and over and over, that would be far more effect, *even* if the message is misleading [1].

    It’s time to get the panic out of the epidemic.

    Btw, so much for “social distancing” in NYC [2].

    [1] https://www.youtube.com/watch?v=89od_W8lMtA
    [2] https://nypost.com/2020/03/30/crowds-ignore-social-distancing-rules-to-watch-usns-comfort/

Comments are closed.