I understand that the reported data on this epidemic may not be accurate, but what else do we have to go on? In any case, I’d like to point to a few oddities, for what it’s worth:
As of today, a total of 10 people in Indonesia, Philippines, Cambodia, Vietnam, Bangladesh, India and Pakistan are known to have the coronavirus. A few dozen had it but are now virus free. Those countries have over 2.2 billion people.
As of today, 11 people in Iceland are known to have the virus. Iceland has 330,000 people. Does this mean that it’s more easily transmitted in cold weather?
In Norway, 32 currently are known to have the virus in a country of about 5 million. In China outside Hubei, about 13,000 had the virus and about 12,000 recovered, meaning about 1000 currently are known to have the virus. Italy has twice as many infected and South Korea has 5 times as many, even without adjusting for population. A random person in Norway is about 8 times more likely to infect you than a random person in non-Hubei parts of China. Is it now safer to visit Shanghai than Oslo?
You can say the Chinese caseload data is vastly underreported, but it does sort of match the Chinese data on coronavirus deaths, which are also increasingly rare outside Hubei. And the World Health Organization seems to believe the data is plausible.
A few weeks ago, most people looked at this as a Chinese problem. People felt sorry for China. They still have borne the brunt of the epidemic. But it now seems possible that by April the Chinese caseload will have fallen even further and output will be rising rapidly, while Europe will have a soaring caseload and be plunging into recession. (With the US somewhere in between?)
That’s something no one would have anticipated even three weeks ago.
(Oh wait, the Chinese stock market has been doing better than the European markets, a fact that puzzled people last week.)
READER COMMENTS
Nathan Taylor
Mar 3 2020 at 10:43pm
One possibility is the virus has rapidly evolved. Generally selection is for more contagious but less virulent. Selection is for the virus spread, and if you kill the host too aggressively, you can’t spread.
Arguably existing flu strains have evolved towards an optimum. So a baseline guess would be to assume selective pressure is towards contagiousness/virulence of existing flu strains.
To be clear, this is a very speculative possibility for why the data is so odd. And since we know the data itself is so poor, I wouldn’t conclude this has happened. But it’s plausible enough to just keep in the background while the data gets better.
nobody.really
Mar 4 2020 at 12:28am
Another theory: We’re not discussing how many people have the virus; we’re discussing REPORTS on how many people have the virus. And the sources differ.
As I understand it, the virus presents with common, flu-like symptoms, and testing for the virus is expensive. Thus, reports on the prevalence of the virus may be influenced by 1) the prevalence of the virus, 2) the political climate for making accurate reports, and 3) the financial capacity for the local health care facilities to administer an expensive test to everyone who presents with flu-like symptoms.
For example, the BBC reports that Iran has reported 2336 cases, and 77 deaths, for a kill rate of more than 3%–higher than reported anywhere else. Does that mean the virus is more lethal in Iran? or that Iranians are peculiarly susceptible? Or, perhaps, that the prevalence of the virus is under-reported, while the deaths are more accurately reported?
RPLong
Mar 4 2020 at 9:07am
No, we can’t say this. I think what you really mean is “If two people were chosen at random, one from Norway and one from a non-Hubei part of China, then the Norwegian would be more likely to carry the COVID-19 virus.” There are, of course, many things that determine whether meeting someone will result in a viral infection. You might be more prone to engaging in behaviors that would make you susceptible to infection. The person you meet might be more prone to engage in behaviors that transmit a virus. You may have a better diet and a healthier lifestyle than another person. The true likelihood of infection is a number that represents all sorts of biological and behavioral factors.
I suspect the that data looks the way it looks because demographic, behavioral, lifestyle, and other factors differ quite a bit across countries and even within countries. Iceland, for example, likely has a higher rate of infection because almost the entire population of the country lives in just one city, and because the country’s primary source of economic activity is tourism.
Alan Goldhammer
Mar 4 2020 at 9:14am
There is really nothing strange at all. The zoonotic outbreak took place in one region of China and that area, Hubei, was hit very hard. Public health reporting always focuses on morbidity and mortality and if there is little occurring, not much is reported. This may be why regions outside Hubei, South Korea, and for some mysterious region the Lombardy area of Italy show no large numbers of infections. this is Epidemiology 101.
When I first looked at the Hubei numbers from the Chinese Health Ministry, I found it striking that case reports were extremely low for those <20 years of age. This points to mild infections in this age cohort and likely a large under reporting of actual infections. As I’ve noted before, what is really needed is large scale testing within the population to establish the baseline infection number. It is only then that actual morbidity and mortality statistics can be compiled. That the US with its large biomedical industry could not implement large scale testing promptly is a huge black mark.
I’m not sure what the quality of public health infrastructures are in the countries noted in Scott’s post. It may be that there are many cases of COVID-19 infection but little morbidity and mortality. IMO, we have no way of knowing.
A couple of lessons that should be learned from this outbreak: 1) animal control efforts to minimize chances of zoonotic outbreaks need to be taken in countries where reservoir animals are part of the diet; and 2) research into rapid vaccine and medicine development needs to be the focus of constant investment rather than being slowed down or discontinued when an outbreak wanes. On the second point, we have novel vaccine approaches that should have been piloted before this outbreak so they can be scaled up rapidly. Small biotech companies, while innovative, lack the infrastructure to produce vaccines at scale.
Scott Sumner
Mar 4 2020 at 12:23pm
Everyone, Some valid points, but none of them even come close to making this data seem anything but extremely weird.
For instance, huge numbers of Chinese tourists visit Vietnam, Philippines and Indonesia, far more than visit Iceland.
And yes, the data misses some young people with mild symptoms, but that’s likely true in most places.
The Iran data is unusually fishy, as was noted, but I didn’t even discuss Iran in the post.
I’m still open to the possibility that there are huge problems with the data that could somehow explain this, but as each day goes by that seems increasingly unlikely.
P Burgos
Mar 4 2020 at 3:16pm
What’s the demographic profile of those Southeast Asian countries? If you have a really young population, and pretty poor detection methods, would it be possible to have a large outbreak without many deaths or severe cases?
Alan Goldhammer
Mar 4 2020 at 5:45pm
Scott, you are missing the point. If you don’t have an accurate denominator, in this case the number of COVID-19 infections, you can not say reliably what the mobidity and mortality is. Yes, there is a lot of Chinese tourism, but if those tourists came from Hong Kong or Shanghai, chances are there was not a transmission of the virus to the countries you mention. As I noted, it comes down to traditional epidemiology and accurate background data.
Jim Lumbers
Mar 26 2020 at 11:49pm
Demographic profile explains a lot. Shorter lives means fewer old people and relatively more young people and a higher secular death rate. So lower detected and reported COVID 19 infection and mortality rates. That said, the differences in numbers are so enormous that they cannot be believable Even allowing for peculiarities in Chinese reporting. All of which stands as a serious indictment of the WHO which has responsibility for global monitoring. The Organization has a duty to be constantly examining the effectiveness and quality of members’ reporting. How frightening that economies are being disrupted and live lost through the deficiencies of this politicised, corrupt organization.
Ahmed Fares
Mar 4 2020 at 1:43pm
Dr. Sumner,
China measures its cases differently. This from a reddit comment:
It is true that the death rate reported by China is heavily misleading. But this is NOT due to an active cover-up. There are 2 main structural reasons:
This is primarily due to the structural method of how China records deaths on their certificate. It is established policy/practice in China to record the final cause of death, rather than all existing conditions and overlapping factors.
For example, if a (say 85 yo) patient in the US with diabetes and an existing heart condition gets nCoV, is admitted in the hospital, is confirmed with nCoV, then dies of heart failure, he is recorded as dying of nCoV AND heart failure with other complications. However if the same patient dies in China, he would only be recorded of dying by heart failure.
This is a well-known issue with China and co-morbid diseases. I don’t agree with it, I wouldn’t do it, but I don’t run China. But this is not a new method they made up to try to hide deaths here, it’s just the way it’s done. This has led to jokes in the epidemiology community that “it’s impossible to die of flu in China”, because they basically don’t record any deaths where the patient has flu. See here this recent article from the Global Times, which is one of China’s state-sponsored newspapers.
This is not something even China is really trying to hide. They just tell us, sorry, our doctors just do things this way, we have no interest in changing it.
The quote above contained a link with reference to the flu measurement problem titled: “Data methods show gap between US, Chinese flu-related deaths”
https://www.globaltimes.cn/content/1177725.shtml
Scott Sumner
Mar 4 2020 at 11:18pm
Interestingly, the death rate in China is higher than in many other countries.
Hazel Meade
Mar 4 2020 at 5:51pm
I think a lot of people are not just not taking into account underreporting of mild cases, they are combining estimates of mortality based on cases NOT being underreported, with estimates of infectiousness based on assumptions of massive underreporting.
EITHER only 75,000 people in a city of 11 million got this disease, and 4% of them died, OR lots more people got it, and a smaller percentage of them died.
David S
Mar 5 2020 at 1:30pm
If you look at the statistics, some interesting patterns appear that should be reported more widely:
If you are under 50, you are in approximately 0% danger if you get the virus
If you are over 50, but are not sick you are in <1% danger if you get the virus
If you are sick, this is bad news – stay away from everyone.
The best real data that we have is from the infected cruise ship. 3,711 people were confined in close quarters with people that were infected. 705 eventually tested positive for the virus. Half of those showed no symptoms at all while sick.There were 6 fatalities, all over 70 years old. So the fatality rate in the US should be under 1%.
The US currently has 80 confirmed cases, and 9 deaths – but this is not a typical scenario. Unfortunately, a medical worker that was working with the first patient in Washington state got infected. He had a night job at a nursing home for old people too sick to stay home. That was really bad, a lot of very old and sick people got the virus, and they couldn’t fight it. But that doesn’t say much about the real world.
The reason China looks so bad is that “everyone” smokes and air pollution in the effected area is so bad that everyone is sick. The reason the statistics are so weird is that the virus typically has no symptoms, so unless you implement testing of healthy people you will not know how far it has spread. So essentially the reported statistics (deaths and severe illnesses) are more closely related to existing illnesses, old age, and environmental factors rather than the actual spread of the virus.
Mark Z
Mar 5 2020 at 9:13pm
I’d be interested in the age distribution of the people aboard that cruise ship, particularly those who tested positive. The stereotype of course is that cruises are mostly popular among older people. 6 fatalities out of over 700 people may be lower than expected for the age group (the median cruise ship passenger is in their 60s; mean however is 47, but I suspect the ages of cases skew upward). The average age of infected people in Wuhan was about 55.5, almost 20 years above the median age, so either young people are a lot less likely to contract it in the first place, or there’s a strong bias among people who test positive toward more symptomatic/severe cases (which tend to be older people).
I think you’re right that the overall fatality rate will be under 1% in the US; it was 0.9% in South Korea and it would seem is now currently below 0.7% as they’ve been testing more people. The US’s abysmal mortality rate is mostly due to so the fact that testing has largely been restricted to people who are already very sick or are close to people who are very sick, thanks to the stringent rationing of tests.
Michael Rulle
Mar 6 2020 at 2:30pm
If China has in fact turned the corner (btw, I assumed government was buying stocks!) and by that I think you mean the absolute number of people with the Virus will be declining, then the “world’s reaction” to this is a factor of crazy we may never have seen before.
There should be, I think, no particular reason why the rest of the world should be any worse than China on a proportional basis——accounting for normal “standard error’ kind of thinking. It does seem odd to have a northern Italy cluster, but no more odd than having a Hubei cluster.
If I am doing my math correct, 1 out of a thousand people in Hubei have contracted Covid-19. In America, 1 out of 5-10 people get some kind of Flu—-which is a pure guess as the estimated range is so wide and not from”counting”. Outside Hubei, the numbers are so small it is practically speaking, close to zero.
If that is the case, the total numbers worldwide will be extremely small. So small in fact, that it is plausible to imagine we have these kind of mini-breakouts of a variety of Coronas and Influenzas that are never detected and whose death rates just end up being swallowed up by the wide ranges that exist in our estimates of Flu deaths—(this year between 25000-45000 in the US alone.)
When enough time passes we will know much more.
Kris
Mar 9 2020 at 6:10am
Some perspective on Iceland. As of now, there are 58 known infections in Iceland, which sounds very for a population of 360K. Apparently all but one infection has been traced to Italy or Austria. Most of the infected people were skiing or vacationing in Italy/Austria, and the other infections that have spread in Iceland have been traced to people who came from Italy/Austria.
It’s a bit of a mystery why the per capita infections are higher in Iceland than in other countries. My guess is as good as anyone’s: Well, the obvious thing is that the source of the currently identified infections seems to be known. Icelandic health authorities have been very proactive screening for the virus and are actually beefing up the screening effort with the help of a major genomics company that’s based in Iceland. All people coming from Northern Italy and other affected areas have been asked to enter into a 14-day quarantine, and most seem to be complying. A small population also might make this more manageable.
58 infections in such a small population is still scary. Not only for Iceland (as the source is known), but in general. It begs the question if the spread of infections is vastly underestimated in other counties?
Kevin
Mar 13 2020 at 4:07pm
I’m not an epidemiologist so I don’t have much context on these numbers, but I am a Data Scientist and this appears to me to be a classic case of apples to oranges. To clarify, what I mean is that the method in which this data is being collected and reported varies so wildly from country to country that trying to compare the infections per capita between say Germany and Indonesia is like comparing Apples to Oranges, despite both countries claiming to be measure the same thing.
What actually led me to this article is that I was trying to figure out the anomaly that is India’s data. Their first case was a month and a half ago, yet to this day they’ve only detected 81 cases, 2 deaths. This is strange as India has 1.2 Billion people and is one of the most densely populated countries on the planet, with substandard water and sewage infrastructure, widespread communal living, and very little centralized control to implement quarantines and mass screenings. While I should clarify that I work with a lot of Indians, and love their culture, history and cuisine, the reality of modern India would suggest that they should have more cases than anyone, yet they’ve only detected/reported 81.
Ravi
Mar 16 2020 at 1:39pm
I am in India currently and I second what Kevin said here. The reason why the case of India seems to be an anomaly is because of the fact that India simply isn’t screening people on a scale comparable to even US right now, and US is being rightly criticized for it. The approach in India has largely been, “if people report symptoms, they’ll be tested”. I suspect there are several thousands of cases in India already, in stark contrast to the reported 114 as of this writing. In fact that too might be a low number. I hope I’m wrong but the available data from the world and the probability alone (of India having such low numbers) leads me to thunk the cases are in reality orders of magnitude higher in India.
susan wakabayashi
Mar 24 2020 at 4:51pm
Once we have a better data set to work with (and antibody testing is being done to more accurately assess the denominator in the death rate), it will be interesting to do deeper analysis to see what the impact was of things like:
Superior healthcare
Flattened curve so as not to overburden health system
General health of a population (less obesity, fewer preexisting conditions like heart disease and diabetes, etc.)
Better immunity of a population (maybe no way to measure, but thinking about problems unique to the developed world – like auto-immune/asthma issues, and the fact that lack of exposure to pathogens has been floated as possibly a contributing factor)
All of this would require controlling for things like different methodologies in how deaths get recorded, whether cause of death is even investigated in people over a certain age, etc. But I’m sure some interesting insights can be gleaned from the post-pandemic analysis.
Comments are closed.