The real infection cases

As more and more countries are exiting the worst phase of COVID-19 outbreaks, epidemiological efforts are underway to understand the true extent of infection.  More RNA and Ab testing is being conducted for a wider population.  A new published study looked at the problem from a different angle (https://www.thelancet.com/journals/lanpub/article/PIIS2468-2667(20)30089-X/fulltext): how classification of COVID-19 affected the documentation of cases in China by Feb 20, 2020.  The authors focused on the first to fifth editions of COVID-19 guidelines issued by the Chinese Health Commission.  They found that, in general, going from one edition to its later version increased the case numbers, such that if the fifth version were adopted from the beginning, China would have had 232K cases instead of the documented 55K on Feb 20.

It is always good to debate specific data, instead of vague, subjective arguments.  The problem with the paper is it was based on pure modeling, and it acknowledges many of the shortcomings itself.  One thing it didn’t acknowledge though, but which is plain to everybody, is they didn’t really need to look at the first to fourth editions and do all the math; all they needed is the newly confirmed cases on Feb 12, divide the number on Feb 11, and multiply the total infections on Feb 20, to get a similar, final number.  Why it is so is simple: there was a big change from the 4^th to fifth edition, such that about 4 times as many cases were counted on Feb 12 as on Feb 11; and this latest change occurred with the most case numbers, near the end of their analysis period, so it carried the most weight.  The authors or their modeling seemed to deduct if one used the fifth edition from the beginning, the cases would be 4X as well.  While this is all good for modeling, it did not reflect reality on the ground.

The most critical assumption by the authors consciously or unconsciously is that 75% of 232K cases were missed because these people were considered not infected based on the editions and sent home and never heard of again.  This was invalid.  There was a very aggressive monitoring system in place: there are always the infected cases, suspected cases, and close contacts and other quarantined.  The last category needed a certain period of close monitoring, like 14 days.  If during this time symptoms appeared he would undergo further testing.  This is the standard practice for any infectious disease in any country, and how COVID-19 was handled since the 1^st edition.  It doesn’t matter which editions one used.  So missing out on a previous edition doesn’t mean missing out forever.  The last category likely includes millions of people in China, far bigger than the 232K.  If they, or anybody else, developed symptoms, they would likely be identified.

But how about one was just missed based on the old edition, and then he infected more people later?  This is essentially what the authors banked on.  While this scenario can never be ruled out, it is unlikely a major contributor, because the people he infected would have a very good chance of being pick up, all still being under the same, aggressive surveillance system.

In summary, the paper is purely a model play that suffers the same faults as others.  That the 80K or 50K infections in China or Wuhan are underestimates is almost certain, but the main culprit is not differences in the guidelines, but how people go for testing, and how any society can trace infection cases.  Most underreporting is due to people not knowing they are infected and hence, not getting the tests.  You don't go to see a doctor whenever you feel sick, just common sense.  This was true in China in Jan and Feb, also true in any other country in Mar and April.  For example, in countries with the highest death rates, like Italy, France, and Spain, we will surely find many many more infected cases in subsequent reports.  Just don't accuse them of hiding something.  

When people look at the data from China, 84K infections, 4600 deaths, they can have a misleading, overall impression.  This is true for any country including the US (see NY), but particularly so in China, and the contrast is striking.  In essence, China had two infection pictures, one in Wuhan, one outside of Hubei (my April 9, 2020 blog).  For simplicity, Hubei outside Wuhan is in the middle, but ignored for the moment.  Of the 10 million residents, Wuhan had 50K cases, 3869 deaths.  Other Chinese provinces, most like many major European countries in size and population, have a total of 16K, 130 deaths.  

Not considering the potential, inevitably underreporting everywhere, the situation in Wuhan is comparable to those in Italy, Spain, France, UK, each with 50-60 million people.  Why is the death rate in Wuhan a bit lower, <8% vs >10%?  One factor is in Europe 20% or more deaths come from nursing homes.  China doesn’t have many nursing homes, such that the concentrated deaths were rarer.  Another factor is that, after the Jan 23 lockdown and initial panicking (my Jan 26 and Feb 9 blogs), fresh supplies and manpower rushed in from all over the country, which greatly relieved the pressure on the local medical system.  This is far less the case in other countries.  An indicator is the infections of doctor and nurses.  It was reported early on, among the local doctors and nurses, 4K were infected, maybe 2-3K from work.  But for the 42K newly arrived doctors and nurses (for the whole Hubei province), none was infected.  Infection of the local staff stopped as well.  Even allowing for minor overlook, this is a tremendous marker of improvement in the situation.  It is hardly difficult to understand whatever happened in Wuhan, if one has paid attention from the beginning.  The final infection and death figures would have been lower without people jamming and crashing the hospitals around Jan 23.

But for provinces outside Hubei, the difference is stark.  It is a tall order to find any country doing any better than, or even close to, any of these Chinese provinces.  Most have 50 million or more people, a few over 100 million.  The most infections are about 1600, including hundreds of foreign imported cases.  The most deaths are 22, a number of provinces have 0.  All the provinces give daily updates independently, patterns and trends emerging and converging consistently.  Also comparable to Hong Kong (with many foreign imports), but HK has < 10 million people.  Western media hardly question these data, but harp on the much worse Wuhan data?  Simply illogical.  Because there is no indication anybody is faking anything in these provinces (none in Wuhan either).  The data can be explained only by science, and have been further validated by subsequent figures from other countries (April 5, 2020 blog).  If one accepts data from these provinces, he must accept data in Wuhan as well, not the least being not far off from the similar European situations. 

My April 9 blog explained a few reasons why these provinces did so well.  Basically, if you have the whole province’s weight on a few hundred cases, anybody can do it.  If Italy or France had only 1000 cases, they would do splendid, too.  Then why did the provinces have only a few hundred cases, not the 200K in Italy?  This is because they acted right after Wuhan lockdown.  One week after Jan 23 all provinces had locked down.  If the maximal severity of Wuhan lockdown is 100%, for these provinces it was 70-80%.  For Italy, perhaps the most stringent in Europe, it was at best 50%.  Thus, despite many many more people from Wuhan reaching other Chinese provinces than overseas, China outside Hubei did much much better.  South Korea also acted very quickly, ending up with 10K infections.  It must be stressed that East and Southeast Asian countries do much better, despite being much closer to China and having the first cases earlier than Europe and the US.  Even European countries aren't the same, with Germany, Austria, and a few others doing better.  COVID-19 has been at worst a feeble pandemic in China.  Its becoming a global pandemic was never a forgone conclusion, disheartening it has.

Another usefulness of infection figures, whose consequence is now being realized in many countries, is that reopening is critically dependent on the data.  When a lockdown or stay-home order is imposed, most people will follow.  Say if only 50% of the citizens do, then human interactions will drop by 1-0.5*0.5, or 75%, which is a low estimate.  Because of this significant drop, newly confirmed infections, based on past human-to-human transmission, will peak in about 3-4 weeks’ time, which is uniformed everywhere, another big knock on the China conspiracy theory (April 5, 2020 blog).  But coming down from the peak to near 0 will depend on how high the peak is and how long you keep the lockdown.  For China outside Wuhan, it was early Mar back to a new normal, so 6 weeks later.  For Wuhan, it was Mar 22, two months after lockdown, when people start going out, or April 8, when planes and trains start operating.  Other countries aren’t going to wait that long, so it will be interesting to see how it turns out. 

      

Ultimately, retrospective studies will give a clearer picture of the true scale of COVID-19 pandemic.  The true infections = reported cases x N, N>1.  N will vary within a tight range among China and other countries.  The key is the reliability of the Ab test: how sensitive it is, the low detection limit, and how specific it is, not detecting other CoVs.  It further depends on the disease: we still don’t know if every patient will develop detectable Ab or how long it lasts.