(Moved from old blog- originally published May 7th 2020)
I entered medical school with a stronger scientific background than most doctors. My literary hero was Mister Spock and I considered myself a champion at scientific reasoning.
In my youthful arrogance I thought this would mean I would quickly run circles around the less intellectually talented docs. But in contests of “what will happen to this patient” and “what should we do for that patient” and “anything having to do with medicine at all” I would lose and lose badly, again and again, like some small dull witted child repeatedly falling for a large dull witted adult who keeps “getting my nose” over and over and over…what the hell was wrong with me?
It turns out that in medicine experience beats smarts and basic science every…single…time. That’s why one of the bedrock principles of medical education is “humiliation is the best teacher” which in turn is why “pimping” (are they still allowed to call it that?) by attendings, residents, nurses, the janitorial staff, is a feature of everyday life for the medical student.
Say it with me: “in complex systems like the human body or human society, experience beats deterministic reasoning and simple mathematical models”. Keep saying it in your head until you believe it. Imagine you had the relevant statistics on all the players in your favorite sportsball league: heights, weights, speeds, strengths, ages—could you predict the outcome of the season reliably? Of course not. You’d be rich off your gambling winnings by now if you could. How did Biff Tannen do it? He had the ultimate experience, a book telling him exactly what was going to happen.
This is why I have a sincere question with respect to the results of epidemiological modeling that filter down to us through the press and other sources. Here is a wonderful page which gives a great explanation of the basic deterministic linear models for epidemics that have been around since the 1920s. They note that the governments are basing their decisions on far more detailed models (I should hope so) but, as they use the simpler models to make their point in the article, I’m going to assume that the complex models do not much contradict the results they show here.
Very cool! I am not an epidemiologist and this guy is. He is, in fact, a digital epidemiologist so he should know better than I. I will try to contact him with my questions and failing that, try to track down some other epidemiologists through my network. If you know one please message me.
The point I would like to focus on is his explanation of why it is a terrible idea to try to get to herd immunity without a vaccine. You can calculate herd immunity from R0, which he states is 2.5 for COVID-19. The formula tells us herd immunity achieved when 1 – 1/R0 transmissions are prevented. For COVID-19 this is 0.6 meaning you need to prevent 60% of transmissions to get your effective transmission rate to 1. Below 1, new cases begin to decline.
The assumption in herd immunity is that this is almost entirely due to people being immune. He correctly notes that under the “do nothing but wait to get infected” plan we achieve herd immunity when 60% of people have had it (and either recovered or died) and are immune (which he questions the certainty of). The problem with achieving herd immunity this way is that herd immunity is not the point at which an active epidemic stops. It’s the point at which new cases each day start declining. If you look at the curve from the model you expect almost as many people to catch it after R effective (Re) dips below 1 as did before. In these models he summarizes that under this strategy, and this simple model, 80% of people will get it before it goes away (and almost that many before our ICUs are no longer overwhelmed). This is terrible scenario and the reason we all locked down all over the world.
Okay, pretty scary but obviously even in the absence of worldwide government orders for everyone to stay inside most of the time, people would do things. Some fraction of people would be so scared they would stay inside anyway. Many more would limit their contact with other people, stop going to events, wear masks in public, cancel travel etc. Many local governments, especially in locales with active outbreaks, would close bars and restaurants and schools temporarily.
Some variation on this is probably what happened in past pandemics. It is difficult to find details on what they did and even more difficult to find details about on what they didn’t. I’m assuming all-the-world was not locked down. Otherwise, that fact would be repeated endlessly in the media and we would know the result.
Here is the Wikipedia page on flu pandemics:
Let’s look at Hong Kong flu as the most recent pandemic with a disturbingly high body count. It happened from 1968-1969. I have done some superficial research about vaccination during this pandemic and find no mention (edit-actually see below there was a vaccine available later in the pandemic). Unless someone can show me otherwise I am assuming that during this pandemic they started ignorant, got fearful and took a variety of measures around the country and around the world. They did not have worldwide stay at home orders as near as I can tell.
According to this article: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4169819/
What could have happened according to the model?
The basic reproduction rate of Hong Kong flu was estimated as 1.8. Running it through the calculation about we see that if they had done nothing, herd immunity would have been achieved when 44% of people had it. Amost as many people would then get at as the curve descended. I’m going to naively use the approximate ratio of herd immunity to people who get it shown in the modeling article above, which means ultimately about 60% of the world would have caught the Hong Kong flu if everyone had just laid back and thought of England.
What actually happened according to experience?
What actually happend? According to Wikipedia greater than 14% got it. Following the reference paper it looks like they calculated this from the excess mortaility and presumed case fatality rate. I’ll have to double check (update- I never double checked). Assume this number is as correct as any others we are using for decision making today. I understand that technically 60% counts as greater than 14%, but the meaning of that estimate is that it probably isn’t much more than 14%.
What happened? why the huge disparity? That is the question I have for epidemiologists. I’ll try to get one to explain it to me. By the way, looking at the table in the wikipedia article I don’t see any flu pandemic close to 80% of people getting it. Yes yes I know that COVID-19 isn’t flu which therefore means it has magical powers that must make it much worse than any respiratory virus pandemic in history until proven otherwise. Have you learned nothing?
My lame attempt at explaining it
These are my ideas about why those previous pandemics didn’t hit their theoretical death toll in the absence of near universal lockdowns, vaccines, testing, and widespread contact tracing.
First, as the modelers will tell you, one of the basic assumptions of the model, a homogeneous society with everyone having equal likelihood of coming in close contact with everyone else, is not an accurate reflection of reality. They know this will have an effect but given that they use it to make their point about how “do nothing” is suicidal, let’s assume they think it is on the smaller error side rather than the larger. Based on my experience with how networks work, this actual error might be surprisingly large.
As we saw in my previous post about the contact tracing study in Taiwan, the average person really struggles to transmit outside his family (and healthcare workers). Of the 100 patients they did their detailed contact tracing on, only 1 managed to give it to an outsider. So how does it manage to spread so quickly between family units? Some people, and some situations, are much much much better than average at transmitting. I can go weeks without spending much time around anyone new. Then my friend Dara has so many friends they actually needed to throw him three separate birthday parties each year; he is that close to that many people. Is Dara the cause of the pandemic? Yes, yes he is, he and people like him. If our society is a network of vertices (people) connected by edges (spew) almost everyone of us would have a tiny number of edges. But a tiny number of people would have a huge number of edges. These account for much, most, almost all? of the spread between families. The other factor is probably regular vertices like me who find themselves in super spreader environments like national conferences, choir practices, and, for some reason, meat packing plants (both literal and figurative).
The thing about those vertices (Daras) with all the edges (lots of close contacts) is that they tend to get the disease sooner rather than later. On the average you expect super spreader people to get knocked out of the game earlier, and the way networks behave this could have an outsized effect. I’m certain the epidemiology modelers have investigated this but I can’t find an answer. Hoping to soon.
A second, obvious reason that past pandemics didn’t reach their full potential is, that people learned about the pandemic and got scared and altered their behavior. If you buy my earlier post about the Pareto Principle, these behavioral changes very likely also had outsized effects in slowing down the pandemic.
Yet another reason is that governments did do things. I can’t find a reference to them doing anything near what we have seen in the COVID-19 pandemic but surely, at least locally they probably did things especially as the flu hit their regions. On the other hand, this was 1968 and the majority of governments around the world didn’t have the ability to do anything terribly sophisticated.
The main reason I wrote this post, is I am hoping someone can point me to answers about the disparity or show me an error in my reasoning. Just to provoke and outrage them let me conclude with simplistic calculation.
If the equations say that 60% of people should have got Hong Kong flu but only >14% did and the equations say that >80% of people should get COVID-19 does not experience (which I contend trumps determinisic reasoning and linear models) tell us we would expect >18% to get COVID-19 assuming we take the approximate measures they took for Hong Kong flu?
Shoot- looks like there was a vaccine
It says they manufactured 9 million doses within 4 months of receiving the new virus but fails to mention how far into the pandemic that was. 9 million is not a lot, wouldn’t get you close to herd immunity but could have helped. Here is a journal article from the period:
This article mentions that ultimately 15% of the population of Hong Kong got it. This was the first place hit and would most likely not have benefited from vaccine so the above still holds.
https://www.scmp.com/lifestyle/health-wellness/article/2154925/how-hong-kong-flu-struck-without-warning-50-years-ago-and The journal article agrees with the 15% number for Hong Kong as of that July and never mentions a second wave for them so I read that as 15% is who got it and there was no vaccine involved in Hong Kong getting over it.
I have thought it about more as I have let this post sit and I think we can infer from the above articles that Hong Kong did not take measures anything like what we see today for COVID-19. First they don’t mention any such measures and hospitals told people to just go home and rest. Sources from the time report that it wasn’t considered especially deadly anywhere until it got to California so it is unlikely Hong Kong would have shut its economy down.
Consider also that Hong Kong is (and was) one of the most densely populated places on Earth and back then was under British control, so if your objection is that the Chinese communist party was fudging the numbers, that would not have been correct for the time.