Mortality of Covid-19

Intro  

  One of the most asked questions related to the novel coronavirus is "what is the mortality rate of the disease?" And most medical professionals and statisticians will choose not to answer it, because so far the data is not consistent enough to tell. Various countries report things differently, have different testing rates and methods and probably different definitions of what it means to be dead or recovered from Covid-19. To give a perfectly informed answer to this is impossible and that is why the people we look to for answers avoid the question, while people who are not professionals are giving all of the possible answers at the same time. I am not a professional, so I can give my answer and you can either trust my way of thinking or not.

  In order to compute mortality with absolute certainty we need several things:

• the pandemic has to be over
• the number of deaths from SARS-Cov-2 has to be exactly known
• the number of people infected with SARS-Cov-2 has to be exactly known

 Then the answer would be the total number of dead over the total number of infected people (100*dead/infected). During the epidemic, though, people tend to use the numbers they have.

Panic!

 One of the most used formulas is: current number of deaths over the total number of infected so far (100*current deaths/current infected). This formula is wrong! Imagine there would be two people A and B. Both get infected at the same time and no one else gets infected after that. A will die from the disease in a week, B will recover in two weeks. If we use the formula above, for the first week the mortality of the disease is 0, then it becomes 50% after a week and it stays that way until the end. If B would die, too, the mortality would be computed as 0, then 50, then 100%. As you see, not much accuracy. In the case of Covid-19 the outcome of an infection is not known for three weeks or even more (see below).

  But let's use it anyway. Today, the 31st of March 2020, this would be 100*37832/786617 which is 4.8%. This is a large number. Applied to the entire world population, it would yield 362 million deaths.

  Accuracy comes from the finality of an outcome. A dead man stays dead, a recovered one stays recovered. A better formula is current number of deaths over the sum of current number of deaths and current number of recovered (100*current deaths/(current deaths+current recovered)). This eliminates the uncertainty of people who are sick, but still have to either die or live. If we would know with certainty who is infected and who is not, who died from Covid-19 and who recovered, this would actually be pretty accurate, wouldn't it?

  If we use it on Covid-19 today, we have  100*37832/(37832+165890), which gives us an 18.57% mortality rate. "What!? Are you insane? That's a fifth of all people!", you will shout, with immediate thoughts of a fifth of the world population: 1.4 billion people.

  So which number is the correct one? Neither. And it all comes from the way we define the numbers we use.

Reality

  OK, I kind of tricked you, I apologize. I can't answer the question of mortality, either. My point is that no one can. We can estimate it, but as you have seen, the numbers will fluctuate wildly. And the numbers above are not the extremes of the interval, not by a long shot. Let's explore that further while I explain why numbers derived from bad data cannot be good data.

  What are the types of data that we have right now?

• deaths
• infected (cases)
• recovered
• tested
• total population of an area

  And we can't trust any of these.

Cases/infected

  One cannot confirm an infection without testing, which is something that for most countries (and especially the ones with numerous populations) it is really lacking. We know from small countries like Iceland that when you test a significant part of the population, half of the number of infections show no symptoms. The rest of 50% are on average also experiencing mild symptoms. The number of severe cases that can lead to death is relatively small. The takeaway here is that many more people can be infected than we think, making the actual mortality rate be very very small.

  So, can we use the Iceland data to compute mortality? Yes we can, on the part of the population of Iceland that was tested. We can't use that number for anything else and there are still people that have not been infected there. What is this significant percent of the population that was tested? 3%. 3% right now is considered large. Iceland has a population of 360000, less than the neighbourhood I live in. 3% of that is 10800 people. The largest number of tests have been performed in South Korea, a staggering number of 316664. That's only 0.6% of the total population size.

  But, using formula number 2, mortality for from the Iceland data would be 100*2/(2+157), which is 1.26%. Clearly this will get skewed quite a lot if one more person dies, so we can't really say anything about that number other than: yay! smaller than 4.8%!

  We can try on South Korean data: 100*162/(162+5408) which gives us a 2.9% mortality rate.

  Yet, assuming we would test a lot of people, wouldn't that give us useful data to make an accurate prediction? It would, only at this time, testing is something of a confusing term.

Testing

  What does testing mean? There are two types of tests: based on antibodies and based on RNA, or molecular tests. One tells you that the body is fighting or has fought an infection, the other is saying that you have virus stuff in your system. The first one is faster and cheaper, the other takes more time, but is more accurate. In all of these, some tests are better than others.

  There were reports that people who were infected before and recovered got reinfected later. This is not how that works. The immune system works by recognizing the intruder and creating antibodies to destroy it. Once your body has killed the virus, you still keep the antibodies and the knowledge of what the intruder is. You are, for at least a number of months, immune to the virus. The length of time for this immunity depends not on how forgetful your immune system is, but on how much the virus mutates and can trick it into believing it is not an intruder. As far as we know, SARS-Cov-2 is relatively stable genetically. That's good. So the reason why people were reported to get reinfected was that they were either false positives when they were detected or false negatives when they were considered recovered or false positives when they were considered reinfected.

  Does it mean we can't trust testing at all and it's all useless? No. It means that at the beginning, especially when everybody was panicking, testing was unreliable. We can start trusting tests now, after they have been used and their efficacy determined in large populations. Remember, though, that the pandemic pretty much started in January and for many countries just recently. It takes time to make this reality the new normal and people and technology work in a "proper way".

  Testing is also the official way of determining when someone has recovered.

Recovered

  It is surprisingly difficult to find out what "recovered" means. There are also some rules, implemented in a spotty way by the giants of the Internet, which determine which web pages are "not fake news", but I suspect that the system filters a lot of the legitimate ones as well. A Quora answer to the question says "The operational definition of “recovered” is that after having tested positive for the virus (you have had it) you test negative twice, 3 days apart. If you test negative, that means that no RNA (well, below a certain threshold) from the virus is found in a nasal or throat swab."

  So if you feel perfectly fine, even after having negative effects, you still have to test negative, then test negative again after three days. That means in order to determine one is recovered, two tests have to be used per person, tests that will not be used to determine infection in people with symptoms or in people who have died. I believe this would delay that kind of determination for quite a while.

  In other words, probably the number of recovered is way behind the number of infected and, obviously, of deaths. This means the mortality has to be lower than whatever we can compute using the currently reported values for recovered people.

Deaths

  Surely the number of dead is something we can trust, right? Not at all. When someone dies their cause of death is determined in very different ways depending on where they died and in situations where the morgues are overflowing with dead from the pandemic and where doctors are much better used for the sick you cannot trust the official cause of death! Moreover, on a death certificate you can write multiple causes of death. On average, they are about two or three, some have up to 20. And would you really use tests for covid for the dead rather than for the sick or recovered?

 Logically it's difficult to assign a death to a clear little category. If a person dies of a heart attack and it is tested positive of SARS-Cov-2, is it a heart attack? If someone dies of hunger because they lost their job during the pandemic, is it a Covid-19 death or not? If an 87 year old dies, can you really say which of the dozen medical conditions they were suffering of was the culprit?

 So in some situations the number of deaths associated with Covid-19 will be overwhelmingly exaggerated. This is good. It means the actual mortality rate is lower than what we can determine right now.

Population in an area

  Oh, come on! We know how many people there are in an area. How can you not trust this? Easy! Countries like China and Italy and others have implemented quarantine zones. That means that the total people in Italy or China is irrelevant as there are different densities of the contagion in regions of the same territory. Even without restrictive measures, geography and local culture as well as local genetic predispositions will work towards skewing any of the relevant values.

  Yeah, you can trust the number of people in small areas, especially if they are isolated, like Iceland, but then you can't trust those numbers in statistics, because they are not significant. As the virus spreads and more and more people get infected, we will be able to trust a little more the values, as computed over the entire world, but it will all be about estimations that we can't use in specific situations.

Infectiousness

  An important factor that will affect the total number of deaths, rather than the percent of dead over infected, is how infectious Covid-19 really is. Not all people exposed to SARS-Cov-2 will get infected. They are not really immune, some of them will be, some of them will be resistant enough to not catch the virus. I need a medical expert to tell me how large this factor is. I personally did not find enough information about this type of interaction (or lack thereof) and I suspect it is a small percent. However, most pessimistic scenarios assume 80% of the world population will get infected at some point. That implies a 20% that will not. If anyone knows more about this, please let me know.

Mortality trends

  There is another thing that has to be mentioned. By default viruses go through the process of attenuation when going through large populations. This is the process by which people with milder symptoms have a larger mobility, therefore they infect more people with their milder strain, while sicker people tend to "fall sick" and maybe die, therefore locking the more aggressive strains away from the general population. In this context (and this context only) quarantines and social distancing are actually bad because they limit the mobility of the milder strains as well as of the aggressive ones. In extreme cases, preventing people from interacting, but then taking severely sick people to hospitals and by that having them infect medical personnel and other people is making the disease stronger.

  However, statistically speaking, I expect the mortality of the virus to slowly decrease in time, meaning that even if we could compute the mortality rate exactly right now, it will be different later on.

  What about local authorities and medical administrators? How do they prepare for this if they can't pinpoint the number of sick and dead? The best strategy is hope for the best while preparing for the worst. Most politicians, though, live in a fantasy world of their own making where words and authority over others affect what and how things are done. There is also the psychological bias of wanting to believe something so much that you start believing it is probable. I am looking at you, Trump! Basically that's all he does. That being said, there are a lot of people who are doing their job and the best they can do is to estimate based on current data, then extrapolate based on the evolution of the data.

  So here is another piece of data, or rather information, that we have overlooked: the direction in which current data is moving. One of the most relevant is what is called "the peak of the contagion". This is the moment when, for whatever reasons, the number of infected and recovered has reached a point where the virus has difficulties finding new people to infect. The number of daily infections starts to decrease and, if you can't assign this drop to some medical or administrative strategy, you can hope it means the worst is behind you. Mind you, the number of total infected is still rising, but slower. I believe this is the one you should keep your attention on. While the number of daily infected people increases in your area, you are not out of the woods yet. 

Mechanism

  Statistical studies closely correlate the life expectancy of a population with the death rate in that population. In other words there isn't a specific mechanism that only kills old people, for example. In fact, this disease functions like a death probability amplifier. Your chances to die increase proportionally to how likely you were to die anyway. And again, statistically, it doesn't apply to you as an individual. The virus attacks the lungs and depending on your existing defenses, it is more or less successful. (To be fair, the success of a virus is measured in how much it spreads, not how badly it sickens its host. The perfect virus would show no negative symptom and increase the health or survival chances of its host. That's how vampires work!)

  I have no doubt that there are populations that have specific mutations that make them more or less susceptible to SARS-Cov-2, but I think that's not statistically relevant. I may be wrong, though. We can't know right now. There are reports of Italian regions in the middle of the contagion that have no sick people. 

Conclusion

  We cannot say with certainty what is the mortality rate right now. We can't even estimate it properly without going into horrible extremes. For reasons that I cannot ascertain, WHO Director-General Dr Tedros Adhanom Ghebreyesus announced on the 3rd of March a mortality rate estimated at 3.4%. It is immense and I personally believe it was irresponsible to make such a statement at that time. But what do I know? A UK study released today calculates a 1.4 fatality rate.

  My personal belief, and I have to emphasize that is a belief, no matter how informed, is that the mortality of this disease, by which I mean people who would have not died otherwise but instead died of viral pneumonia or organ failure due to SARS-Cov-2 overwhelming that very organ over the total people that have been exposed to the virus and their immune system has fought it, will be much less than 1%. That is still huge. Assuming a rate of infection of 80%, as many scenarios are considering right now, that's up to 0.8% of all people dying, meaning 60 million people. No matter what proportion of that number will die, it will still be a large number.

  The fact that most of these people would have been on their way anyway is not really a consolation. There will be loved grandparents, people that had various conditions and were happily carrying on with their first world protected lives, believing in the power of modern medicine to keep them alive. I really do expect that the average life expectancy, another statistic that would need thousands of words to unpack, will not decrease by a lot. In a sense, I believe this is the relevant one, though, in terms of how many years of life have been robbed from people by this virus. It, too, won't be easy to attribute. How many people will die prematurely because of losing their job, not getting medical attention when they needed it, getting murdered by people made insane by this whole thing, etc?

  Also, because the people who were more likely to die died sooner, or even got medical attention that they would otherwise not gotten, because pollution dropped, cars killed less people, etc, we might actually see a rise of the life expectancy statistic immediately after the pandemic ends.

  Bottom line: look for the daily number of newly infected people and rejoice when it starts consistently decreasing. After the contagion, try to ascertain the drop in average life expectancy. The true effects of this disease, not only in terms of mortality, will only become visible years after the pandemic ends.

  Update: mere days after I've written this article, BBC did a similar analysis.