The data on coronavirus spread is thin, patchy, and likely wrongly estimated. It is essential that we understand this to deal better with the next crisis.
Kshitiz
Should you allow a few vulnerable people to potentially die or suffer? The answer is obviously a resounding no. And So it should be. But what if the process of saving those people necessitate measures to send a very large number of people in penury, joblessness, and potentially death? The answer surely is not the opposite of before. It is neither a yes, nor a no. Instead, the question begs for data, the pros and cons of each action, or lack of it. Even then, it is walking on a razor’s edge of a significant moral dilemma. In the wake of the spread of coronavirus, the developed world has decided to go with one option over the other. Is it a resoundingly correct option? It is not convincing that locking the world down to prevent a rapid virus spread was necessarily the best option, although no option is likely to be a good one. There are multiple reasons behind the claim, but the chief one is the near complete unreliability of data on the spread of this virus.
How good is the data of COVID-19 spread?
By now everyone I know understands exponential curves, and the need to flatten them. I completely agree, and have no issues with it. However, do we know how good the data really is? WHO initially mentioned a mortality rate of 1–3%, which is many times higher than the seasonal flu (0.1–0.2% of infected). With the rapidity of spread of coronaviruses between humans, it made sense that the spread had to be curtailed to prevent a sudden spike in fatalities and morbidities. Draconian measures made sense. However, what if the mortality rate is much lower? Would the policy of a complete lockdown be the same? More importantly, do we know the real numbers?
The first evidence of the unreliability of the data, which raised the suspicion, was the very large differences in rate of mortality between S. Korea and China. Both countries have reasonable health care services, governments are responsive, though Korea may have more developed infrastructure. Still, the differences in percentage of fatalities are very large. It ranges from ~10% in Italy (4,032 death in 47,000 cases) to 0.1% in Germany (577 deaths in 19000 cases). Countries therefore present a difference of 100 times in the rate of fatalities. Korea had a death rate of 0.1%, while Wuhan in China, the epicenter of this epidemic, had a death rate of 5%. What does it all mean?
Are the Italians and the Chinese especially vulnerable to the virus, or their healthcare could not cope with the sudden outbreak, thereby increasing fatalities. Both these scenarios are correct in my view. Strangely, Italy has continuously registered an unusually high rate in flu epidemic related deaths in the last decade (68,000 deaths in Italy between 2014–2017 for a perspective, i.e. 22,000 flu related deaths per year), possibly owing to a large, integrated, elderly population.
I think another important factor, and a basic one, is the total number of people tested for. It is not physically possible to test everyone, and countries would differ on the basis of the number of tests they could realistically perform. But, if the number of tests are so low, we are likely going to get a patchy picture of the underlying data of viral spread. Worse, it may be severely biased, since only the really sick people will get tested, particularly if the tests are few in number. My hunch is that not only the data of lethality rate is patchy, but it is likely to be severely biased towards a higher fatality rate than reality.
There is some evidence of this bias. Germany, which has been doing extensive testing, has a very low rate of lethality. The denominator is therefore higher for number of people tested for, and those that die in response to the infection. Since deaths are easily reportable, even though the reported numbers are going to be lower than actual deaths, it is a safe bet that number of people who have encountered the virus is underreported. By how much? If Germany is the extreme example, it is 0.1% death rate. But. Germany has only 19000 cases confirmed as of now, which is likely to be a very low estimate of the actual number of infected people, or those who have encountered the virus.
I consider that in a nation where the virus has spread as an epidemic, a much larger population has encountered the virus than is reported. 0.1% seems to be a lower bound, which is already similar to the seasonal flu. Of course, even in Germany, the likelier death figures are going to much more than the 68 deaths reported so far. But we are talking of a much lower percentage of fatalities than 1–3% previously thought of as likely. May be the death rates is correlated with very different levels of healthcare facilities, or emergency preparedness present in different European countries?
A fatal flaw in determining the fatality rates
My hypothesis is that in parts of the world, this new virus is rapidly infecting a population that is naïve to its spread, but killing a much smaller percentage of population which encounters the virus. So the worry is about the rapid spread in the wide west, immunologically speaking, and not so much the fatality rates.
More than 99%, and may be 99.5% of the population coming in contact with the virus is unaffected, or act as mediators to spread the virus to possibly vulnerable population. How many of those remaining are vulnerable: It is crucial to determine if it is 0.5%, or 0.1%, or even lower. How could it be determined? The only good way is a randomized test in untested populations where the virus spread has halted, e.g. in Wuhan, China. How many of those untested were found to be positive for the virus, or developed immunity against it will be crucial to understand the epidemiology of its spread. The very thin data on a Japanese cruise ship suggests a strangely low 1% fatality rate, where the spread is expected to have been complete among the inmates. There is another evidence of a semi scandalous coronavirus testing on NBA players where 450 players got tested with scarce kits. The data showed that 65 players had tested positive, about 15% of the players,without symptoms (which may still develop). This suggests that many asymptomatic people have already contacted the virus, they just could counter it, and may not need to be tested. Scandalous though it may be, I consider it is a valuable piece of information about the epidemiology of COVID-19 spread, when such data should have be collected in a systematic manner in a defined untested population. This would have brought the scare down a bit. What this data mean is that a large subpopulation is not showing any symptom, and so the viral spread is fast, but the fatality rate is not really 20–30 times higher than flu as indicated.
The lesson here: a need for rapid, randomized testing on small subpopulation to ascertain the real value of how many people are infected, how many respond, and what are the fatality rate, and morbidity rates. This data is completely unreliable currently, and even though policies cannot admittedly wait for the data, the data is not difficult to obtain with focused studies. WHO, where are you?
Other Factors at Play
For a protracted shut down, our policies should be based on more complete data, which need to be systematically collected and shared now. This is necessary to separate the claims that 50 million people could die, as was alleged by WHO, or a hundred thousand people who were already suffering for many ailments would. A repeated perspective again. Last year 26000 people, mostly old, died in Italy due to seasonal flu. In the USA, CDC estimated a figure of 12000–61000 deaths annually due to seasonal flu. These are damning numbers, and must be placed in perspective amidst all the panic. H1N1 flu, which caused some panic many years ago, killed 151,000 to 575,000 people within one year in the world according to CDC.
Human death is not an easy event to quantify, and cannot be taken for lightly. However, data driven approach with a properly collected data (right now the data is clearly just numbers, and not a scientifically collected data) will help in devising the best policy for the current, and future epidemiological scares.
Interestingly, for countries which are presented as those that stemmed the spread of the disease, other factors may also be on play. As I am writing this article, chloroquine, the standard and an old drug to cure malaria (it was used, as a herbal source, by Peruvians to treat severe fever for centuries), is said to be working for COVID-19 in a limited manner. This is interesting, because malaria is not a viral disease, and so the mechanisms are likely to be quite different. Another interesting factoid is that the COVID-19 spread seems to be much more limited in malaria endemic countries. Temperature is another factor. These factors need to be accounted for before a particular governmental policy of complete shut down (e.g. Singapore) and considered as being instrumental in preventing the coronavirus spread.
The cost of a shutdown
This is not easy to determine. Of course. I come from India, a poor country with a large population, a significant portion of which comprises of daily wage earners, but with a relatively good record in controlling epidemics considering its socio-economic situation. How do you quarantine such a population? Those who twitter and facebook may have the luxury, but for the vast majority of people there is no option but to work. A shutdown that is temporary is bearable, but where each year more than 5000 farmers commit suicide because the basic liquidity is not there to deal with a single seasonal crop failure. In a protracted shutdown, how many will lose their jobs, will have nothing to bring home, will have no option but to let their kids die because the little savings that protected their precarious future is used up in this shutdown?
This is no exaggerated focus-on-a-single-victim scenario, this is a widespread situation all around the world for the bottom half of the population. My father lived in Africa for many years, where the situation is similar, if not worse.
Beyond a few days, there is no choice in India but to work. But even in the USA, but for direct social support (which will be leaky, not properly directed, and never adequate) there is a domino chain of economic activities which will fall down. The restaurants are closed with their hourly workers, who cannot pay their rent or mortgages, which leads to the rentiers and the banks to suffer. This is just one example, and this example extends to all spokes of economic networks. The stocks are shattered, and if they are any reflection of the reality of the broader economic activity, we may be looking at a severe worldwide depression. Today headlines suggest 5 million jobs will be lost in the USA because of this shutdown. One has to take data of economists with some salt, but the effect is likely to be big.
This will have severe consequence on the well-being of humanity, both at the level of economics and life. Just anecdotal examples suggest that restaurant workers will not be paid beyond 15 days, personal gym trainers lost their earnings, and a startup I know of fired 20% of its very large temporary employee base. Quarantines are purely class-based in most parts of the crowded world where cheek-to-cheek living is the norm for poorer classes. How do you maintain social distancing in the shanties of Mumbai and the slums of Mexico City, even if you do not work? They are trying to find the trail of an infected man who took a communal bath in his shanty in Mumbai, a bath used collectively by hundreds of people. Provocatively, are we collecting herd immunity among the unwashed masses to protect the quarantined, the “tweeting us”?
This is all possibly bearable, and is probably welcome to prevent the collective suffering, if it is temporary, say a month or a bit more. This is exactly what is needed if we could hammer it down! After all, Hebei in China has finally stopped having any more cases for 2 days in a row. But if it is for much longer, then the effect of these policies also need to be properly weighed. Just like the trolley dilemma, it is a difficult moral compass to navigate with.
For the next crisis, lets learn for now
Although the WHO, CDC etc. have conveyed it accurately, many people have responded to the crisis believing that they may themselves get the virus, be infected and be adversely affected. Current data suggest that is highly unlikely for most people, but what it will prevent is a spread to the vulnerable population which might get affected. Flattening the exponential curve: the phrase has taken hold. A comparative perspective is necessary about fatality rates of normal flu, and novel flu viruses (again: 22000 deaths per year in Italy, 16000–61000 deaths per year in the USA; H1N1 killing about 175000 to 575000 within one year in the world). These numbers should be repeated alongside live tracking of deaths, which my alma mater Johns Hopkins Medicine, is admirably presenting, and which is lapped up by tweeting masses like a macabre sports display card. Causing fear, and panic without the appropriate context and perspective.
It is an admirable response of the world, and tells us that in response to an external crisis, people as a collective can take a very altruistic approach. This crisis has spread so quickly that it not correct to blame the governments for doing what they thought was right in the very first encounter. The time to marshal a population in varied political arrangements to significantly alter their collective social behavior was too little. My purpose is not to apportion blame, but to present an argument that complex scenarios do exist, and wrong data could lead to scare-mongering, panics, and devastation of economic activities at the cost of the most vulnerable parts of our population. Should we encourage panic in masses who think they will contact the disease, or instead, actively encourage them to quarantine to protect the vulnerable others. Instead, should we selectively identify vulnerable populations and quarantine them.
Kshitiz is an Assistant Professor in the Department of Biomedical Engineering, UConn Health. He is not an epidemiologist, so here is a rider. The views expressed are his own, and are developing as the virus spreads and new data emerge. The objective is to not question the lockdown, since it is a rapidly developing situation, but to insist we learn from it for future disaster management based on good data.
