## A “20 million lives” pay off: how to curb the effects of the COVID-19 pandemic

• ###### J. Bughin

Dr Jacques Bughin, UN consultant, Solvay Business School ULB, Portulans Institute and G20Y, former Director McKinsey Global Institute, and senior partner McKinsey & Company.

### 1. Introduction

#### March 3

In yesterday’s companion article, I have mixed the ranges of values of three key fundamental drivers of pandemic development, based on data collected from academic studies and international health organizations such as the WHO, to derive a possible distribution of global fatalities arising from the diffusion of COVID-19 virus. The outcome is that the potential of fatalities has a mode at 20 million people worldwide, with a standard deviation of 5 million, or a pandemic more than tenfold the severity of a flu pandemic, if COVID-19 is able to continue its curse freely. History of self-protection and health policy has usually been able to curb pandemics, even if more difficulty with viruses like new forms of H1N1 influenza. Assuming same success rates as in the past, the actual pandemic may take a toll of 2 to 4 million lives worldwide.

A casualty of 4-20 million lives lost is a big figure, adding, if it ever materializes, about 5% to 25% additional increase to the world crude number of deaths per year. Furthermore, using as a very rough proxy, the various demographics of infections and deaths from South Korea and China, 40 to 50% of the deaths seems to cluster around the age of 45 years old. This means that 20 years of human capital for those with ability to work is lost, leading to a value loss at purchasing power parity of wages.1 A complete sizing implication is deferred to another future article, but suffice is to say that it is important to take actions to limit the COVID-19 virus to spread freely across the world for human and economic matters.

This article introduces some of methods to limit the diffusion of the virus, with in background the question as to how to limit its spread to become more like a severe flu (>10 times fewer casualties), even stop its curse as soon as possible in the next quarter or so. We will also look at the case example of China, as this was the first country affected. In fact, after underestimating the virus and signals given by some of their doctors in Wuhan, China has taken many measures, from focused measures (building additional care capacity, or sending enough doctors), to more radical measures such as the requirement of large containment of the population in the main Chinese provinces around the Wuhan epicenter. The good news at this stage is that those active strategies seem to have flattened the spread of the disease, even if the epidemy is still ongoing. Early reproduction rates, $R_0$, inferred from the free outbreak were above 3, with the effective reproduction rate, $R_t$, to come down significantly.

The messages we wish to convey, from a review of cases and from own model simulation, are as follows.

• It is still plausible to confine the disease and make it become more like a severe flu
• This however requires agility, speed and massive scale actions
• Consistent adjusted social behaviors are key, meaning that population may have to cooperate without panic for success

Two other side messages are also in order:

• Vaccines have been important to limit the reproduction of new outbreaks. Similarly here, the size of the pandemics may be large enough to sustain a search for it, but looking at past experience, it might be that any vaccine may not play a role in this first cycle of this virus outbreak, given long time, possibly more than one year, to develop secure its use as well as to execute the logistics of vaccination worldwide.
• This is a global phenomenon - meeting the three messages above will be most effective if the approach is coordinated, rather than segmented. This is especially important as the world continues to be more and more connected.

### 2. Do not play small- this is a “scale” game

Curbing a pandemic is a game of scaling actions. Consider that in the case of the current drivers of COVID-19, the actual $R_0$ is indeed like 1.9, as per our high-level simulation in the previous post. At this level, a vaccine that protects the population is easy to compute as:

Thus $V$ will have to be adiministered to about 50% of population to hope contain the virus outbreak, not a small number of the population worldwide (3 to 4 billion of individuals) to fully eradicate the disease.2 Likewise, the total mortality rate (fatality per population) needs to come down manifold for COVID-19 to behave more like a severe flu. As the mortality rate is driven by the product of three key drivers (reproduction rate, skewness of social contagion, and fatality rates), each of them may have also to be cut by half (50%) to meet the aggregate objective.

### 3. Controlling the outbreak: scale + agility and speed

Consider each of the three key drivers of the virus diffusion.

#### Reducing the fatality rate

When it comes to the fatality rate, the rate is typically dependent on the institutional quality of healthcare, which is rather « given » in the short-term. The main scale play is then concerned with enough capacity of health units and doctors. China, furthermore, has shown how agile they could be in building additional capacity in matters of weeks in the epicenter of the virus development. China quickly redirected many doctors to support enough capacity to handle severe contaminated cases.

An important driver of the fatality rate evolution is how well one manages to limit the false negatives, that is patients with not enough discriminative symptoms to be diagnosed,and sent back outside hospital confinement. Not only this person is at risk, but may lead to large secondary contagion. In general, reducing the false negative is not easy: in the case of the COVID-19, tests used have a 90% specificity success (they are able to identify you as not being contaminated when you are not), but with only a 40% sensitivity (ability to identify you as contained, when you really are).3 At this level of sensitivity, one might need up to 8 tests for uncertain patients to find out whether they are or not contaminated, with 95% certainty (versus 50%/50% for any new patients).

Producing and executing such a number of tests is obviously a really intense and scale consuming game. It is likely impossible to maintain in periods of early explosion of the pandemics. Agility is then the name of the game, - in this case finding quickly new ways to improve the poor rate of sensitivity. Doctors in the Hubei regions went on a systematic agile testing, and discovered that sensitivity score has been usually low because of their reliance on typical CT machines for diagnosis. They experimented successfully with a combination with genomics and other scanning machines that improve the effectiveness of the sensitivity diagnosis, (even if a multiple diagnostic sequence remains necessary to limit risk of false negatives).

Regarding tests, South Korea is also an example of agility. When realized the human transmission of COVID-19, South Korea’s CDC approved and released a « beta » - series of their country lab companies’ diagnostic tests, even if not sure if they were fully effective, in the hope to better grasp the extent of contamination. The idea was not to be perfect, but to reduce the asymmetry of information with the disease, and act fast, with quick checking on how tests work and adjust accordingly. In effect, South Korea has been the country with the largest number of population testing, and with one of the best flattening of their pandemic curve.

Consider now the super-spreaders. As said in my previous article, super-spreading is a natural outcome of many viruses. Typically, super-spreaders in the population obey a power law like Pareto where the top 20% of population accounts for 80% of the contamination.

In the case of influenza viruses, we have a less skewed distribution of contribution, where the top 20% is more likely accountable for 40% to 65%, average 55% of the contamination. In such an average case, an individual in the top 20% of the population will be about 7 times more amenable to social contagion than someone in the other 80% of population. With a population average of reproduction rate at $R_0=1.9$ (our best case for COVID-19 to date), the contribution rate of the top 20% is 6 (each person contaminates 6), while it is more like 0.9 for the bottom 80% (each person contaminates less than 1).

The good news here is that the COVID-19 long tail of contaminated individuals does not have an exponential propensity to contaminate, which kills the outbreak. Yet, the bad news is that the top 20% are the real contributors to a fast exponential contagion. Given how contagious they are, simple maths show that one needs to spot and contain 75% of them to have the average population reproduction rate fall below : $R_0<1$.4 Alas, a reduction of 50% of the top 20% still make $R_0$ decline from 1.9 to 1.3, and we remain still in a situation of not controlling the pandemic, but we are converging to the $R_0$ of a typical flu.

In practice, it will be rather difficult to spot as many of those super-spreaders. But multiple agile methods may be pursued to identify at least a good portion of them. The first is to trace people who may have come into contact with an infectious individual. By mapping the origin of the individual, one is better informed about the size of secondary cases caused by a single individual. The mapping of online social networks of individuals may lead to clue as to who might be super-spreader. Second, viral genome sequences may provide information on both the timing of the outbreak and structure of secondary cases.5

Third, logic may be used if one cannot spot personally identify super-spreaders. Super-spreaders will be a fortiori dangerous when they are in close social contacts with others. “Super-spreading” events, with long time contacts, such as concerts, fares, etc. can be cancelled, as a way to limit the risk. Further, for major non controllable outbreaks, one can go broader, limiting locations of high frequency contacts such as schools or work. While indeed costly for the economy, a few weeks closure should be the rule, and be just longer than the incubation and contamination time of the virus. If the total is 20 days, and contamination $R_0 = 2$, we should possibly go for some smart shutdown, of 40 days (hence, « quarantine »), in the case of COVID-19.

#### Limiting reproduction rate via control of contamination and susceptible pop

Independently of the super-spreaders, contact tracing is an important play, and a very powerful response strategy, especially when the transmission of viruses appears after the onset, -or concurrent-, to symptoms. It is claimed that this is how the SARS was finally controlled as the virus contagion appeared after the onset to symptoms for SARS, providing an easy case of spotting effectively any infected person. The same is true as well for the Ebola and MERS outbreaks.6

Tracing contact must be critically important and large, at least spotting as many of the super-spreaders; thus again a scale game. But this is also a matter of agility and speed. Leveraging simulations published in the Lancet (see above, and footnote 6), to a population with a reproduction rate estimated to be at $R_0=1.9$ for COVID-19, (as per our base case of $R_0$ to date for COVID-19), a 50% chance to control the pandemics for 50% of contacts is achieved only when: a) intervention before isolation is done in 1 to 2 days; b) disease transmission before symptoms, happens in less than 10% of cases; and c) social distancing is launched at the very early stage of the epidemy, when they are less than a few tens of contaminated cases.

In most of times, the condition c) is difficult, at least when it comes to a new virus. In such a case, the containment tactics to be successful must be even larger and must target not only the contaminated person, but the susceptible person. China after missing its start, constrained quarantine for all symptomatic individual.7

### 4. Ensure an active and appropriate role for the population

The above set of levers can curb the epidemy, but can be socially disruptive, especially when it comes to affect less and less targeted sub populations, and affect a broader part of population. In general, also levers that limit social contacts or use personal information may be seen as contrary to freedom and privacy.

The role of population is nevertheless critical and also is a matter of scale. COVID-19 infection mechanisms require people to wash hands, limit contacts, and generally behave to protect others in case of sneezing, coughing, etc. Those behaviors are not that difficult and challenging, and work very effectively to the extent everyone does it, as proclaimed by the concept of herd immunity. Hence, it is critical to communicate more and clearly about those small changes, versus the case of necessary confinements if eradication becomes more and more difficult.

Viruses are unaware of borders, so it is important to coordinate among connected routes. Likewise, this is also about international aid in the form of best practice sharing, capabilities and capacity support to regions and countries will lower resources, like developing countries. An ability to play a coordinated role will be a proof that globalisation may be an asset, not a risk to our society.

© Jacques Bughin. Written March 03 . Comments more than welcome. All errors are mine. References listed as they are found in the text

1. Riou et al, 2020, adjusted age-specific case fatality ratio during the COVID-19 epidemic in Hubei, China, January and February 2020; MedrXiv.

2. Using estimates from Gavi and WHO, a highly indicative fully loaded cost of producing, packaging and distributing the vaccine per year would like be in the range of likely 10 to 30 billion USD, at purchasing power parity, but this may be a small investment versus the disruption costs in health and economics. See http://www.gavi.org/library/gavi-documents/supply-procurement/. According to audit reports, and industrial data for reference, the top three vaccine blockbusters today include Pfizer’s Prevenar 13 against pneumococcal virus, (5.5 billion USD), Gardasil by Merck (3.5 billion USD) for papillovirus, and Shingrix by GSK against zona (2.2 billion USD). Those costs do not include cost of distribution, etc. — and usually have yet to cover the globe, especially developing countries.

3. See Chen, 2020, Towards data science, https://towardsdatascience.com/statistics-and-unreliable-tests-coronavirus-is-difficult-to-contain

4. Alas, a reduction of 50% of the top 20% still make $R_0$ decline from 1.9 to 1.3, and we remain still in a situation of not controlling the pandemic, but we are converging to the $R_0$ of a typical flu.

5. See Hellewell, et al. 2020, Feasibility of controlling COVID-19 outbreaks by isolation of cases and contacts, The Lancet

6. See Maier, 2020, effective containment explains sub-exponential growth in conﬁrmed cases of recent COVID-19 outbreak in Mainland China, Arxiv.

Learning from the curve

An open source research project on COVID19 and economics. A collaboration between academics to reach out to policy makers and the general public.