What Comes After COVID

Juan Cambeiro

The next pandemic is coming. Is it possible to say when?


From the beginning of agriculture 12,000 years ago to the early 20th century, infectious diseases have probably killed far more people than famine and violence combined. 1 When hunter-gatherers started to farm, they also built permanent settlements and domesticated animals — greatly increasing the likelihood of zoonotic spillover, which would in turn spread more readily in dense, interconnected human populations. Influenza, smallpox, and measles are just some of the infectious diseases that made their appearance in the human population during this period.

As late as 1900, infectious diseases were responsible for nearly half of all deaths in the United States. Infectious diseases continue to kill more people in low-income countries than noncommunicable diseases do. Worse yet, humanity as a whole faces the possibility of devastating future pandemics that could constitute an existential risk.

As we continue to grapple with the ongoing effects of COVID-19, we must start thinking about how to prepare for the next pandemic. The question is not whether it will occur, but when.

What Does It Mean to Forecast the “Next Pandemic”?

Before we can make any forecasts about the next pandemic, we have to be specific about exactly what that means. What counts as a “pandemic”? What’s the time frame we’re looking at? What in particular about the “next pandemic” do we want to know? I will use the following two specific questions. 2

(1) Will a pandemic caused by a novel pathogen begin between 2023 and 2032 and result in the deaths of more than 20 million people?

Since there is no officially recognized threshold for “pandemic,” I’m defining it here as the spread of a novel pathogen that ultimately kills more than 20 million people and that is present worldwide.

Instead of looking at just confirmed deaths, which often undercount true fatalities, I prefer to use the number of excess deaths from all causes during the pandemic. The Economist estimates that COVID-19 caused 20.3 million excess deaths as of the end of 2022, compared to the World Health Organization’s confirmed death count of 6.7 million.

To qualify, these 20 million-plus deaths should occur within a limited window of time — let’s say three years — instead of over many decades. For example, a pandemic for which there are confirmed deaths beginning in February 2030 would have to result in 20 million-plus deaths by February 2033.

I set 20 million as my threshold because it may be of particular interest to ask about the probability of a COVID-like pandemic or worse.

(2) If a pandemic begins in the next decade, what type of microorganism will be the causative agent?

This question is useful for thinking about countermeasures — for example, how we might want to focus prototype vaccine development. To avoid infohazard issues, I won’t get into more detail than looking at different virus families. I’ve divided potential causative agents into four categories: “Orthomyxoviridae virus family” (influenza), “Coronaviridae virus family” (coronaviruses), other viruses, and nonviruses such as bacteria or fungi.

Initial Forecasting Process

To better understand the probability of a future pandemic, it is useful to look at the relevant reference class — in this case, the frequency and characteristics of past pandemics. The past is oftentimes an excellent guide to what might happen in the future.

But how far back do I want to go? All the way back to the Plague of Justinian in the sixth century? Probably not, for a few reasons: A lot about modern-day life is different (we now have antibiotics, for example) and our knowledge of infectious diseases has vastly improved (in particular, we now know that specific microorganisms cause particular diseases). We also have very sparse information about pandemics in the distant past, making it difficult to construct a useful data set.

A logical (and clean) cut-off is 1900. We have substantially more comprehensive data on pandemics or near-pandemics during this time. We also have high confidence about which pathogens caused the outbreaks that occurred after this point. For example, we’ve identified the virus that caused the Spanish flu as the A/H1N1 influenza strain, but we’re not actually sure whether the 1889 Russian “flu” was, indeed, caused by a member of the Orthomyxoviridae family. This period is also characterized by a high degree of globalization, which makes it more applicable as a reference class to today’s level of interconnection.

Here, I include both events widely considered to be “pandemics” (rapid worldwide spread of a disease) and what I somewhat subjectively call near-pandemics: novel pathogens with high lethality potential that saw limited human-to-human transmission and were snuffed out or died out on their own before they reached the pandemic stage. I do this because these smaller outbreaks provide more data about the types of microorganisms that might be most likely to cause future pandemics.

The clustering of near-pandemics in recent decades is probably just because of an unfortunate lack of adequate public health surveillance in the early to mid-20th century — smaller outbreaks likely occurred but went unrecorded.

Next, let’s take a look at our two questions in the context of the above historical data. 3

(1) Will a pandemic caused by a novel pathogen begin between 2023 and 2032 and result in the deaths of more than 20 million people?

In the 123 years between 1900 and 2022, there were a total of 12 known pandemics, epidemics, or outbreaks. Of these, two were pandemics that killed more than 20 million people within three years, for an annualized probability of ~1.6% in a given year. 4 Over 10 years, this would translate to a probability of ~16%. Thus, 16% is my starting forecast.

(2) If a pandemic begins in the next decade, what type of microorganism will be the causative agent?

In the table I constructed, there are four general categories of microorganisms:

Orthomyxoviridae: members of the influenza virus family, which caused ⁵⁄₁₂ events = 41.7%. This includes ½ of the events that meet my “pandemic” criteria.

Coronaviridae: members of the coronavirus virus family, which caused ³⁄₁₂ events = 25.0%. This includes ½ of the events that meet my “pandemic” criteria.

Other known virus: viruses from any of the other 24 known human-infecting virus families, which were responsible for ³⁄₁₂ events = 25.0%.

Nonvirus or unknown virus: any other microorganisms, including viruses from unidentified human-infecting virus families, bacteria, and fungi, which were responsible for ¹⁄₁₂ events (cholera) = 8.3%.

From the reference class, we can clearly see that pandemics and near-pandemics have occurred fairly frequently throughout recent history and will likely continue to occur in the future. However, the probability of a future pandemic should also take into account the current state of technology and society — which is what we’ll take on next.

Adjusted Forecast/Inside View

(1) Will a pandemic caused by a novel pathogen begin between 2023 and 2032 and result in the deaths of more than 20 million people?

I’ll start with my initial forecast of 16% and adjust it based on factors I think are relevant.

Factors that lower the risk of a pandemic in the next decade:

Better sanitation, hygiene, and general health: 16% → 13%

Increased likelihood of speedy detection of a novel pathogen and subsequent rapid development of vaccines/therapeutics: 13% → 10%

Improved medical practices and medical system infrastructure: 10% → 8.5%

Possible use of nonpharmaceutical interventions like widespread masking and physical distancing: 8.5% → 7.5%

I ended up with a probability of 7.5%, down from my starting point of 16%. The most important factors here are better health and sanitation, increased likelihood of speedy detection, and the development of vaccines and therapeutics. Global rates of malnutrition, which greatly increases the likelihood of infection, are significantly lower than they were in 1900. In the realm of medicine, our response to COVID has shown what is possible: In the U.S., it took only 326 days from the first U.S. laboratory-confirmed case on January 20, 2020, until the first FDA-authorized vaccine on December 11, 2020. The COVID vaccines likely saved nearly 20 million lives worldwide by the end of 2021.

However, there are other factors about the current context that increase the probability of a pandemic:

Increased likelihood of unintentional lab leak of a naturally occurring pathogen or pathogen that underwent gain of function: 7.5% → 12%

Continued democratization of biotechnological tools that makes the ability to deliberately engineer deadly pathogens more accessible to bad actors: 12% → 16%

Increased interaction between humans and animals, which creates more opportunities for zoonotic spillover: 16% → 18%

Air travel enabling fast widespread transmission: 18% → 19%

I updated from 7% all the way to 19%, largely because of the increased likelihood of an unintentional lab leak and the continued democratization of biotechnology. My concern for lab leaks is due to the fact that Biosafety Level 3 and 4 labs, where dangerous pathogens are handled and sometimes experimented on, are increasing in number. The continued democratization of biotechnological tools, meanwhile, is providing more and more actors with the capability to engineer dangerous pathogens. My final forecast for a pandemic killing more than 20 million people in the next decade is 19%, or roughly a one-in-five chance.

(2) If a pandemic begins in the next decade, what type of microorganism will be the causative agent?

I’ve arrived at a probability of 19% that a pandemic will start in the next decade and kills more than 20 million people. If this were to happen, what might the causative microorganism be? Eleven out of the 12 microorganisms in my reference class are viruses — five are influenza viruses (Orthomyxoviridae), three are coronaviruses (Coronaviridae), three are other known viruses, and one is a nonvirus. Viruses have historically been more likely than other microorganisms to cause a pandemic and are still epidemiologists’ biggest concern, because of their high replication rate and the lack of a broad-spectrum antiviral.

Let’s take each of these categories in turn. First, I’ll adjust for the fact that events that met my definition of “pandemic” were caused once in 1918 by a flu virus and once in 2019 by a coronavirus. I’ll give more weight to Orthomyxoviridae and Coronaviridae.

Orthomyxoviridae: 41.7% → 50.0% 

Coronaviridae: 25.0% → 30.0%

Other known virus: 25.0% → 15.0%

Nonvirus or unknown virus: 8.3 → 5.0%

On the other hand, while we have experience developing vaccines for influenza viruses and coronaviruses, we do not have a prototype vaccine for 11 of the 26 human-infecting viral families. Many of these other human-infecting viral families are also just generally less well characterized, which increases the danger they pose. So I want to increase the likelihood of outbreak from the other known virus and nonvirus or unknown virus categories. I also want to account for the fact that we are actively concerned about specific flu viruses that can spread between mammals. I think overall this should negate about a third of my previous adjustment for Orthomyxoviridae and fully negate it and then some for Coronaviridae, while substantially increasing other known virus and nonvirus or unknown virus.

Orthomyxoviridae: 50.0% → 47.2%

Coronaviridae: 30.0% → 20.8%

Other known virus: 15.0% → 23.0%

Nonvirus or unknown virus: 5.0% → 9.0%

So, according to my forecasting, there is roughly a one-in-two chance of the causative agent of the next pandemic being an influenza virus, a one-in-five chance of it being a coronavirus, a one-in-four chance of it being some other virus from a known human-infecting viral family, and a one-in-11 chance of it being some other microorganism.

The devastation caused by previous pandemics, including COVID-19, serves as a reminder of the destructive power of infectious diseases. We do not know when the next pandemic will strike, or what form it will take, but by approaching the question with a probabilistic mindset, we can give ourselves the best chance of meeting it prepared.

  1. Calculation by Charles Kenny in The Plague Cycle.
  2. These questions are similar to two included in a biosecurity-specific forecasting tournament I launched on Metaculus.
  3. Another thing to note: If I went back further in history, I’d probably want to adjust for the world population size. But because of the time scale we’re looking at, it’s not necessary to do this: It would not result in any additional events that meet my “pandemic” definition.
  4. HIV/AIDS killed more than 20 million people, but not within the span of just three years — so it doesn’t count.

Juan Cambiero is a Superforecaster. He placed first in multiple forecasting tournaments, including the IARPA FOCUS 2.0 COVID Forecasting Tournament, and is a biosciences analyst at Metaculus. He is also a graduate student in epidemiology at Columbia University and can be reached at juancambeiro1015@gmail.com.

Published March 2023

Have something to say? Email us at letters@asteriskmag.com.