Skip to Main Content

In December 2021, one year into the Covid-19 pandemic, my colleague Alina Deshpande had the idea of searching historical disease records for potential “black swan” disease outbreaks. These are unexpected outbreaks notable for their duration, spread or severity.

Why bother doing this? Identifying common features among exceptionally large outbreaks could serve as warning signs of future pandemics or unusually devastating outbreaks.


The spread of disease depends on three major factors: One is a pathogen’s ability to adapt to a wide variety of hosts. Another is the population of susceptible hosts that could provide a home for the pathogen. The third is environmental conditions that support — or hinder — the pathogen. If a newly mutated coronavirus has characteristics that enable it to infect humans, disease spread is likely. If a pathogen infects people who already have antibodies that can recognize and neutralize it, the pathogen would stop spreading. And if a new pathogen infects a person who is isolated and dies before they interact with another person, the disease stops there. In these scenarios, lack of susceptible hosts and/or unfavorable environmental conditions constrain the outbreak.

To understand if each of these factors is equally important, our research team at Los Alamos National Laboratory turned to a visual analytics tool Deshpande had developed in 2012 called Analytics for Investigation of Disease Outbreaks, or AIDO for short. We had designed it to help public health professionals understand and respond to new disease outbreaks by comparing them to historical ones. (AIDO is freely available to public health officials, researchers, and others.)

AIDO includes a database of detailed information about more than 600 outbreaks of 40 distinct diseases: chikungunya, Ebola, malaria, measles, plague, and more. It also includes analytical packages that show the user an outbreak’s trajectory based on comparisons with the database. We used AIDO to see if there were black swan events in its collection of disease-specific outbreaks — there were — and identify common clues that could be used in future outbreaks as warning signs for a black swan event.


We defined a black swan event as an outbreak with more than 10 times the number of cases compared to the average case counts of other outbreaks of the same disease. Such events were seen in almost every disease in AIDO.

We then identified differentiating factors for each event and classified them in three categories: pathogen, host, or environment. Taking it a step further, we recognized that environment could be subdivided: natural environmental factors could include a recent hurricane or climate-caused changes in an area, while human-made environmental factors include the built infrastructure (like a poor or robust public health system) or human behavior. The results of this analysis offered plenty of food for thought.

A 2011 anthrax outbreak in Zambia is a good example. Numerous wild hippopotamuses had been infected with Bacillus anthracis, the bacteria that causes anthrax, which live in soil. Dry weather had forced the hippos to forage on land, rather than in the water, where their grazing stirred up the bacterial spores, which they inhaled. Local people, already in a situation of sparse food supplies, used the dying and dead animals as food. Nearly 500 people became ill from eating the hippo meat. In the AIDO library, the average case count for anthrax outbreaks is less than 50, making this a black swan event.

Given the size of the hippos, a significant amount of meat was shared for consumption across the community, including long-term products such as jerky that could be eaten over weeks or months. Anthrax spores can survive for years under such conditions, spreading the disease beyond the initial event. Extensive educational campaigns were needed in all the nearby villages to contain the outbreak.

The 2016 Zika outbreak in Brazil was another potential black swan event. Zika virus, like the virus that causes Covid-19, was an emerging pathogen in that area. Cases were initially misdiagnosed as dengue fever, which is endemic in Brazil. Two invasive mosquito species, Aedes aegypti and Aedes albopictus, spread Zika virus among a large, immunologically naive population. During this outbreak, about 15,000 cases were reported from Brazil's Ceará province alone. The initial misdiagnosis and the presence of a new pathogen contributed to delays in establishing appropriate public health control measures and information campaigns.

What surprised us the most was the role of human behavior in black swan outbreaks. In 1994, for instance, a small outbreak of plague, which is caused by the bacterium Yersinia pestis, in Surat, India, caused approximately one-quarter of the city's nearly 2 million residents to panic and flee — many on public transportation — ultimately spreading the disease to other parts of India. The outbreaks were not happening because a new disease had entered a susceptible community; it was the human response that so effectively spread this old pathogen.

This early experience with AIDO suggests that it can be used to give public health experts a heads up before a potential pandemic springs to life, to provide actionable information for an unfolding outbreak, and to offer a quick estimate of an outbreak's trajectory. That trajectory can then be used to make decisions about how to mitigate or control an outbreak in its early stages.

We weren't able to use AIDO to initially assess Covid-19 because, as an emerging disease, there were no historical Covid-19 outbreaks against which to compare it. In December 2021, though, we applied the AIDO model to the Covid-19 data available as of March of 2020, essentially pretending we were back in that time, and saw unmistakable signals that this outbreak, then spreading beyond China, had all the warning signs of a potential pandemic or black swan event.

The AIDO insights into the effect of human behavior on disease spread aligns with what has been seen during the Covid-19 pandemic, with behavioral responses to the timing and extent of lockdowns or mask-wearing recommendations having had major effects on either blocking or spreading the disease.

Predicting the spread of infectious diseases, it turns out, isn't nearly as difficult as predicting human behavior. Maybe an AIDO for that should be our next task.

Nileena Velappan is a bioscientist in the bioscience division of the Los Alamos National Laboratory in Los Alamos, New Mexico.

Create a display name to comment

This name will appear with your comment

There was an error saving your display name. Please check and try again.