Bostonians’ collective memory of the year 1918 is inextricably tied to the Curse of the Bambino and the last Red Sox World Series victory of the 20th century.
But for MIT’s Richard Larson, Sox fan though he is, 1918 calls to mind a more significant event. What would become one of the worst plagues in modern history reached its East Coast peak that fall. Boston was the first urban center of the great influenza pandemic, and those infected in the city would spread the disease to the rest of the country and to Europe. In September alone, about 1,000 Bostonians died of the disease.
The influenza pandemic lasted through the next year. About 500 million people (nearly a quarter of the world’s population) became infected, resulting in as many as 700,000 deaths in the United States and at least 50 million worldwide. Influenza is normally much more lethal in the very young and very old, but for reasons that remain unclear, the 1918 strain claimed a disproportionate number of healthy young men and women. More Americans died in the epidemic than in all the wars of the 20th century combined.
A few milder influenza epidemics have occurred since then, including the 1968 Hong Kong flu, which kept Larson in bed for a full week. But many researchers fear that we are in for another big one. The U.S. Department of Health and Human Services estimates that an epidemic similar to the 1918 outbreak would kill almost two million Americans.
If such an epidemic strikes in the foreseeable future, we won’t be able to stop it. As far as modern medicine has come since 1918, it still takes longer to develop a vaccine than it would for a new strain of the flu to sweep the globe. If the virus were as bad as 1918’s, Larson says, the death toll could surpass 200 million worldwide.
But Larson also has a hopeful message. The course of a pandemic is not inevitable, he says, and individuals, families, employers, and governments can all do things to contain the spread of the disease. Larson believes that traditional epidemiological models have painted an overly bleak picture, leading to feelings of helplessness among the public and a lack of good planning by the government. “The science that informs the policy treats the flu’s spread as a fait accompli,” he says. But he begs to differ.
Larson, a professor of engineering systems, is in the midst of a two-year project to mathematically model how behavioral changes might contain a future pandemic. As part of that work, he is evaluating state governments’ preparedness plans and recommending ways to improve them. In January, his research team will host an on-campus event where it will provide feedback to representatives from about half the states and several universities. The researchers will also present a “master plan” consolidating ideas gleaned from the states’ plans and from their modeling research.
A New Model
Influenza epidemics happen when a strain of the virus adapted to other animals, such as chickens, mutates to become easily transmissible between humans. Public-health officials are nervously tracking H5N1, the virulent bird flu that has killed 200 people (mostly in Indonesia and Vietnam). Although it has passed between humans only rarely, no one can predict when or whether a particular virus will acquire the ability to do so efficiently. And because a virus that acquires that ability is by definition new to humans, it’s much harder to fight off than the seasonal flu, which at least resembles something our immune systems have seen before. Lack of established immunity to a new strain leaves us naked against its assault.
Quarantines and border controls won’t do any good: people newly infected with any flu are contagious for a full 24 hours before they show symptoms. As epidemiologists put it, these people are “shedding” the virus with every breath. The science may change in the coming years, says Larson, but today it takes six months to develop and test a vaccine once a virus emerges. Six months is plenty of time for the virus to engulf the world. Although the secondary bacterial infections associated with influenza can now be treated with antibiotics, we’re nearly as vulnerable to the disease itself as we were in 1918. With a denser and more mobile population, we may be even more vulnerable.
Epidemiologists debate how often an epidemic as bad as the one that struck in 1918 is likely to occur. Some say it happens only once a millennium, others once a century–in which case our number is up. But epidemics are such complex systems that until one is under way, no one can predict much. It’s impossible to gauge the lethality of a virus before it emerges, and viruses are always mutating in unforeseeable ways. They tend to strike in waves, but the duration, timing, and location of those waves cannot be forecast. “Any serious researcher is humbled by the pre-flu limitations on what we know,” says Larson.
More important for Larson, human behavior is unpredictable. Most epidemiological models take into account neither the heterogeneity of the population nor the fact that people’s behaviors will change over the course of an epidemic. Larson calls this a “fundamental error.” His contribution to influenza modeling is to feed human variability into the equations. Because the flu has no inevitable course, encouraging people to change their behavior can slow its spread and reduce its toll.
Models don’t predict how many people a virus will kill–there are too many unknowns for that. Instead, they predict the speed at which an epidemic will spread through the population before burning out. One of the most important parameters in these models is Ro, the average number of other people who will be infected by each infected person in the early days of the pandemic. If Ro is greater than one, the disease initially spreads exponentially and then subsides, as most of the population considered in the model has been exposed. That is, according to most traditional theories, there’s no containing the spread of the disease: it must simply run its course.
“The idea in good currency,” Larson says, “is that Ro is a fixed constant.” Fixed-Ro equations, which epidemiologists initially adopted in the 1960s and ’70s, are based on concepts developed in Germany in the 19th century to predict population growth. (In the original equations, Ro was the reproductive constant, the average number of female babies a newborn female baby would have in her lifetime.) They provide “an accurate description of something that moves very slowly,” like human population growth, says Larson. But they do a poor job of predicting fast-moving phenomena such as the spread of a highly contagious respiratory virus. That’s why Larson treats Ro as a variable with separate components.
Larson, whose work focuses on operations research, says that because he has not been trained as an epidemiologist, he sees the problem of flu modeling with “new eyes,” and that makes it easier for him to question epidemiologists’ assumptions. “No one knows how any particular virus will work its way through the population,” he says. “It’s not like an apple falling,” where the position of the apple can be predicted at each point during its fall.
The math behind a variable Ro is not complex. In Larson’s models, the number of people a flu sufferer will infect is the number of people he or she encounters multiplied by the probability of infection in the event of an encounter–both of which can be controlled to a great extent. (The probability of infection does have unpredictable, uncontrollable inputs, such as characteristics of the virus itself, but human inputs–such as hygienic procedures and the distance between an infected person’s desk and her coworkers’–can be managed.)
“Our key motivation is that a lot of people think once the flu hits, there’s not much we can do,” says Larson. He is hoping to show that such fatalism is misplaced.
One of Larson’s simplest models, developed with MIT doctoral student Karima Nigmatulina, stratifies a hypothetical population of 300,000 into three groups of 100,000 people who have either a high, medium, or low number of face-to-face encounters each day. In this model, if no one’s behavior changes, the epidemic will peak after 11 generations of the flu (about 33 days); almost all the highly social people become infected, 70 percent of those with medium levels of social activity are infected, and 28 percent of the least sociable people are infected.
If, at any point before this hypothetical epidemic peaks, every individual either chooses or is compelled to halve the number of people he or she encounters each day, the situation changes dramatically: thousands fewer people are infected, even if no one’s behavior changes until the 10th generation (day 30). And the earlier people’s behavior changes, the better. If just the people with the highest level of social activity reduce their interactions by half, but they do it on day 9, only 73.9 percent of them will be infected. And their decrease in social activity has protective effects for everyone else: only 47 percent of those with medium activity levels will be infected, and just 15.6 percent of those with low activity levels will be infected. Although the model doesn’t predict how many infected people will die, fewer infections would mean fewer deaths.
Convincing people to change their behavior early in an epidemic is likely to be a challenge. It’s plain that those with influenza should, if not stay in bed, limit face-to-face contacts and keep their distance when encountering healthy people. But the flu’s 24-hour asymptomatic period complicates matters. When there’s plague in the land, we all have to reduce our social contacts before we feel sick.
Larson fears that without major educational programs and support from the government and employers, people aren’t going to stay home if they feel fine; it’s inconvenient and could even cause economic hardships. But the government and many major employers do not have plans in place to support what are called social-distancing measures–literally putting space between people.
On the individual level, social distancing means “dampening nonessential day-to-day contacts,” Larson explains. Telecommute if you can. Have groceries delivered instead of going to the supermarket. “We can’t contain [influenza] geographically, but we can invoke social distancing aggressively, early, and for a long time,” he says.
Employers and government officials must be prepared to support such measures–even if it might hurt the economy in the short term. For instance, they should develop a more flexible sick-time policy so that employees can protect their health and that of their families without fear of losing their jobs. If possible, offices should open windows and rearrange employees’ desks so that they are farther apart. The government must also be ready to cancel public events. And it should close public schools–and make sure freed high-school kids don’t then congregate at the mall–before the disease spreads through the schools.
Learning from the Past
Some think the 1918 epidemic originated at a military camp in Fort Riley, KS. But Boston, which was a main routing point for American soldiers being sent overseas in World War I, was the major node of the disease’s spread. Two weeks after the virus arrived in Boston’s port, it had infected 2,000 men in the First Naval District, and it quickly spread to civilians.
That September, an unidentified military doctor at a camp outside Boston described the disease. The men at first appeared to suffer from ordinary influenza, but their condition rapidly deteriorated, their faces darkening from lack of oxygen as they suffocated. In late September, he saw about 100 soldiers die every day.
Still, the city of Boston did not cancel or postpone a September “Win the War for Freedom” parade that sent navy sailors and civilian shipyard workers–and the virus–marching through the streets. Later that month in New York City, a day after 324 new cases of influenza had been reported, the city’s health commissioner refused to close schools, theaters, and other public gathering places. It’s as though cities like Boston and New York were practicing reverse social distancing in 1918, says Larson–a good model for what not to do next time.
Evaluating States’ Plans
Larson is working to influence policy at the state level, because the federal government has left it to the states to plan for pandemic influenza. He and co-principal investigator Stan Finkelstein ‘71, a medical doctor and senior research scientist in the engineering systems division, have a grant from the Sloan Foundation to evaluate the plans of all 50 states.
Five of Larson’s students worked full time over the summer scoring each state’s plans on about 60 measures, many of them related to hygiene and social distancing. This fall they will also evaluate the preparedness plans of several universities, including MIT. Like the military barracks of World War I, today’s dorms and lecture halls bring many people together in close quarters, ideal circumstances for accelerating the spread of an epidemic. Yet if all the universities in Boston were to send all their students home at the same time, Logan Airport would be overwhelmed, Larson says.
Larson is not ready to release the states’ scores or rankings, but he says they demonstrate that state governments still have a lot of work to do. “Some of the states seem not to take this very seriously,” he says. “They are going through the motions.”
The plans generally do a very good job of addressing how hospitals will deal with a surge in patients, prioritize antiviral drugs for their staff, and stockpile equipment such as respirators. But most fall short in planning for social distancing. Making matters worse, the existing lines of communication between the states themselves, and between state governments and the private sector, are inadequate, Larson says.
Clear and efficient communication is crucial for monitoring the progress of a pandemic and for guiding public response. “Rather than panic, we want to get good information to the public, to let people know they have control,” says Larson. He notes that Hong Kong did a good job of containing SARS in 2003 because it had the infrastructure in place to disseminate information about risk locations–where and when contagious people had been out in public. Those who might have been exposed to the virus then stayed home and reduced social contacts in case they were sick but not yet symptomatic.
Larson’s model shows that social distancing can significantly mitigate the impact of a pandemic. But when illness strikes, the government has little time for deliberation–about when or whether to close schools, for example, or about who should receive medication that’s in short supply. The longer we spend arguing, the greater the number of people who will be infected, and the more people who will die. Given finite resources, and life-or-death consequences, “a plan including all decisions needs to be worked out before the flu hits,” says Larson.
“Are you scared?” he asks. “You should be.”
DeepMind’s cofounder: Generative AI is just a phase. What’s next is interactive AI.
“This is a profound moment in the history of technology,” says Mustafa Suleyman.
What to know about this autumn’s covid vaccines
New variants will pose a challenge, but early signs suggest the shots will still boost antibody responses.
Human-plus-AI solutions mitigate security threats
With the right human oversight, emerging technologies like artificial intelligence can help keep business and customer data secure
Next slide, please: A brief history of the corporate presentation
From million-dollar slide shows to Steve Jobs’s introduction of the iPhone, a bit of show business never hurt plain old business.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.