Research Stories

by Diane Boudreau

You want to talk about tough decisions? Talk to Bob England.

On April 29, 2009, the Centers for Disease Control and Prevention (CDC) confirmed Arizona’s first case of swine flu in an 8-year-old child. At the time, no one knew how infectious or deadly this brand-new strain of flu would be. CDC guidelines recommended shutting down any school with a confirmed case. But the CDC recommendation was just that—a recommendation. As director of the Maricopa County Department of Public Health, England was responsible for making the decision on a local level.

Colorized negative stained transmission electron micrograph (TEM) of the H1N1 swine flu virus (Image courtesy of CDC).

Better to be safe than sorry, right? That’s easy to say when you aren’t the one getting angry letters from single parents who can’t work while their kids are at home. Or when you don’t have to wonder how many of those kids depend on school lunches for their nutritional needs. On the other hand, taking action to stem an epidemic only works if you start early. England couldn’t wait to find out how dangerous this virus would become before making a decision.

“If you take action, you have to do it immediately. The only rational thing to do was intervene early to prevent the spread until we could collect enough information to know otherwise,” he says. As a result, he closed three schools early in the outbreak.

Closing schools is one way to combat flu. Other means include vaccines (when available), antiviral medications, and additional social distancing efforts like canceling public events. All of these measures carry economic and social costs, and none of them are 100 percent effective. Public health officials have to weigh the costs and benefits of these options, usually with incomplete information.

These decisions are daunting. If you have a vaccine and the supply is limited, who gets it? If you mount an aggressive antiviral campaign, will the disease become drug resistant? If you close schools, will kids hang out at the mall and continue to spread the virus anyway? And what is the cost of closing schools or businesses or transportation networks in an already fragile economy?

To help answer these questions, England and his colleagues turn to mathematicians who specialize in modeling. Like the computer simulations they run, these experts generally work in the background, out of the media limelight. But their work contributes to decisions that affect every one of us.

When scientists have questions, they usually conduct experiments as a first step toward finding answers. With questions about infectious disease, however, experiments aren’t always possible, practical or ethical. Researchers can’t go around infecting people with flu viruses to test out which interventions work best. So they use models instead. They create abstract virtual worlds where a myriad of scenarios can be tested without hurting anyone.

“We do this through computer and mathematical calculations, which is a lot cheaper than waiting through three epidemics and seeing which approach works best,” says Carlos Castillo-Chavez, director of ASU’s Mathematical, Computational and Modeling Sciences Center. “Modelers carry out experiments in a virtual world that we can’t perform in the real world.”

Flu by numbers
Modelers use two quantitative methods: statistics and dynamics. Statistics describe quantities and their distributions. For instance, scientists use statistics to express the likelihood of dying from a particular disease. A key number in studying disease outbreaks is the reproduction number. This number describes how fast a disease spreads through a population. A disease that spreads quickly and kills many will warrant more extreme and costly containment measures than a less aggressive virus.

Modelers use statistics to estimate the reproduction number and its distribution in order to evaluate the effectiveness of prevention measures. “We don’t test these measures with a statistical model. We use models that study disease dynamics,” says Castillo-Chavez.

Dynamic models incorporate statistical information, but they don’t just regurgitate data. They simulate what happens over time and build a realistic virtual world where the effectiveness of future interventions can be evaluated, using limited information from past or ongoing outbreaks.

Dynamic models usually rely on differential equations. A differential equation states how a rate of change in one variable is related to other variables. For example, suppose you are taking a trip from Phoenix to Tucson, and you want to know what time you will arrive. If you hop on a magic carpet and fly there at the same speed the entire way, you can figure out the answer easily. But if, like most of us, you drive a very non-magical car, you will change your speed often due to traffic lights, varying speed limits, and other factors. To figure out how long your trip will take in light of these obstacles, you need to solve a differential equation.

“With disease, we might look at how many cases there are on June 1, June 2, June 3, etc. From all that information you want to determine how many people will become infected over time, knowing that the number of susceptible people goes down as more people get infected, or knowing behaviors that will change infection rates,” explains Castillo-Chavez.

You could also use a differential equation model to study how different school closure options would affect the spread of a disease. In fact, ASU researchers have done exactly that.

Decision Central
In February 2009, a group of people gathered in a small, windowless room in the heart of downtown Tempe. The room, surrounded on all sides with projection screens, is the “drum” of ASU’s Decision Theater. The people were health officials (including England) and public school officials. They came together to simulate a flu pandemic situation and evaluate school closure options.

County education and public school officals in Arizona respond to scenarios in a pandemic influenza exercise in February 2009.

Unbeknownst to the group, at that moment H1N1 swine flu was making its secretive debut, most likely in the Mexican town of La Gloria. At the time, many people assumed that the H5N1 avian flu would cause the next pandemic.

The core of the simulation exercise was a dynamic model that simulated the spread of the H5N1 virus through space and time and across different age groups. Overlaid on top of the model was a set of fake news reports designed to mimic a real-life media response. Participants were asked to make decisions at key points throughout the exercise. The model then simulated the effects of those decisions.

Unlike in real life, however, the participants could go back and change their decisions and then compare outcomes in order to determine the best course of action.

“Mathematical models help us to map out the likelihood of certain possibilities and allow us to make decisions using a wider set of information,” explains Castillo-Chavez. “But by no means will they tell us what is going to happen. They tell us what is likely to happen under certain circumstances.”

It seems like a stroke of good fortune that ASU hosted a flu pandemic exercise right before an actual pandemic broke out. But Tim Lant, research director for Decision Theater, says it was actually an example of good prep work.

“When these situations emerge, they emerge as crises. But we can only know what to do if we prepare for them in advance,” he explains. “There are a lot of other situations we take the same approach with. If done correctly, it will appear that we were lucky, but actually we were just preparing ourselves well.”

For example, he says that ASU conducted a flu pandemic exercise in 2007 dealing with university closures. Ultimately, ASU did not shut down as a result of swine flu. But university officials were prepared in the event that it became necessary.

England, on the other hand, did close a total of four schools due to the swine flu outbreak. Three of these closures happened shortly after the virus broke out, before anyone knew how bad it would be.

But England says it became rapidly apparent that the disease was more widespread than anyone had realized, and that there were probably a lot more cases in schools than the ones that had been confirmed.

“We also knew that most of the people weren’t very sick,” he adds. “As soon as we had enough information to know that it wasn’t a grave threat, we treated it like regular flu.” That meant backing off the CDC’s school closure guidelines.

“We were the first community in the country to back off on the CDC recommendations,” says England. “The CDC changed their recommendations a few days later, but those few days made a difference. I’d have had 20 to 30 schools closed by that point.”

He did, however, close one more school even after realizing that the epidemic was mild. That particular school jumped from a 2 percent absenteeism rate on a Friday to 20 percent the following Monday. In two grades, fully half the students were absent.

“We talked to school officials and made a joint decision to close the school with that level of illness,” he says.

England notes that the Decision Theater exercise played a role in the real-life school closure decisions he faced just two months later. “It reinforced some of the insight we had before and gave us some new insight as well.”

For example, the simulation highlighted the risk of reopening closed schools too early. England says he took this risk into consideration when deciding when to reopen schools. “We knew this was going to be a pandemic, from the first days,” he says. “We knew we were going to need a lot more help from these guys [at ASU]. We were making decisions without much real information. Healthcare providers everywhere do that all the time. You make decisions based on best guesses and your best opinion. But it’s a bigger deal when you’re doing it for 4 million people at once.

“It’s a little scary going against the CDC, but we thought it was the right thing to do,” he continues. “The upshot is we did the right things at the right time. I’m confident of that.”

Read more about the nature of flu pandemics in "Lessons from pandemics past."

For more information about the Decision Theater school closure exercise, visit: http://asunews.asu.edu/20090225_pandemic