Ebola, Zika modelers aim to inform policy decisions

March / April 2016 | Volume 15, Issue 2

By Karin Zeitvogel

​Mathematical modelers face numerous challenges as they try to predict the course of epidemics, such as Ebola and Zika. Their forecasts can inform policymakers' decisions on how to most effectively deploy resources to contain and manage the outbreaks.

Risk of local Zika transmission

Heat map of Americas shows how Zika virus could spread from South America to the Caribbean, Mexico and the US
Courtesy of Dr. Kamran Khan et al

Source: Anticipating the international spread of Zika virus from Brazil,
The Lancet, January 14, 2016 (online)

To discuss these complex issues, Fogarty recently convened a meeting of disease modelers and U.S. government officials. They considered how to improve models to provide actionable information early on in outbreaks, when the right interventions can be critical, and foster closer collaboration between government and academia. In addition, they reviewed the results of an Ebola modeling exercise.

The difficulty of predicting the course of the current Zika outbreak was examined, but just as at the start of the Ebola pandemic, the paucity of data on the mosquito-borne viral illness that has swept Brazil and traveled to at least 13 other countries makes modeling difficult.

"We don't have many good previous Zika outbreaks to work with," noted Dr. Lone Simonsen, a research professor at George Washington University's Department of Global Health, in an interview conducted on the sidelines of the meeting. "We don't know the link between Zika and microcephaly. We don't know the rates of how many pregnant women have this problem. We don't know the duration of immunity. If you get infected in childhood does that mean that it protects you when you are a pregnant woman one day?" she asked, calling for more case control studies to try to assess if there is a link between Zika and microcephaly.

But the lack of data should not prevent modelers from trying to understand the virus, she suggested. "Modeling is a thinking tool. We can say the link between Zika and microcephaly is real, can say one in 10 pregnant women are going to be affected by this if they are infected, and can use different incidence intervals and come up with different predictions."

By working through the process, scientists will better understand which pieces of information are needed to answer the important questions. And that can be useful now, Simonsen said.

The scientists also reviewed what they learned from an Ebola modeling exercise, which Fogarty helped to organize, under the Research and Policy for Infectious Disease Dynamics (RAPIDD) program. Eight teams from U.S., U.K. and Canadian universities and several U.S. government agencies were tasked last year with predicting when Ebola would peak in Liberia and how it would progress between September and December 2015. Each team had been given four different Ebola scenarios to model, each with different levels of containment. The results were presented and compared at the two-day forum.

Teams were free to choose the type of model to use and the parameters they would add to their model to "interrogate the reality" of the Ebola outbreak in Liberia, as one modeler noted. For instance, the opening of Ebola Treatment Units and the deployment of healthcare workers around Liberia - both of which played a key role in bringing the epidemic under control - were implicitly modeled by lowering the transmission rate parameter, said Fogarty's Dr. Cecile Viboud, one of the challenge's organizers.

As more data became available, all of the models became more accurate. Most accurately predicted when the epidemic would peak in Liberia. Even the weakest model held useful information, showing that transmission is not exponential at the beginning of an outbreak.

But the lead of the CDC's Health Economics and Modeling Unit (HEMU), Dr. Martin Meltzer, said being able to forecast accurately with ample data is not enough. "By the time you're getting more accurate as measured by these metrics, [policymakers] have already spent money, made decisions and deployed resources," Meltzer told the
gathering.

"It's nice to know it gets more accurate when you have more data, but is it good enough at the beginning so that policymakers can rely sufficiently on it to say this is where it's generally pointing if you don't do anything, and this is where it will go if you do X, Y and Z?" he said.

It was Meltzer's warning in September 2014 that up to 1.4 million people in West Africa could be infected by Ebola within months if nothing was done to stop the disease that spurred the global health community to take action. He noted that many of the policy decisions that have been made in the Ebola or current Zika outbreak have been made early on.

As modelers work with the little data available and try to cut through the fog of war that comes with epidemics, they need to keep in mind that policymakers need "actionable" information from models, a White House official told the meeting.

But Meltzer said models need to show more than just what can be done. "I think for a model to be useful to policymakers, it has to show the impact of different interventions, both individually and as a sum," he suggested. "Very often, in fact, we've found through one of our analyses that the combined effect of the interventions is greater than the sum."

More Information

To view Adobe PDF files, download current, free accessible plug-ins from Adobe's website.

Footer