Skip to main content

Ebola Infections Fewer Than Predicted by Disease Models

Improvements in health care and other uncertainties make accurate forecasts difficult

A few months ago the U.S. Centers for Disease Control and Prevention predicted that up to 1.4 million people in Liberia and Sierra Leone could become infected with Ebola by mid-January. In a recent address to the Senate, CDC director Tom Frieden said that worst-case scenario would not pan out.

That is partly because health care workers in the Ebola hot zone are engaged in a battle to contain the epidemic. It is also because of assumptions about human and viral behavior that are built into the mathematical models used to predict the spread of infectious diseases. Assumptions are inherent in these models. “You take islands of data from different places and build bridges of assumptions that link up all these islands,” says Martin Meltzer, senior health economist at the CDC. Meltzer’s model, which predicted the 1.4 million infections in Liberia and Sierra Leone, worked on the assumption that things would not improve. “Our forecasts are based on the idea that nothing will change,” he says.

But things have changed. About 3,000 U.S. military personnel have been deployed to West Africa since September. They’ve helped build Ebola treatment units and laboratories and train local staff. Alongside them, health care workers from international aid agencies such as Doctors Without Borders and Partners in Health are working with local doctors, nurses and epidemiologists to identify and treat Ebola patients. Meltzer says those interventions were not built into the model because predicting their impact is difficult. “It’s easy for me to say that suddenly 100 [Ebola treatment units] have been built,” he says. “In the model that’s just two lines of code. But the reality of building those units and seeing that they work, that’s much harder.”


On supporting science journalism

If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.


Human behavior is also difficult to predict but education campaigns raising awareness about Ebola and teaching safe burial practices seem to have worked, says Bryan Lewis, a computational epidemiologist at the Virginia Bioinformatics Institute at Virginia Tech. His original estimates suggested that several hundred thousand Liberians could become infected with Ebola. “But now we think we won’t cross the 100,000 mark,” he says. Lewis’s team did not incorporate potential improvements to health care systems or human behavior into their model. “But looking at things now, behavior really is changing and that’s why I think things are improving on the ground,” he notes.

Modelers are forced to build some assumptions into their programs because of a lack of data. That’s especially true at the beginning of an epidemic when efforts to stop the outbreak take precedence over accurate data collection and communication. “Some of the data I was receiving early on wasn’t reliable,” Meltzer says. He worked with CDC staffers on the ground in west Africa to access information such as case counts and death reports. “There were case reports that sat on a desk for some time before they were reported and so the dates of symptom onset were inaccurate,” he says. “You just have to do make do with whatever data you can get.”

Making projections far into the future can also introduce inaccuracies into disease models. “There’s just so much uncertainty,” Lewis says. “Things are always going to be wrong when you look that far ahead.” His team decided to focus on short-term projections instead of looking many months into the future. Those projections were closer to the actual number of cases that have been reported: more than 17,000 people have been infected and more than 6,000 have died since the outbreak began.

Some regions have seen vast improvements in the number of reported cases. Lofa County, once Liberia’s Ebola hotspot, went from reporting hundreds of Ebola cases per week in August to four new cases in the first week of November. But where things can get better they can also get worse. Public health officials have been reluctant to declare these numbers a success because of fears that the virus could surge. "It's like saying your pet tiger is under control," Bruce Aylward, assistant director general for Polio, Emergencies and Country Collaboration at the World Health Organization, told reporters at a news conference in October.

Since then an additional nation, Mali, has seen eight cases of Ebola, six of them fatal, and the virus has continued to spread in Sierra Leone where more than 6,000 people have been infected.

It’s not as bleak a picture as some early models painted but Meltzer believes those projections made people sit up and pay attention. “We were telling policy makers that if we don’t do something, this is what will happen,” he says. “We needed them to know that we could see millions of people infected with Ebola.”

Seema Yasmin is director of the Stanford Health Communication Initiative at Stanford University, where she also teaches science journalism and global health storytelling. She is an Emmy Award–winning reporter and author, medical doctor and frequent contributor to Scientific American.

More by Seema Yasmin