After a week of attention-grabbing Ebola headlines, including urgent calls for action from President Barack Obama, it’s clear the international community recognizes the virus as a serious threat.
But how serious a threat is unclear. The dizzying array of projections, including this week’s from the World Health Organization (WHO) and the Center for Disease Control (CDC), places projected cases anywhere between 151,000 and 1.4 million. The radically different estimates complicate everything from assessing the effectiveness of interventions to how to parcel out aid.
The gulf illustrates how small differences in methodology and approach can lead to radically different estimates – and underscores how much of a guessing game the outbreak is. These estimates, which can include a political calculation of how seriously the outbreak is being taken, are what the international community works with as it considers a response.
On Monday, the WHO released its latest projections: 20,600 cases by early November in Sierra Leone, Liberia, and Guinea. The next day, CDC released its own analysis for Sierra Leone and Liberia that includes a worst-case scenario of 21,000 cases by Sept. 30 and 1.4 million by Jan. 20. (Guinea was not included.)
But the 1.4 million CDC figure is the high end. The same report issued Tuesday includes lower estimates of some 8,000 cases by Sept. 30 and 550,000 cases by January.
For lead responders like Doctors Without Borders, the projections – whether 20,000 or 1.4 million – can sometimes seem like needless squabbling if not enough attention is being paid or action being taken.
"This is already an unprecedented Ebola outbreak, we are in unknown waters," argues Charles Gaudry, who heads Doctors Without Borders in Kenya, by e-mail. "We already know we need massive scale-up right now.... Whether the predicted infection level is 20,000 or 1.4 million does not change that.... We need more centers, we need more qualified staff, and we need them right now."
The problem of 'underreporting'
The CDC’s highest projection is based on an assumption in its calculations that for every person listed with Ebola, there are 1.5 more people not listed. This is known as an underreporting factor.
The use of underreporting as a factor “is up for a lot of contention and it’s been widely debated,” says Maia Majumder, a computational epidemiology specialist at HealthMap, a disease research team run out of Boston Children’s Hospital. HealthMap does not include underreporting in its estimates, nor does the WHO. They estimate 14,000 Ebola cases by the end of October, and a maximum of 151,000 cases in January – if the response to the crisis in West Africa completely collapses. (Editor's note: This sentence was edited to clarify the dates of HealthMap's estimates.)
Most Ebola data projections work off only reported cases, though they include a caveat that numbers could be higher due to underreporting, says Stephane Helleringer, a professor of population health at Columbia University in New York. He is concerned that the way CDC uses underreporting to calculate numbers may be incautious.
To be sure, actual underreporting of the virus on the ground in West Africa, apart from debates about statistical methods, is a major concern. In the most affected African nations, people don’t always report cases due to distrust of the health system, the stigma of contraction, or fear of going to Ebola medical centers.
Yet experts say the trouble with including underreporting in projection models is that underreporting is known typically to decline as outbreaks go on.
“If the unreported cases occurred early on but we’ve gotten increasingly better at detecting … the growth becomes a much flatter one,” Dr. Helleringer says, noting that he would guess underreporting is declining in this outbreak. “Typically in the epidemics that’s what happening. Early on, you get a lot of cases undetected and then the surveillance system kind of springs into place.”
Another reason for discrepancies between the WHO and CDC figures are differing estimates of Ebola’s incubation period – the time from infection to showing symptoms.
The CDC’s model assumes this time is a six-day period. But WHO data collected in the past month in West Africa show a 12-day incubation period. In the CDC model, cases multiply much faster, Helleringer explains.
Incentives and estimates
Finally, disease projections numbers are part of a political context, and impact the success and funding of health agencies and programs. So there’s an incentive to estimate high.
“The effectiveness of your response probably will be much larger against the scenario of 1.4 million cases than the 20,000 we’ve predicted,” Helleringer says.
CDC Director Tom Frieden stated clearly this week that high projection rates like 1.4 million are designed to catch the world’s attention and urge greater action.
“Part of the point of having a projection of what might happen if we don't take urgent action is to make sure that it doesn't happen. And that's what we hope and anticipate this will result in,” he said