CFR vs. PFC: Assessing the true toll of the West African Ebola outbreak, 2013-2015

As the West African Ebola outbreak winds down, it leaves behind many lingering questions. One that piques my interest in particular is, “How many lives were truly lost?”

From publicly available data sources alone via the WHO, it’s possible to construct a simple chart that takes a look at how the proportion of fatal cases (PFC) has changed over time for the three most affected countries in the region (Figure 1):

The distinction between the PFC associated with an outbreak at any given time and the more commonly used case fatality rate (CFR) is as follows:
PFC = N(deaths) ÷ [N(deaths) + N(recoveries) + N(hospitalized)]
CFR = N(deaths) ÷ [N(deaths) + N(recoveries)]
Thus, as the number of hospitalized cases approaches 0 (as it should as an outbreak comes to an end), PFC should approach CFR. (However, it’s worth noting that these terms are often used interchangeably, though they really shouldn’t be.)

As an outbreak progresses, PFC should (in theory) increase and ultimately stabilize at the true CFR – unless there’s a notable change in case management or prevention that results in improved outcomes (or mild case discovery) over time. Here’s an example, which I drew up near the end of the South Korean MERS outbreak (Figure 2):

figure_2This is about as close as you can get to a perfect depiction of what we expect to see – especially when handling real, publicly-sourced (and often very messy) data. It should come as no surprise that this is largely due to the South Korean government’s dedication to transparency (and their ability to actually follow through thanks to adequate resources).

If we compare what we saw with Ebola in West Africa to MERS in South Korea, it becomes quickly apparent that something is amiss. In West Africa, we have been witnessing what seems to be a drop-off in PFC – a highly unlikely outcome given that no “miracle drug” was developed and distributed on a mass scale during the outbreak (and as cases grew, contact-tracing and case-finding activities understandably suffered).

Right now, the overall for all three countries is barely 40% – less than half of what we’ve seen in previous Ebola outbreaks. There are (at least) two reasons this might be happening:

(1) Our follow-up data are incomplete. It’s possible that final outcomes were not successfully reported for some (perhaps many) cases that were initially registered into the WHO database.
(2) The denominator is inflated. Because of the scale of this outbreak, only 53% of the nearly 30,000 cases that have been reported thus far have been lab-confirmed. It’s possible that a large chunk of the remaining 47% weren’t Ebola at all (Figure 3):

figure_3However, it’s also possible that these remaining 47% of cases were Ebola, but they were milder than what we’ve documented in the past. Such cases may not have been reported as Ebola in the absence of such a large outbreak; they may have instead been written off as malaria or dengue – both of which are endemic and exhibit similar symptoms. This brings to light the possibility that previous outbreaks may have been significantly larger than what we filed away on record; however, especially in the early years of Ebola, it is possible that severe cases were the only ones diagnosed. If this is true, the CFRs associated with previous outbreaks may be significantly lower than what we believe them to be due to the artificially deflated denominators.

Nevertheless, if we assume that all fatal cases of Ebola were lab-confirmed, we can use information from Figures 1 and 3 to quickly estimate that the percentage of lab-confirmed cases that have died thus far is ~75%. (Perhaps unsurprisingly, this is a relatively close approximation of the ~67% fatal cases we’ve seen so far in Guinea, where a whopping ~87% of reported cases have been lab-confirmed.)

…So, what will be the true case fatality rate (CFR) associated with the West African Ebola outbreak? The truth is, I’m not quite sure – and I hope it’s now clear why this particular metric is not as cut-and-dry as it often seems.


*I’ve addressed this question before in an article I wrote for the Disease Daily (which was then cross-posted here). In Summary: Back in early September 2014, I used optimization methods to estimate the true CFR associated with the outbreak. At the time, I found that the true CFR was likely much higher than the PFC of 53% that was being reported at the time. The content above remains fairly consistent with my earlier estimate, but uses 13 more months worth of data and explores a few reasons why – despite the fact that the outbreak is now nearing its end – the PFC may still not be a reasonable estimate of the true CFR.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s