2.8

## London School of Hygiene & Tropical Medicine

Skip to 0 minutes and 14 secondsJUDITH GLYNN: Ebola causes so much fear because of the high proportion of deaths. In this lecture, we'll look at some of the problems behind measuring fatality rates. Imagine a refugee camp where there's cholera outbreak. There are 100 cases of cholera, and 10 patients die. The case fatality rate will be 10 divided by 100, or 10%. In general, the case fatality rate is the number of deaths divided by the number of cases. In the first Ebola outbreak, 149 cases acquired Ebola by person-person transmission. And 119 of them died. The case fatality rate was therefore 119 divided by 149, or 80%. In the Ebola outbreak in Kikwit in 1995-- 310 cases, 250 of them died.

Skip to 1 minute and 4 secondsCase fatality rate is therefore 250 divided by 310, or 81%. In the West African outbreak, as of 16 of November 2014, there have been more than 15,000 cases, and more than 5,000 deaths. But the case fatality rate is not simply the number of deaths at that point divided by the number of cases at that point. Let's look at why not. Well, first, we have to think who is included as a case, which deaths are Ebola deaths, and how are they included? And how do you actually measure the case fatality rate? Standard case classification criteria for Ebola divide them into suspected cases, probable cases, and confirmed cases.

Skip to 1 minute and 47 secondsThere are several different criteria for suspected cases, including someone with a sudden high fever who's had contact with an Ebola case, or people displaying typical Ebola symptoms which include fever and a range of other symptoms, none of which are specific for Ebola. They can occur in other conditions, including malaria, which is very common in the region, typhoid, and other fevers. Or also any sudden, unexplained death is enough to put someone in the suspected category. A probable case is a suspected case with medical opinion that they're probable, or any person who died from suspected Ebola who has epidemiological contact with a confirmed case. And confirmed cases are those who've got a laboratory test that's positive for Ebola.

Skip to 2 minutes and 36 secondsIf we look at the 15,000 cases mid-November, although more than 9,000 were confirmed, there was quite a lot of probable cases, and 3,600 suspected cases. So we're not clear how many of those 15,000 actually had Ebola apart from those that were definitely confirmed. So are cases being missed, even from the list of susceptibles? Are we undercounting? Well yes, we will be. It's a very large area. There's remote areas where cases are not going to be recognised and reported. And the symptoms are not unique to Ebola so that people may be sick, and it's not recognised. It's also possible that non-Ebola cases are being counted as Ebola. Because, again, the symptoms are not unique to Ebola.

Skip to 3 minutes and 20 secondsAnd in the suspected and probable cases, we're not having a laboratory confirmation. What about the deaths? Are we missing deaths from Ebola? Well, yes. Again, it's a large and remote area. And the symptoms are not unique to Ebola. And also, there's a reluctance to report deaths because of the implications for funeral and follow-up. So people who die and not already on the suspect list may be missed altogether from the figures. It's also possible that people who die from diseases other than Ebola are being counted as Ebola deaths because the symptoms are not unique Ebola.

Skip to 3 minutes and 54 secondsAnd if someone is a suspected case dies and has had contact, they'll become a probable case, so will also be included in those figures without laboratory confirmation. So to estimate the case fatality rate, we really need cases who actually have Ebola. And we need people who die who've actually had Ebola. But that would suggest restricting us to confirmed cases. And there's a problem with doing that. Because if you exclude those who died before confirmation, you will underestimate the case fatality rate. So we can't just look at the confirmed cases to get it. Another problem is that we need to know the outcome for all the cases.

Skip to 4 minutes and 30 secondsIf we look at a particular point in time, at how many cases there've been and how many deaths there've been, many of those cases will still be sick. The outcome won't yet be known. And so, if we simply take the number of deaths at that point divided by the number of cases, we'll underestimate the case fatality rate. So you might think it's easiest to measure what's happening in the treatment centres. We know who's a case. We know who goes on, what the outcomes are, and who dies. But again, the problem with that is you might underestimate the case fatality rate as it would exclude those who die without reaching the centres.

Skip to 5 minutes and 4 secondsFor example, as of September 2014, looking at the cases with known outcomes, if we looked at all cases, the case fatality rate is 71%. But if you looked just in the hospitalised cases, the case fatality rate was 64%. This might suggest that hospitalisation was improving the outcome, and that might be part of the explanation. But it could also be that those who were very sick had died before getting to the treatment centres, then would contribute to this difference.

# A deadly disease: measuring case fatality rates

This lecture looks at the problems in estimating cases, deaths and the case fatality rate in Ebola.

The case fatality rate (CFR) is the proportion of people with a disease who die from it.

CFR = number of deaths / number of cases

In the current Ebola outbreak there are many problems in estimating the number of cases, the number of deaths, and the way they are put together to estimate the CFR.

The case definition for Ebola is divided into suspected, probable and confirmed cases. The full definitions are given by WHO in Annex 1 of the Ebola situation reports:

‘Suspected

• Any person, alive or dead, who has (or had) sudden onset of high fever and had contact with a suspected, probable or confirmed Ebola virus disease (EVD) case, or a dead or sick animal

• OR any person with sudden onset of high fever and at least three of the following symptoms: headache, vomiting, anorexia/ loss of appetite, diarrhoea, lethargy, stomach pain, aching muscles or joints, difficulty swallowing, breathing difficulties, or hiccup

• OR any person with unexplained bleeding

• OR any sudden, unexplained death.

Probable

• Any suspected case evaluated by a clinician

• OR any person who died from ‘suspected’ EVD and had an epidemiological link to a confirmed case but was not tested and did not have laboratory confirmation of the disease.

Confirmed

• A probable or suspected case is classified as confirmed when a sample from that person tests positive for EVD in the laboratory.’

It is clear that the suspected cases will include people who do not have Ebola. The symptoms listed occur in other illnesses, including malaria which is common in the region.

Probable cases will include a higher proportion of true Ebola cases, but note that individuals who die are more likely to be classified as probable, which has implications for calculating the CFR.

While confirmed cases should all have Ebola, confirmation requires a blood test or other sample which usually implies the patient has survived long enough to get to the hospital. Therefore if you calculate the CFR using only confirmed cases this could underestimate the CFR. Similarly an estimate using only hospitalised cases would likely underestimate the CFR. However, if the hospitals are overflowing and only the sickest patients are admitted, the CFR could be overestimated if based only on hospital data.

In addition, cases and deaths will both be missed from the figures, either because they are not recognised as Ebola or because they have avoided the health authorities. There may be mild or even sub-clinical infections: how common this is is not known. Deaths may be deliberately hidden from the authorities due to concerns about interference with burial practices.

It is important not simply to calculate the CFR as the number of reported deaths divided by the number of reported cases on a certain date. This would underestimate the CFR, as many of those cases would still be ill and might yet die.