Various fatality rates have been mentioned in connection with Ebola. WHO in its fact sheet on the disease
(http://www.who.int/mediacentre/factsheets/fs103/en/ ) reports a CFR in previous outbreaks in Africa of up to 90% and on its main Ebola page (http://www.who.int/csr/disease/ebola/en/ ) implies the CFR in this West African outbreak is 53%. It does not give the method by which it arrives at this figure. The Sierra Leone Ministry of Health and Sanitation, on the other hand, is reporting a CFR of 37.7%.
The Case Fatality Rate is defined as the percentage of Ebola patients who actually die of the disease. Simply, if the total number of Ebola patients is N1 and the number of Ebola patients who die from the disease is N2, then:-
Case Fatality Rate, CFR =
Determining the CFR is not as straightforward as it might at first seem. One has to first define who is an Ebola patient and who is not. This causes much difficulty. The hospitals and clinics see a large number of patients who could be Ebola patients. Ebola ingeniously disguises itself with the symptoms of many other commonly occurring diseases, including malaria, cholera, typhoid, influenza etc. The only certain way to diagnose it is to send a blood sample for a sophisticated laboratory test.
This test can only be carried out by a few laboratories and even test results are subject to some uncertainty, but for the purposes of this article we shall ignore totally this aspect of the problem. So, if we assume that Ebola is only conclusive when the laboratory test returns positive, do we in our calculation of CFR include only those patients who have undergone a lab test? Some patients never make it to a hospital or clinic. Particularly in this outbreak, some, probably many, Ebola patients have died unrecorded in their communities and have been quietly buried there. In the normal flow of village life deaths occur routinely and are often not diagnosed by medical personnel or reported.
Some patients are taken to a hospital or clinic almost at their last gasp, and they expire before there is an opportunity to observe or test. In the frantic congestion of a national epidemic few send blood samples of a dead man for testing. Some patients are admitted to an isolation ward for observation and they quickly recover (Ebola has, in at least a few cases, been reported to produce no more than a headache, even when lab tests conclusively detect the virus), without need for a test. In order to determine the true CFR these patients should be included in our numbers, but they go unrecorded. Survivors who are not recorded lead to an overestimation of the CFR, whilst Ebola deaths that are not recorded lead to an underestimation of the CFR or the true severity of the disease.
In addition to “genuine” uncertainties caused by the nature of the outbreak and the disease, there are “illegitimate” uncertainties caused by poor recordkeeping and perhaps even a deliberate desire in the early stage of the outbreak to underreport and hide the true extent of the problem.
One must also consider whether different CFR’s could or indeed should be calculated. What is the CFR of hospitalized patients? What is the CFR of non-hospitalized patients? What is the CFR of one facility, perhaps MSF’s Ebola Management Center, as the organization is careful to call it in its literature? What is the CFR of another type of facility, perhaps Kenema’s Ebola treatment center?
What is the CFR of patients who present as possible Ebola cases, but who upon testing are declared Ebola negative? The answers to all these questions are critical if we are to choose the best response to this outbreak. We should attempt to answer these questions in order to effectively respond to Ebola. Some answers might be easy to get. Some might be difficult, if not impossible. In general it should be easier to get reliable statistics for institutions than for the general public IF the institutions have been keeping accurate records.
The Sierra Leone Ministry of Health, in its most recent (August 28) Facebook page gives the following statistics:
Total Survived and Released Patients = 244
Cumulative confirmed deaths = 372
Cumulative confirmed cases = 988
Now adding the survived and released patients (assuming they were all laboratory confirmed) to the cumulative deaths we get 616, far less than the 988 cumulative confirmed cases. What has happened to the 372 cases that are not listed as having survived or died? Some could still be in hospital, but we know that until recently there were only two treatment centers, at Kailahun and Kenema, and the number of patients and beds there was far fewer than this. It is possible, even probable that the Ministry of Health itself has no idea what has happened to these patients. So all we can say from the Ministry of Health’s figures is that in this set of patients (WHICH MAY OR MAY NOT BE REPRESENTATIVE OF THE NATIONAL SITUATION):
Minimum CFR = (372/988)*100 = 37.7%, making the unlikely assumption that all unaccounted patients have survived.
Maximum CFR = (744/988)*100 = 75.3% making the unlikely assumption that all unaccounted patients have died.
OUR MINISTRY OF HEALTH USES THESE FIGURES TO ARRIVE AT A PRECISE CFR OF 37.7%. THIS IS MATHEMATICALLY INCORRECT, MISLEADING AND POSSIBLY EVEN DANGEROUS TO THE EXTENT THAT IT LEADS TO AN UNDERESTIMATION OF THE SEVERITY OF THE DISEASE. IN ANY SET OF PATIENTS ONE CAN ONLY ARRIVE AT A PRECISE FIGURE FOR CFR IF ALL THE OUTCOMES OF THOSE PATIENTS ARE KNOWN. THIS MATHEMATICAL PRINCIPLE IS GENERALLY TRUE FOR ALL SUCH RATIOS OR PERCENTAGES. I would suggest that a competent statistician from Statistics Sierra Leone join the Ministry of Health’s data collection unit.
Thus, all these figures from the Ministry of Health tell us is that in this set of patients the CFR was somewhere between 37.7% and 75.3%. Since these are confirmed cases, their blood samples were taken and thus they presented at a health facility. One could assume that many of them would have been hospitalized, particularly those whose outcomes are known (died or survived), but some may have declined hospitalization, particularly those whose outcomes are not known. From the Ministry’s situation report of August 28 (http://health.gov.sl/wp-content/uploads/2014/08/Ebola-Situation-Report_Vol-92.pdf ) we know that a total of 684 patients have been admitted at the two Ebola treatment (case management) centers in Sierra Leone since the start of the outbreak. Therefore this data includes some 304 patients who were not hospitalized.
MSF in its web page of August 27 ( http://www.msf.org/article/liberia-msf’s-new-ebola-management-centres-already-overwhelmed ) says: “MSF started its Ebola intervention in West Africa in March 2014 and is now present in Guinea, Liberia, Nigeria and Sierra Leone. The organisation runs five Ebola case management centres with a total capacity of 415 beds. Since March, MSF has admitted a total of 1,885 patients. 907 of these patients tested positive for Ebola, of whom 170 have recovered. MSF has deployed 184 international staff to the region and employs 1,800 nationally hired personnel.”
Here the organization has lumped together the statistics for all affected countries. Once again we are faced with some uncertainty. 170 patients have recovered but we are not told either how many have died or how many are still hospitalized. We know the capacity of 415 beds is as a result of very recent additions to capacity. All we can say from this is that in this set of patients:
Maximum CFR = (737/907) * 100 = 81.3%, assuming all those unaccounted have died. However, in practice since some of those unaccounted would still have been admitted in the MSF facilities the actual CFR would be less than this.
There is one other important piece of information here: the organization admitted a total of 1,885 patients, who must have had symproms consistent with Ebola. Thus, approximately 52% of patients who presented with symptoms consistent with Ebola did not according to the lab test have Ebola. It would be interesting to find out the outcomes of those cases.
In its web page of August 15, (http://www.msf.org/article/international-response-west-africa-ebola-epidemic-dangerously-inadequate ), the organization breaks its statistics down by country. Finally here we get information that is almost complete.
For Sierra Leone:
204 confirmed cases
50 patients in center.
Therefore, assuming the patients who have not recovered and are not still in the center have all died (the only other option is that they left the hospital alive without being certified as recovered, which is unlikely) and discarding the 50 patients in the center whose eventual outcomes are not known, we have 102 deaths from 154 confirmed Ebola patients all of whose outcomes are known:
Sierra Leone MSF CFR = (102/154)*100 = 66.2%
For Guinea Conakry:
80 confirmed cases ***inconsistent with MSF August 8
50 recovered ***inconsistent with MSF August 8
4 patients in treatment center
Guinea Conakry MSF CFR = (26/76)*100 =34.2%
For Guinea Guekedou:
197 confirmed cases
11 in treatment center
Guinea Guekedou MSF CFR = (141/186)*100 = 75.8%
From the above we see that the Sierra Leone MSF CFR is not very different from the Guinea Guekedou MSF CFR. These numbers are also not very different from the maximum CFR given above from the Sierra Leone Ministry of Health figures, which also included patients not hospitalized. We are told specifically that all the MSF cases were admitted at MSF facilities. The Guinea Conakry figure is much lower, but is based on a smaller sample and there are also inconsistencies with earlier figures.
The Sierra Leone Ministry of Health and Sanitation gives figures (http://health.gov.sl/wp-content/uploads/2014/08/Ebola-Situation-Report_Vol-92.pdf ) for Ebola patients at the two Ebola facilities in Sierra Leone, at Kenema and Kailahun. The Kenema facility is run by Ministry of Health staff separately from the MSF Kailahun Case Management center, so it would be interesting to compare CFRs for these two different facilities.
For the Kenema facility:
Cumulative admission 390 (Note: the data does not indicate whether these were all lab confirmed with Ebola)
Cumulative discharged 175
Currently admitted 65
Cumulative deaths 162
The discharged, currently admitted and deaths exceed the admissions, which indicates a discrepancy in the data, albeit relatively small. Using the data as given, for this set of patients:
Kenema Ebola Center CFR = (162/325)*100 = 49.8%
For the Kailahun MSF facility, using the Ministry of Health data
Cumulative admissions 294
Cumulative discharged 69
Currently admitted Not indicated
Cumulative deaths 180
Assuming all admissions are accounted for, as they should be in a well run facility we can conclude that currently admitted are 294-69-180 = 45. This is close to the figure given by MSF itself above (50), although the dates are different. For this set of patients:
Kailahun MSF CFR = (180/249)*100 =72.3%
This is fairly close to the 66% figure obtained above using data from the organization’s web page of August 15.
Conclusions: Using all the available data from the two Ebola facilities in Sierra Leone we get CFRs that range between 50% and 72%. This is considerably higher than the 37.7% CFR that our Ministry of Health is reporting. The method used by MOH is mathematically incorrect and leads to an underestimation of CFR.
By Paul Conton, MSEE
Tuesday August 02, 2014