Rough Analysis of MERS-CoV case data to date

This analysis is done using the case data presented on this page. The same caveats on that page applies - the data is rough, gleaned from various sources both official and not and contains errors and omissions. The conclusions here are conditional on these data and should not be consider anything but indicative of possible trends. The tables and charts are now live and will reflect the data as it is updated although the text does not change automatically.

This page was prompted by some discussion by Helen Branswell [@HelenBranswell], Ian Mackay [@MackayIM], [@FluTrackers] and others on Twitter. It was also motivated by two, widely made, observations. Firstly, that most severe cases have occurred in people who are older and more male. Secondly, health care workers (HCW) are commonly amongst the known secondary cases (cases with known contact with primary cases) but seem to not have such severe disease and fewer fatal outcomes. The aim of this posting is to look at the numbers and see what we can infer about the epidemic. Ian Mackay has posted some analysis on his blog but I thought I would try and dig a bit deeper into the numbers.

I will consider the 405 laboratory confirmed cases for which I have data. Of these, 109 (27%) have died. My list is missing some deaths because of insufficient information to assign some notices of deaths to known cases.

For 251 cases that are not known to be secondary there are 77 deaths (31%) whereas for known secondary cases this drops to 32 out of 154 (21%):

Thus there is a significant difference in outcome between these two groups. Because most of the cases with reported comorbidity are primary infections the numbers are more striking if we divide the data set by that:

So there is a significant difference in outcome between those with existing comorbidity and those without. 

So are HCWs less likely to have a fatal outcome than the other occupations? The simple numbers suggest yes:

However, the other cases include most of the cases with existing chronic illness. Most of the HCWs are recorded as having had contact with other confirmed cases so a better question to ask is whether HCWs are less likely to suffer fatal outcome than other known secondary contacts? Here are the outcomes for HCWs vs family and other (non-specified) contacts (excluding those cases with known comorbidity).

The statistical test suggests there is no significant difference in outcome between these two groups. 'Other contacts' could include HCWs that are not identified as such. So we can compare known family contacts with HCWs.

Again there is no significant difference in outcome between these two groups. The numbers are getting quite small here so lack of power may be an issue. 

Finally, we can look at all secondary cases at whether comorbidity affects outcome. This would better account for ascertainment bias towards severe cases (although it won't eliminate it):

So the better outcome for secondary patients (and HCWs) is likely due to the fact that fewer have existing chronic conditions. 


What I think this means is the cases are a mixture of different exposure risk groups and different outcome risk factors. It seems likely that the secondary cases represent the likely disease outcome in the general population better than the primary cases. Indeed for secondary cases that do have known existing comorbidity, the outcome is poorer with 14 dead out of 20 (only one HCW is included in this group but that probably reflects that working HCWs are likely to be younger and healthy).

The first thing to consider is that HCW are obviously the most likely to be in contact with severely infected individuals and are clearly at high risk of infection. Severe cases may be shedding viruses at higher rates than mild or asymptomatic cases that are controlling the viruses better. Individuals with existing medical conditions may be more likely to be infected or may be more likely to have a severe outcome and thus be counted. Thus far, all mild and asymptomatic cases have been found by through contact with a known case. Thus primary cases are biased towards being severe and secondary cases are probably more representative of the likely outcome for an average person. However, HCWs are going to consist of working-age adults without severe existing morbidity.

Viruses don't tend to infect one gender more than the other but exposure or risk-factor may be be gender-biased. The male bias in the severe primary cases may reflect a bias in the likelihood of having conditions that either confer susceptibility or increase likelihood of severity rather than the male gender being at higher risk. I don't have data for KSA but in many countries the health workforce is female biased with women comprising around 75% [REF]. This is pretty much the gender ratio for HCW MERS cases. The other secondary cases have a more balanced gender ratio. In this case it is not being female or a HCW that is the risk factor but exposure to severe cases and severe cases are in hospitals. In the Al-Hasa outbreak most of the secondary cases were other patients (with a few visitors and HCW as well) showing that having chronic illness and being exposed to severe cases is the worse case. 

It seems unlikely that people with chronic illness have a greater exposure to a source of infection than the wider public so either they are at greater risk of infection per exposure or it is only the severe cases that are being diagnosed and there is a large tail of mild and asymptomatic cases. ​If the latter then the case fatality rate may be much lower than the 50% currently observed for all patients. Perhaps closer to the 20% observed for secondary cases without comorbidity but even this will be biased against asymptomatic cases. 

I am not aware of any definite cases of mild or asymptomatic cases transmitting the virus further. It is possible that such cases have a low probability of onwards transmission.