The 2020 Novel Coronavirus Outbreak
The Estimated Novel Coronavirus Case-Fatality Rate Is Not Good
Why the deceptively good 1% case-fatality rate of novel coronavirus is no reason for optimism.
My previous post is highly related to this post: Why the mortality rate of novel coronavirus is miscalculated, and not important.
A new study just came out, and I’m sure it’s going to be published in the media soon.
It will state that the mortality rate of novel coronavirus is:
- 18% for severe cases.
- 1–5% for mild to severe cases.
- 1% in total.
On the surface, it looks reassuring for the general public. Since most people are not infected yet, they will probably avoid the “severe” category and put themselves in the “total” category, concluding that their probability of dying is 1% if they are infected.
Summary of the Study
Problems with case-fatality rate
Before calculating the case-fatality rate (which people have mistaken for the mortality rate), the study identifies the typical problems:
- Surveillance often detects severe cases first. Asymptomatic infections don’t make people seek treatment, and mild cases won’t be prioritized. This leads to overestimate of CFR, as severe cases are more likely to lead to death.
- Lag-time problem: deaths over cases assume that all currently alive people will recover, which is not necessarily true. This leads to underestimate of CFR.
The study deals with the surveillance problem by creating different estimates for different populations, where some estimates include only severe cases, and others include mild cases as well.
It deals with the lag-time problem by estimating the time from onset of symptoms until death, using 26 early cases, and incorporating this information into the CFR estimate.
First, we have to acknowledge the massive uncertainty of the estimated CFRs. Intervals are given for the most likely range of values. A wider interval should indicate more uncertainty in the figure.
(Note:  contains a 95% confidence or credible interval.)
Importantly, there are still sources of uncertainty which cannot be incorporated into the estimates.
- The researchers do not know the sensitivity of the diagnostic tests (proportion of infected people who return a positive result). There is already evidence of symptomatic patients returning false negative results, and there was one patient who only tested positive on the 4th attempt, after using a more intrusive procedure.
I could go on and on, but I have already mentioned them in my previous posts. Overall, we should keep in mind that these numbers are more like preliminary estimates to help policy makers act under uncertainty, rather than final estimates for assessing the outcome.
(1) Estimate for Hubei
- Population: people who are in Hubei (province of China, includes Wuhan). Data from public reports, up to Feb 5.
- Likely range of cases: symptomatic, severe cases, as diagnostic tools are in short supply.
- Estimate: 18% [11–81%].
(2) Travellers (International Detection)
- Population: publicly reported data (up to Feb 8) for travellers from China, who were diagnosed in other countries.
- Likely range of cases: mild to severe cases, as screening is more stringent.
- Estimates: 5.1% [1.1–38%], 5.6% [2.0–85%], 1.2% [0.9–26%].
(3) All infections
- Likely range of cases: asymptomatic to severe cases.
- Estimate: 0.9% [0.5–4.0%], 0.8% [0.4–3.0%]
- Calculated by scaling the previous estimates to account for undetected cases.
- Scaling factor is estimated from looking at the number of infections detected on four repatriation flights.
Considerations following from previous post, given new data.
My opinion, expressed in my last post, is that the spread of novel coronavirus is far more important than case-fatality in determining the severity of the outbreak. This is because a widespread outbreak will cause hospital systems to run out of capacity.
The spread problem: 1% overall CFR comes with many, many undetected virus carriers
It is tempting to say that the outbreak is not severe, because the estimate for the CFR of all infections is estimated to be 1%, or at least something seemingly low. Someone might even say that it is only half as severe as initially claimed, because they saw that it was 2% before. But what they don’t know is that in the process of arriving at this 1%, the researchers estimated that the number of reported cases was underestimated by a factor of 19. They arrived at this figure because they estimated 24,000 infections (95% CI: 13,000–44,000) having onset of symptoms, and only 1,242 of these being confirmed 4 days later, where 24,000/1,242 ~= 19. If this is the case, then it would imply that there is a large number of undetected cases which we aren’t accounting for. So we now have more people inadvertently spreading the virus across borders and into new communities. What happens if the virus spreads so easily and we can’t contain it?
Lack of medical supplies probably contributed greatly to higher CFR
The hospitals in Hubei were overwhelmed by patient numbers very quickly, hence they ran out of medical supplies and could not provide their usual quality of care. The study itself notes that CFR will vary depending on the “clinical care offered to severely ill cases”. Now it is clear why the 18% Hubei estimate is significantly higher, even when considering uncertainty ranges. Wuhan doctors have already noted patient deaths that could have been prevented if certain options such as ECMO (extracorpreal membrane oxygenation — a procedure that basically replaces the lung) were available.
We should not be assuming that 18% is for Hubei, and 1% is for everyone else. If anything, these estimates are a warning. Countries which cannot contain an outbreak will run out of resources and experience the 18% CFR (and probably even more, as Wuhan is better resourced than most places, and has the full backing of the central government and the 2nd largest economy). Only a well-resourced, well-prepared country with strong containment policies will experience the lower rates of 1–5%.
Thanks for reading. Your thoughts and feedback are appreciated. Send them in a response below, or on Twitter at @newapollo__ (link).