What the tip of the corona iceberg tells us (March 15th 2020 update)
Today I have invited a medical doctor, Thomas Grünfeld, MD MPH, to co-author my blog, recognizing that a multi-disciplinary approach is needed.
As known by many, the lion´s share of an iceberg is below the surface and thus can´t be seen. Normally 90%. Knowing this ratio, the size of the iceberg can be estimated.
A similar approach may possibly be used to assess the number of corona-infected by country and region. Such a route is needed as the testing procedures are inconsistent and most often scattered.
This blog expands on the methodology, sum up the results and extract some implications. What we find is scary, especially for the US.
1. The methodology
Simplistically a certain numbers get infected Day D. Over the next week or so the sickness evolves. Evidence indicates that the symptoms vary widely. Some develop very severe conditions, whereof a certain ratio die.
The mortality rates vary depending on age, health conditions, social conditions, genetic background, quality of treatment and the timing of treatment. Still, here we assume a similar ratio when comparing countries.
Assessment of prevalence of disease based on tests is highly unreliable, due to strong variations in test frequency and practices on who to test. However, mortality data should be more consistent across regions.
With access to data on fatalities and on the mortality rate (fatalities/infected) one may “backcalculate”, and estimate the number of infected by region, at least where the sample is of some size.
One though also has to adjust the timing, i.e. how the number of infected may have increased since Day-D.
We call this the tip of the iceberg approach.
2. The data
Chart 1 shows the ratio between persons tested and infected in each country and the ratio of fatalities per 100 infected.
Some countries have launched a system of broad-based testing, most notably Russia, Turkey, Vietnam and Hong Kong.
A number of countries, e.g. Netherlands, Australia, Sweden and Norway, currently only tests the ones with clear symptoms and/or that are considered most vulnerable. The UK, as the main rule, only tests those who have been hospitalized.
The Netherlands has, interestingly, tested randomly (March 10th) all employees at hospitals in a region (Brabant), finding 4% of the staff to be infected. If a similar ratio is assumed for all hospital employees, that would imply 5000 infected hospital employees. This is 5 times the “official” number of infected for the entire country. The Brabant-region was, however, among the most infected regions in the Netherlands.
Analytically we split the countries in three categories; 1) “The testers”, 2) “The samplers”, and 3) By indication only.
One may safely assume that the number of infected exceeds the official number with a vast margin, especially for the category 2 and 3 countries.
The UK estimates the actual number of infected in the UK to be 12 times the official figure.
Based on the above one may assume 1) that the number of infected are in general way above official estimates, and 2) that the mortality, measured as fatalities relative to infected, is way below the 3,4% figure quoted by the WHO.
Some numbers are though needed. Given the size and coverage of the South Korean data (more than 250 000 tests), and with South Korea being among the countries first exposed, we will use the South Korea ratios in our following analysis.
4. The country figures
We believe it is fair to assume that number of deaths from Covid 19 is fairly reliable, as we assume that seriously ill patients are tested. Hence, this indicator is the most reliable “tip of the iceberg- indicator”.
We know for sure that there are huge “unknowns” with regards to the unrecorded numbers of infected cases.
However, the ratio of deaths to reported infected patients gives us an idea of the number of unrecorded cases, especially when taking the build-up of the number of cases over time into account.
This again gives us an indication of the “contagiousness reservoir”, as symptom free, but infected persons are more likely to interact socially and thus infect others.
This is then also a key rationale for widespread testing: To identify the contagious sources.
5. The countries that stand out
The implications of the findings based on these assumptions are severe for the countries with high death figures relative to the registered number of infections.
These countries most likely have substantial (unidentified) contagious reservoirs. The ratio for Italy stands out, indicating that the number of infected is in the hundreds of thousands, as opposed to the quoted figure of about 20000.
Other countries that stand out are the US, Switzerland, France and Japan.
The US holds other conditions that also are most worrisome: A very fragmented health care system, with a large number of the population uninsured, which is likely to limit the testing activity (also possibly leading to Covid 19 deaths remaining unrecorded).
The reported number of tests conducted in the US is about 10 000, about the same as Norway, which has 1/60 of the population. 50 deaths have been reported in the US (As of morning March 14).
6. Taking timing into account
The picture gets even more worrisome if timing is taken into account. Assuming a (high) mortality rate of 1% implies that there were 5000 infected two weeks ago in the US. Assuming a daily growth rate of 30%, below the rate seen in other countries, this would imply that the prevalence is about 150 000 infected today, if not substantially higher, spread out across the US.
With very limited overview of infections, these numbers are likely to continue to grow exponentially unless stronger restrictions are introduced.
7. Comparing what´s needed with the March 13th plan
Our work on this blog started as Thomas and I discussed whether the March 13th plan of Donald Trump meets the needs that will emerge the next weeks.
The answer is “No”. Hopefully we are wrong, but the response appears to be too little too late. A ramp-up of testing could have worked two weeks ago, but not to meet the challenges that will emerge over the next weeks as thousands of hospital beds will be needed.
* * *
If the “tip of the iceberg” assumption is correct, far more proactive measures need to be made in the US.
For the economy, it is a question of “biting a small bullet” now or taking a big bullet some weeks down the road (in a few weeks).