So I was looking at Worldometer today and the top ten countries at death per million (excluding San Marino, Andorra, and Saint Martin, which seem like outliers based on the size of the country are):
Belgium (490)
Spain (437)
Italy (391)
France (302)
UK (237)
Netherlands (215)
Switzerland (161)
Sweden (152)
Ireland (124)
USA (123)
We’ve (obviously) had plenty of news coverage of the situations in Italy, France, and Spain. And no shortage of discussion on the US and UK. (And a thread dedicated to Sweden).
But, what the heck is going on in Belgium? This article suggests that Belgium is using a more permissive reporting system (and runs the risk of inaccurately over-reporting). I know that a number of jurisdictions have been starting to report “probable” deaths as well as “confirmed” deaths. Is it just that Belgium has been doing that all along and the numbers will even out in the long run?
I suppose, since Belgium sits geographically between France (#4 deaths per capita) and the Netherlands (#6 deaths per capita), you would expect it to be bad there. But if I’ve done my math right (never guaranteed), the CFR for Belgium isn’t materially higher than France or the Netherlands: Belgium (14%); France and Netherlands (~12%); UK (13%) – wait a minute, is the mortality rate really that high? But (same math) it’s 3% in Germany? Belgium looks like it’s done more testing (per capita) than France, Netherlands, or the UK, but less than Germany. But if Belgium was over-reporting (as compared to other countries) wouldn’t it’s CFR be much higher?
I would bet their reporting is much more accurate rather than running the risk of over reporting. Most countries are probably seriously under reporting deaths in homes/nursing homes.
In Belgium it’s the other way around: unless the death can be specifically attributed to some other illness, all deaths in nursing homes are noted as Covid19. Deceased are not tested post mortem to make sure.
As you mentioned, small countries can be outliers and Belgium is relatively small, so could it be related to its size?
USA has a wider range of deaths per million than the countries you list:
NY 1038
NJ 590
FL 45
California 37
Texas 20
So what explains those differences? Population density, lockdown rules, stage through the curve (Belgium seems to be decreasing rates, other may catch up), etc.
Density is difficult to apply because it will vary within a country or state - Belgium has twice the pop. density of NY state, but NY city has 80 times the density of Belgium!
No, as above, Belgium is using a different methodology to count deaths. It’s one that may be more accurate and might slightly overestimate, rather than greatly underestimate (as most other nations probably are at the moment).
This is one of the issues with nation level statistics. Everybody is doing things a bit differently from everybody else, so direct comparisons are difficult to make. Every countries numbers would likely rise if they used Belgium’s method.
So, take direct comparisons between countries with a block of salt. They let us know hotspots, but they aren’t at the level where the error bars are sufficiently small. So comparing 300 vs 350 from different countries is not going to be as meaningful as we’d like.
I understand the Belgium is using a different methodology (indeed, the article in the OP seems to indicate that they are knowingly counting an–unknown–number of non-Covid deaths as Covid deaths).
But wouldn’t that be reflected in the CFR? If I did the math right, the CFR in Belgium is pretty similar to two of its neighbors (France and the Netherlands).
Note: Here’s my logic on the CFR and I’m not a math or statistics guy. In a typical country, you test a bunch of people and some number (less than 100%) of them die. Belgium does that and then adds in all nursing home deaths (if the home has a Covid outbreak) – all of these cases is both a “case” and a “death” (i.e., the CFR for Belgium’s nursing homes is close to 100%). Wouldn’t that (artificially) increase the CFR as well?
Really depends on the denominator and testing methodology (how many tests per capita and who gets tested), wouldn’t it? Also, the particular tests that do get conducted.
If you skew testing more to obvious symptomatic cases, which will also skew, this will produce higher CFRs.
Broader scale testing of the general population, including asymptomatic or mild symptoms, will presumably produce lower CFRs.
There’s also demography, but age/income/health statistics for France, Belgium, and the Netherlands are likely similar enough.
It’s a classic problem in statistical analysis - if your data is already biased, the resulting statistics are problematic. Think telephone polling for politics. If you randomly select landline telephone numbers, this seems fine. But the population of people who still have landline telephones and are further willing to answer a telephone poll is not necessarily representative of the general population. The data still has value but it has to be taken in context.
If Belgium and France and the Netherlands are all testing in similar fashion, then this is something strange. But any differences in how they test (and there are differences) and how much they test can shift numbers. That’s why direct comparisons between countries is difficult. It’s not just fatalities we don’t have a perfect handle on. We also don’t have a perfect handle on how many people are actually infected in the first place, especially as most testing is restricted to more obvious cases.
This. It cannot be emphasized enough; the current stats on deaths are estimates, and you should not assume they are specifically true. Differences between jurisdictions in reporting protocol will make huge differences.
Eventually we will be able to come up with much more accurate estimates through study of mortality rates.