I have made some further improvement to my flow (github.com/janvda/node-red-corona-comparison-hubei) as you can see in below example
The main new features:
-
running average : it is averaging the data for a configurable number of days.
-
new chart "reproduction rate (based on new conf cases)" : the reproduction rate at date X is defined as
(new confirmed cases at date (X+7 days)) / (new confirmed cases at date X)
Note that this is a very rough formula to calculate the reproduction rate as it gives a too pessimistic outcome in case the country is increasing the number of tests. If you want a better calculation of the reproduction rate see my other project "COVID-19 country history charts based on total tests"
Observations
- The number of cases dropped really rapidly in the province Hubei. Showing also a very low reproduction rate (< 0.5) in the last month.
- Italy is well beyond its peak. New confirmed cases and deaths are decreasing but not at the same rate as in Hubei. So the measures taken in Hubei were substantially more effective to reduce the reproduction rate compared to Italy. FYI It Italy would be as effective as Hubei then it would report the last confirmed case on 21 of May.
Comparison Sweden with Belgium
Observations
Sweden is one of the countries that are taking the least measures.
So I have compared this with Belgium which has 3 times more death today (scale factor = 2.995).
Of course it must be said that Belgium is counting also the deaths in eldery care houses (which accounts for about 54% of all reported deaths).
The main conclusion is that the downward trend that is clearly visible for Belgium is not that clear for Sweden. The number of deaths and confirmed cases stays high. Also the reproduction rate for Sweden stays around 1. So, this means that the measures most countries in Europe are taking which Sweden is not taking do clearly have an effect on the controlling the virus spread !
They are also counting "suspected" cases not just confirmed ones.
That's a very astute observation. Our government has in fact relied heavily on a model from the University of Washington Institute for Health Metrics and Evaluation (IHME) that is fundamentally the same as yours. It depends on fitting local or national data to that from Wuhan and Italy. You can take credit for accomplishing nearly as much as a team of professionals.
Unfortunately, politics enters here, and the news is bad. The Trump administration started using the IHME model extensively at a time when the conventional models were predicting in excess of 200,000 deaths. The lower IHME estimates were treated as good news, that the threat was less serious than previously believed. This was used to justify a more limited response, without considering that the Wuhan and Italy data already had a very vigorous public health response baked in. We may now be seeing what would have happened if other countries had delayed or limited their response.
1 Like
A first look, for what it's worth, is not encouraging:
1 Like
Brazil versus Italy
Observations
As Italy has currently far more deaths than Brazil, it is also very clear that Italy is way beyond its peak and is controlling it very well while the number of cases is still rising rapidly in Brazil.
Especially the chart with reproduction rate looks very concerning for Brazil. It is around 1.5 for several weeks with no tendency to decrease (this means that every week 50 % more people became effected than the week before).
Something else you might consider is an estimate of the case fatality rate (or perhaps more accurately, the infection fatality rate). Several investigators have estimated this as
( deaths at t+8) / (cases at t),
where deaths and cases are either cumulative or newly reported. The time lag (8 days) is an estimate of the average time from diagnosis to death for patients who die. Like the reproduction number, this is obviously not an intrinsic property of the virus, but variations by location and over time could reflect changes in testing and treatment protocols.
1 Like
As proposed by @drmibell: I have added the infection fatality rate % chart in release 2.4
This chart replaces the "total deaths (limited to 7 days in future)" chart.
Here below the comparison between US and Italy with this new chart.
Observations
- Infection fatality rate % in Italy is around 12% which is almost the double of the infection fatality rate in the US (around 6.5%). So I am still amazed that of the relatively recent confirmed cases in Italy still about 12% has died.
new charts for Belgium comparing it with Italy as a reference:
Observations
- based on averaging over 7 days : the "new deaths" and "new confirmed cases" are no longer decreasing. So "new confirmed cases" is rising 4 days in row (5,6,7,8 may). This is also visible in "reproduction rate" chart. This is very concerning as the most obvious explanation is that the virus is again spreading more rapidly.
- The fatality rate in Belgium (around 15%) and Italy (around 12%) is very high and remains high. This is considerably higher compared to other countries. For Belgium it can at least partially be explained as it is also counting the deaths outside the hospitals (including those that are not confirmed but presumably had covid-19).
Russia compared to UK
Observations
- Russia (Putin) yesterday announced to reduce the lock down measures. This is strange as reproduction rate is still considerably > 1 (around 1.5).
- The fatality rate in Russia is very low around 1.5% compared to 10% (and more in the past) for the UK. I have no good explanation for this big difference. Is Russia properly counting all its covid deaths ?
Update
Russia is testing more than the UK, so will also detect more infected people with mild symptoms that will survive. So that at least is a partial explanation of the fatality rate difference - see also charts below.
I also think UK starts to count people that died outside the hospitals (elderly care homes), which most countries are not yet counting in their statistics.
Good point.
Does the data have age values? I mean is it possible to extrapolate deaths in age bands? I'd expect to see higher death rates for 60+ and very low numbers for 20-
Data like that should be used in any lockdown considerations (IMO)
Edit...
Re reading the above seems a bit clinical and cold - sorry. I should add these numbers are horrific and the families affected must be devastated. Its a strange old time.
Some googling showed me the following:
-
As of 2018, the average life expectancy in Russia was 67.75 years for males and 77.82 years for females.
-
Life expectancy at birth in the UK in 2016 to 2018 was 79.3 years for males and 82.9 years for females
So you triggered a very valuable point. There is a difference of almost 12 years for man in life expectancy. So Russia has a much lower percentage of people in the most vulnerable category for covid-19. This certainly also explains to a considerable extend the much lower fatality rate.
There is also a significant difference in life expectancy between different parts of the UK as there will be in Russia too. When you look at mortality rates compared to urbanisation and poverty, you get a much clearer picture. There are several clear factors that seem to be influencing mortality rates:
- Urbanisation - higher density = higher transmission, potentially overwhelming local services.
- Poverty - lower income means poor diet, poor living conditions and lower education levels all of which influence life expectancy and mortality rates across most illnesses.
- Pollution - obviously also related to the previous points but there is a very clear link between pollution and health/mortality.
- Availability of ICU beds and specialist staff - this is probably clearest in Germany. Where there is a higher % of specialist beds & staff, mortality from COVID is lower.
Obviously, these are not restricted to COVID apart from maybe the last one, they hold true for all health.
Update for anyone using this data to compare with The Netherlands, starting last Friday the reporting criteria for hospital admittances have been changed. In the day-to-day stats only people admitted with a primary diagnosis of covid-19 are counted, people with covid-19 as comorbidity are later added to the totals. Here, the hospital admittance day-to-day rate is one of the primary forms to calculate the R-factor. Combined with (still) a lack of testing the value of the combined set of data from The Netherlands as available through JHMC in Baltimore it’s hard to use any of this data now for comparisons especially in the flows like these.
There is strong correlation between age and case fatality rate as you can see in the chart "Coronavirus: case fatality rates by age".
Also note that according latest statistics the case fatality rate of UK (9.9%) is now very similar to the case fatality rate reported by Germany (8.5%) (see "infection fatality rate % chart" below). In March and beginning of April the difference was much more - most likely - because Germany did more testing while UK only tested the people that were very ill.
1 Like
Note that the charts don't show any data related to "hospital admittances". It is all based on confirmed cases and reported deaths.
Of course confirmed cases is also impacted by the number of tests performed which differs from country to country and reported deaths is also impacted by the way the deaths are counted. As communicated earlier - in Belgium also deaths are counted that are not confirmed - basically this means that we are counting twice as many deaths as other countries would do. So this should be taken into account when comparing absolute numbers. Also the "new death" curve in Belgium should be shifted a bit to the left as the deaths outside the hospital (which is 50%) are often reported with delays of several days in these stats.
Observations
At this point the most interesting chart is the "reproduction rate" chart. So for Belgium and the Netherlands it looks very similar. It was well below 1 during the last 3 weeks which is very good but the rate is again rising. In Belgium they said that this was due to the new way of testing - in the Netherlands it might be due to more testing ... or due to relaxing of the measures.
Not sure how to read your charts, NL currently has 42.900 confirmed cases and 5475 confirmed deaths, your charts show something completely different.
That is because all the charts for the Netherlands are scaled on the Y-axis. The scale factor used = 1.586 which makes the current total deaths for the Netherlands equal to Belgium. I did that to make comparison visually easy.
FYI Here below the same charts without scaling. Note that the numbers still differ a little bit from the ones you have provided as I am taking a 7-day average (this way I am filtering out the weekend fluctuations you would otherwise see).
Brazil versus Italy
Observations
- Brazil still has a high reproduction rate (1.5) at 2nd of May. So problem is getting worse and worse. Of course as we are now (= 13th of May) 10 days later the current reproduction rate might be lower.
- Infection fatality rate of Brazil is very similar to the one of Italy (around 12%)
- Last (averaged) number of confirmed case for Brazil was almost 9000 which is substantially more than the peak of Italy (5600).
All this indicates that situation in Brazil will substantially become more severe than it ever was in Italy.
