[Flow] Corona country comparison with Hubei (China)

Hi, thanks for a great tool. I have had no problems for the last few days and have been using the update function for Germany but today there appears to be no data in the github. Has something changed?
Andy

1 Like

Hi Andy,

The issue is that the flow was retrieving 2 files from https://github.com/CSSEGISandData/COVID-19 that are no longer updated. Instead we have to use the following 2 files:

My github repository is already updated for this !

1 Like

Updated charts for Italy with lock down data 2020-03-05

Notes:

  • The Hubei data is smoothed (new feature which is documented in README.md)

Observations:

  1. The Hubei factor has increased a little bit from 6.186 (22 March - 6 days ago) to 6.259 (20,000 deaths)
  2. It must be noted that in terms of curve fitting (chart "total deaths (limited to 7 days in future)") a "lock down date" of 2020-03-07 (2 days later) gives a slightly better fit and would give a hubei factor of 8.064 (25,000 deaths)
  3. The new confirmed cases for Italy will go rapidly down from 31 March (or maybe 2 days later) onwards if it is following the Hubei curve.
  4. The number of deaths for Italy will go rapidly down from 7th April (or maybe 2 days later) onwards if it is following the Hubei curve.

Charts for Spain with lock down date 2020-03-15

Observations and Notes:

  1. I have taken lock down data 2020-03-15 instead of 2020-03-12 (see my post 11) as this resulted in better fit with the Hubei data.
  2. The Hubei factor is 10.605 (33,000 deaths) which is worse than Italy (hubei factor between 6.259 and 8.064 - see my previous post)
  3. A lock down data of 2020-03-15 also means that Spain is about 8 to 10 days behind what is happening in Italy.
  4. Confirmed cases will go rapidly down from 10th April onwards if they follow the Hubei curve.
  5. Deaths will go rapidly down from 17th April onwards.if they follow the Hubei curve.

Charts for the Netherlands with lock down date = 2020-03-16.

Observations

  1. lock down date = 2020-03-16 is chosen as this currently gave the best fit (see chart "total deaths (limited to 7 days in future)").
  2. The hubei factor is 1.305 (4,000 deaths). Compared to post 15 where the lock down date was 2020-03-15 the Hubei factor was 1.088 (about 3,500 deaths).
  3. So problem in the Netherlands is more severe in absolute numbers than problem in Hubei (and in fact in whole China as outside Hubei there are hardly any deaths reported in China).

Charts for Belgium with lock down date = 2020-03-12 but also including following manually added stats for today = 2020-0-28:

  • new deaths = 64
  • new confirmed cases = 1000 (just a rough guess - this number is not that important at this stage).

Observations

  1. Hubei factor is now 0.449 (1,400 deaths) while in post 18 ( 3 days ago) it was 0.324.
  2. You can also see in chart "total deaths (limited to 7 days in future)" that Belgium is not following the Hubei chart. The number of deaths is more rapidly increasing for Belgium. We get a better fit when using 2020-03-17 (hubei factor = 0.843) or 2020-03-18 (hubei factor = 1.023) as lock down date. This is a troubling as it indicates that the measures taken by the Belgium government at 2020-03-12 don't have the same effect as the lock down measures taken by Hubei.

Butting in for a minute, but some additional background on this as while the Netherlands is now finally taking measures, there’s no actual lockdown going on when comparing to other countries with this statistic listed. The prime minister calls it an “intelligent lockdown” at times. It’s more or less similar to how it was for the UK before they entered their lockdown, but it’s not enforced as such, nor called/named like that. People are asked to use common sense, but police still having to break up “anti coronavirus parties“ following the new “no events, only 3 people together if family, otherwise 1.5 metres apart” guidelines shows that common sense can be hard to find. Furthermore, number of confirmed cases is inaccurate due to a lack of tests done, and as a new report from earlier today shows the virus has been present in the country for at least a week before patient zero was diagnosed as such. The number of deaths might be inaccurate too as doctors are mentioning that patients with suspicions of the virus who die aren’t tested, and aren’t counted in the statistics either

Thank you for the great tooling you create with these flows :slight_smile:

2 Likes

It is safe to say that all of the stats should be taken with a very large pinch of salt.

Test numbers and who is being tested varies greatly, death records (including death categorisations) also vary (note Russia in particular as an example). Quality of testing also plays a part. Only really the general trends are worth noting for most of our countries.

Of course, then there is the issue of understanding the difference between deaths that would be considered "normal" for this time of year vs those where COVID-19 is an influence. For those countries on lockdown there is a further complication in understanding the overall impact of the pandemic - lockdown will be reducing other categories of deaths such as car and other travel accidents. It will probably be years before we really understand the full impact - if we ever do.

I also though wouldn't want anyone to confuse an analysis of stats with personal impact. Every death, illness and isolation takes a personal toll.

1 Like

It is true that there is a big gap between the actual number of cases and the confirmed cases. Moreover not every country is testing with the same intensity. In Belgium there are currently only testing the hospitalized people and medical staff that is ill. So comparing the number of confirmed cases between different countries (or even between Hubei province in China with other provinces) doesn't make much sense.

The issue with the death statistics is different. There the issue is not that different countries are counting them differently (especially not for the countries for which I have shared the charts). But that they are not a good statistic to predict the severity well in advance. E.g. for Hubei: once the "confirmed" cases started dropping, it still took a week to see a comparable drop in the number of deaths.

That said, comparing with Hubei, showed that up to now, most countries are following a very similar curve. The major difference is the shift in time and its severity (hubei factor). So it might give a reasonable good prediction for the coming days and weeks.

3 Likes

OK, this is my last post on this because we are a long way off topic.

However, I have to point out that I don't believe that statement to be entirely correct. In fact we know for sure that some countries are counting deaths under different categories and that this impacts the stats. We will also now be getting many instances where people come into an overloaded ICU without having been tested, not recovering and not being tested at all.

Anyway, whether we are looking at this from a crisis perspective or from an analytics one, I think this topic is now too far off track for the Node-RED forum.

1 Like

Is there some interest to extend the flow so that it is also possible to draw the charts at the level of an individual US state ?

I have found an interesting repository that has the input data for this:

1 Like

Yes there is for those of us in the US. (well at lease I'd be interested)

1 Like

It would be interesting to be able to compare & chart one country with another. For example UK & Italy, instead of all comparisons being made against Hubei.

2 Likes

This is your regular reminder to take these figures with a MASSIVE pinch of salt. We are seeing more and more irregularities with the numbers.

Japan, for example has shown a weird hike after the announcement of the delay of the Olympics. Russian figures are suspiciously low, USA have very poor testing, China is being accused of wildly inaccurate numbers, ... there are many other issues with them.

Even here in the UK where I am rather closer to the numbers, there are all manner of issues. I think people are trying to get them right, certainly many colleagues that I know I believe will be trying, but there are so many issues on the front line that numbers may not always be upper-most priority for obvious reasons.

3 Likes

Hi Julian,

Thanks to these charts I was already aware that you need to be careful when comparing confirmed cases (see example below).
If there are other anomalies with the data then these charts might as well be a nice instrument to make this clear.

E.g. Here below the comparison with province Henan with lock down date = 2020-01-23

Observations

  1. Only 22 deaths are reported for province Henan (population 96 million) - also note that excluding Hubei this is the province with most death in China ! In the context of this it is also chocking that the president of the US (population 327 million) is claiming to expect between 100,000 and 200,000 deaths while in the most affected province in China (excluding Hubei) they only reported 22 deaths.
  2. The deaths and confirmed cases happened during the same period as they were also reported for Province Hubei.
  3. In chart "Total confirmed cases" you see that for Henin 3 times more confirmed cases are reported per death. This is strange but possible explanations are: 1/ Henan is more thoroughly testing to find back all/most cases to control the spreading compared to Hubei (similar discrepancies we see now in Europe between Belgium and Germany. 2/ Another explanation might be that 50 deaths were not reported as Corona deaths in Henan.

I have updated the flows so now it is possible to compare with another country instead of Hubei (China).

Don't forget to push on update charts button as the charts are not automatically refreshed when changing/setting the location1 and location2 parameters.

Here below I have compared United Kingdom (lock down date = 2020-03-23) with Italy (lock down date = 2020-03-07).

Observations / Notes

  1. date shift = 16 days, this is just the difference between the lock down dates specified for both locations. In other words it means that UK took 16 days later similar measures as Italy.
  2. scale factor = 1.211 means that UK has currently 1.211 times the number of deaths compared to Italy 16 days ago (date shift) . So if UK follows the same path as Italy it will get 21.1% more death than Italy.
  3. chart total death (limited to 7 days in future) shows that the curve for UK is a bit more rapidly rising than Italy. This indicates that the UK measures at 23rd March don't have the same effect as the measures taken by Italy at 7th of March. So UK will most likely even have more than 21.1% death in the end compared to Italy.
1 Like

I added the US states data from https://github.com/nytimes/covid-19-data

E.g. here below the charts for New York (based on lock down date 2020-03-23).

Here below a comparison of New York (lock down date = 2020-03-25) with Italy (lock down date = 2020-03-07)

1 Like

Nice update @janvda
Charting UK & Italy, it shows the peak of new UK cases for next weekend, which is not dissimilar to the Government predictions.
With all the data anomalies & recording issues, it will be interesting to see how this plays out.

2 Likes

@Paul-Reed indeed the charts are not that bad in predicting the evolution. I hope UK is better prepared than Italy because they will even be hit harder than Italy based on current comparison.

1 Like

Comparison of UK (lock down date = 2020-03-23) with Italy (lock down date= 2020-03-23).

Observations

  • The chart is inline with the predictions made 5 days ago (now scale factor = 1.161 compared to 1.211 5 days ago).
  • So UK seems to be in an even worse situation than Italy (16% more death) but with a delay of 16 days.
  • The date shift of 16 days also means that the UK has waited 16 more days than Italy to let the virus wildly multiply in the UK before taking measures to control the spreading. This is very painful as this might mean that the total number of deaths is maybe up to 10 times more than it could be (e.g. total confirmed cases in the UK at 2020-03-07 was around 50 while it was around 1000 at 2020-03-23 (16 days later))
1 Like