Clarity on French Data

[CSSEGISandData]

We are issuing this announcement to explain the logic behind our handling of cases, deaths, and recoveries in France. We recognize this has been a source of great frustration among developers, particularly since our numbers are greater than the official numbers from the French government and lower than those of other data aggregation sites, but we have needed to validate multiple details before we were comfortable publishing our method.

This is the general logic:

(France confirmed & probable cases in ESMS (via French Government)) - (confirmed cases in Guadeloupe, Mayotte, French Guiana, Reunion, & Martinique (via Regional Governments))

We will address the two items differentiating us from other methodologies here:

1. Subtraction of cases, deaths, and recovered from Overseas Regions and Departments (DROM): the data published by Sante Publique include cases in DROM as well as continental France. We have these regions designated as level 1 administrative regions in the dashboard, and thus have their own entries. These numbers have to be removed from our count for continental France or these cases will be double-counted and the case burden in France will be artificially inflated. This distinction explains why our case number in France is lower than in other data aggregation sources who are not making this correction.

2. Inclusion of probable cases: we recognize that this is a controversial decision, however we believe this is the best means of estimating the ground truth case burden in continental France. The reason we are implementing this approach in France is for three reasons:

   a) France has implemented strict criteria for an individual to qualify as a probable case that severely restricts the likelihood of a patient that is truly negative being included in the count. For that reason, we have confidence that these cases overwhelmingly represent true infections.
   b) France has made the explicit decision to not test these cases and they will not be reflected in the official case count. We believe this masks the true extent of the infectious spread in the country and should be captured by those interested in understanding the case burden across time and space.
   c) France is quantifying and publishing these numbers, in contrast to several other governments that have implemented similar approaches.

Again, we apologize for our radio silence on this issue but needed to ensure the legitimacy of our approach. We appreciate the feedback we receive on the dashboard but please do recognize that we are a small team managing a highly scrutinized dataset and keep your comments kind and civil.

The CSSE team.

[JiPiBi]

Concerning point 1 , as I made my own fixes in your datasets :

• When do you apply this new rules ?
• Will the timeseries be corrected from the beginning ?
  Edit : In your errata.csv , in last lines , you indicate that it was applyed from 4/14
  Seems correct from that date for example for deaths with my french official values.

On point 2, you have made a choice, dont know if you have the same policy for all countries : for example in Portugal

[CSSEGISandData]

Hello

This rule has been applied since April 14th, which is, as far as we can tell, the earliest date where the French government delineated suspected and confirmed cases in ESMS.

The suspected cases in Portugal are not the same as probable cases in France. Probable cases in France are symptomatic and live in nursing facilities with 2+ confirmed cases of SARS-CoV-2 infections and will not be tested via PCR. Suspected cases in Portugal are symptomatic and will be tested for the virus.

For more on the Portugese strategy please see: https://covid19.min-saude.pt/estrategia-e-captar-o-maximo-de-casos-suspeitos-de-covid-19/

[lautaroblatto]

Hi

I'm wondering if you could help me understand the huge jump in French confirmed cases from April 11th to April 12th. The number goes from 93790 to 120633.
Are these the correct figures? I see it is a few days before the adjustment of the rule, so I thought I should ask.

Thank you very much for any explanation you may be able to give me, and for all your work.

[JiPiBi]

@lautaroblatto
the explanation is above, JHU decided to count possible cases in nursing homes in the same counter than confirmed ones.
Possible has not the same meaning than probable in french. So you add potatoes ans carrots, and you say you count potatoes, no you count vegetables.

Considering suspected cases in Portugal, the ratio between suspected and confirmed is so huge , that it is hard to understand that they can test all suspected, but it is their issue.
Concerning France, governement announced to prepare itself to test up to 700000 person a week . Wait and see.

To come back to the confirmed values, in the period of high symptomatic numbers of people is of low interest IMHO .
The main KPI are new hospitalyzed and new ICU .
After the end of confinment , it could become again interesting to measure the virus circulation, but the main interest of testing will be to isolate confirmed ones to block propagation.
For the moment the relationship between test and death is not obvious : look at japanese and italian situation on worldometers.

[lautaroblatto]

Thank you very much for the reply!

[traut21]

Where's the explanation or proof, how the numbers in France are guessed?
If data is published within a file named "confirmed" I don't feel it's justified to use estimated, but unconfirmed numbers instead.

WHO does use the officially confirmed cases: https://covid19.who.int/region/euro/country/FR
European Centre for Disease Prevention and Control (ECDC) does use the officially confirmed cases: https://qap.ecdc.europa.eu/public/extensions/COVID-19/COVID-19.html

That's why I do use the numbers from https://www.ecdc.europa.eu/en/publications-data/download-todays-data-geographic-distribution-covid-19-cases-worldwide instead now - and JHU still does not handle issues here properly, I feel.

[JiPiBi]

FYI Worldometers makes also the same addition confirmed+possible.

[traut21]

I know worldometer - and they emphasis that their numbers are "total numbers" instead of "confirmed numbers". Earlier on they named that they calculated those numbers. I do not find that information within their sources any more - but you can find out, that their numbers are estimated.

[JiPiBi]

Even french official values in confirmed are estimated, it is clearly written in the reports.
They seem unable to extract an unique list of confirmed persons from the positive test list. Probably a dataset or confidentality issue.

[drbansal0]

Добрый день !

Сообщение от: - "БОЛЬНИЦА ЧИСТОЙ ВОДЫ", США.

Нам нужны срочные / серьезные доноры органов, которые готовы продать следующие органы ниже ...

(1) Почки
(2) печень
(3) поджелудочная железа
(4) Легкие
(5) яички
 
Мы заплатим 850 000 долларов за любой из вышеперечисленных органов.
Доноры должны быть клинически здоровы с любой из групп крови, A, B или O положительных в возрасте от 18 до 55 лет.

ПРИМЕЧАНИЕ: ДЛЯ ТОЛЬКО СЕРЬЕЗНЫХ И Зрелых ЛИЦ.

По всем вопросам, отправьте нам письмо по электронной почте

(nationalkidneyfoundation156@gmail.com)

или отправьте нам прямые сообщения на ..

Через WhatsApp (+1 559 744 3171)

[Epi-Mic]

Hi, where do you get the number of probable cases in france?

[k-r-b]

The total cases in France here are 14,000+ less than on the JHU dashboard (https://coronavirus.jhu.edu/map.html). Is there a way to pull the data that feeds into the JHU dashboard?

[servoz]

French numbers are still wrong?

Unless I am mistaken, the following data, coming from the French Ministry of Health, are good (for example for the 2020-09-09 we observe):
casConfirmes: 344101 (this is the total number of confirmed cases. i.e. including EHPAD and EMS).
deces: 20319 (this is the total number of deaths outside of EHPAD and EMS)
decesEhpad: 10475 (this is the total number of deaths in EHPAD and EMS).

So casConfirmes gives well the total number of confirmed cases since the beginning of the epidemic, whereas to have the total number of deaths since the beginning of the epidemic it is necessary to add deces and decesEhpad.

In summary, based on the data from the Ministry of Health, for 2020-09-09:
Total number of confirmed cases = casConfirmes = 344101
Total number of deaths = deces + decesEhpad = 20319 + 10475 = 30794

We can find these numbers, 344101 and 30794, also here.

I had already noticed that the total number of confirmed cases for France was wrong from CSSEGISandData. I just realized that the total number of deaths for France is also wrong.
Indeed if we look, for France at the 2020-09-09, the total number of confirmed cases and the total number of deaths we see:
Total number of confirmed cases: 363751 (344101 from the French Ministry of Health, vide supra)
Total number of deaths: 30643 (30794 from the French Ministry of Health, vide supra )

I can't understand why the results for France have been wrong here for so long when the method to get the correct numbers are now known (Total confirmed cases = casConfirmes and Total death number: deces + decesEhpad) ?