Read on Twitter
Here's why I think this #hrdoxychloroquine "study" data is fake
https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)31180-6/fulltext#sec1
A thread....
@99freemind @realDonaldTrump @PressSec
https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)31180-6/fulltext#sec1
A thread....
@99freemind @realDonaldTrump @PressSec
(1) the study says that they received data from 600 hospitals up to mid-April, and it was published end-May.
This is impossible. If you have ever collected clinical research data you will know how impossible it is.
This is impossible. If you have ever collected clinical research data you will know how impossible it is.
You have to get doctors to enter information and they are completely unreliable, requiring to be chased all the time. If you're lucky some units have a research nurse who is much better, but the data reports are either on paper (needing mailing/faxing) or an electronic repository
Electronic repositories are faster but you need ethical approval, suitable web servers set up, access set up for each hospital (and ideally each worker). This is what the authors say happened - they say they questioned existing repositories
I'm sorry but these don't exist. Repositories do exist that maintain some of this data but it is mostly unlinked, relies on it being filled in correctly and exists on different systems that don't talk to each other.
Add in different countries this stuff is just not accessible
Add in different countries this stuff is just not accessible
They are trying to claim that they are able to access this information in other countries.
But
(1) on all clinical record systems information is always patchy. "ex smoker", "BMI" and "current medications" are notoriously not present.
But
(1) on all clinical record systems information is always patchy. "ex smoker", "BMI" and "current medications" are notoriously not present.
(2) for each hospital system providing access you have to go through a complex ethics approval system (otherwise your hospital has just given your personal data to some strangers).
THIS DOES NOT HAPPEN QUICKLY, IF AT ALL - we are talking months to set up
THIS DOES NOT HAPPEN QUICKLY, IF AT ALL - we are talking months to set up
Even if you get ethics approval, then you would have to implement a system in your PAS (patient admin system) that allows external access. This is *really* frowned upon in all Western countries
https://www.efa.org.au/Issues/Privacy/privacy.html
https://www.efa.org.au/Issues/Privacy/privacy.html
An ethics approval of this magnitude would take months but it would take even longer to get the system up and running to allow secure access to hospital data systems externally.
IT JUST DOES NOT HAPPEN.
IT JUST DOES NOT HAPPEN.
Feel free to try and navigate the rules and regulations for NSW (Australia) yourself....
https://regis.health.nsw.gov.au/
https://regis.health.nsw.gov.au/
And here is the #covid19 specific guidance...
The reason it is complex is to protect patient privacy and ensure that the ethics are followed which go all the way back to the "Nuremberg code"
https://www1.health.gov.au/internet/main/publishing.nsf/Content/Clinical-Trials
The reason it is complex is to protect patient privacy and ensure that the ethics are followed which go all the way back to the "Nuremberg code"
https://www1.health.gov.au/internet/main/publishing.nsf/Content/Clinical-Trials
Anyway, moving on from the fact that it is absolutely impossible to get all this data from different countries (or even your own) reliably and quickly, then we have the data itself.
There is an appendix, here https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)31180-6/fulltext#sec1
There is an appendix, here https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)31180-6/fulltext#sec1
This gives more of a breakdown of the data and uses something called "propensity matching". This is a method of matching up similar patients.
So for instance you want to get the same proportions of black/white patients in each group. Same for BMI, smoking, etc
So for instance you want to get the same proportions of black/white patients in each group. Same for BMI, smoking, etc
Each time you do this for each factor you reduce the pool of patients to compare. So to get the same number of black, non-smoking, BMI 25, hypertensive people in each group is really tough (5 factors)
This group did it with 23 factors, all matched up "perfectly"
This group did it with 23 factors, all matched up "perfectly"
This is pretty much statistically impossible, and they are claiming that they got this matching in 7000 patients out of a pool of 96000 patients for which they received high quality information from 671 hospitals.
Nope. Sorry, the data is too "clean".
Nope. Sorry, the data is too "clean".
But wait, this is how they did it, right? So they only took PCR-positive patients?
OK, now we're opening up a can of worms.
Do you know that pathology systems and hospital record systems are completely separate?
OK, now we're opening up a can of worms.
Do you know that pathology systems and hospital record systems are completely separate?
So we look at the "Johns Hopkins" data from around the world and assume it's contemporaneous. It must come from somewhere (but we don't know where).
OK, well it must come from the pathology labs, yeah?
https://gisanddata.maps.arcgis.com/apps/opsdashboard/index.html#/bda7594740fd40299423467b48e9ecf6
OK, well it must come from the pathology labs, yeah?
https://gisanddata.maps.arcgis.com/apps/opsdashboard/index.html#/bda7594740fd40299423467b48e9ecf6
OK, so the pathology labs around the world (including Senegal, Turkey, Pakistan and of course #china) automatically feed their patient-level information to Johns Hopkins using their electronic feeds that they set up long before the #pandemic right?
OK, let's just assume this exists (because there isn't a hotline for manual reporting btw, so it has to be electronic)....
That could only come from pathology labs *not* hospital patient systems (PAS). PAS systems can talk to pathology lab systems, but....
That could only come from pathology labs *not* hospital patient systems (PAS). PAS systems can talk to pathology lab systems, but....
the data is not imported, it's linked.
If you work on a clinical record system you can click a link to a particular pathology result, but it's a link to the pathology lab. Those results are not (usually) imported as a static record into the PAS
If you work on a clinical record system you can click a link to a particular pathology result, but it's a link to the pathology lab. Those results are not (usually) imported as a static record into the PAS
Anybody who knows how databases work would understand this.
So, this group are telling us that, not only have they got access to 671 hospital clinical record systems but (and this is a problem)....
So, this group are telling us that, not only have they got access to 671 hospital clinical record systems but (and this is a problem)....
They have access to all the pathology providers linked to those 671 hospitals and THEY CAN DATA LINK THE PATIENT RECORDS.
Not a chance.
This didn't happen, and if it did it would be a massive privacy scandal.
Not a chance.
This didn't happen, and if it did it would be a massive privacy scandal.
There would be serious financial and potentially criminal repercussions - look what happened to @Google only recently when they reidentified patient data without authorisation https://www.chicagomaroon.com/article/2019/6/27/uchicago-google-sued-federal-class-action-suit-dat/
So, either this data is completely fabricated
-or-
The authors have been able to overcome impossible hurdles of data sharing and probably broken a number of laws in the process.
I know which one I'm going for.
Oh, btw @TheLancet feel free to challenge me on these claims.
-or-
The authors have been able to overcome impossible hurdles of data sharing and probably broken a number of laws in the process.
I know which one I'm going for.
Oh, btw @TheLancet feel free to challenge me on these claims.
ADDENDUM:
For those who believe medical journals are not subject to political bias please feel free to read this released this week (an open access article from a closed access journal) https://www.sciencedirect.com/science/article/pii/S0090825820310921?dgcid=rss_sd_all
For those who believe medical journals are not subject to political bias please feel free to read this released this week (an open access article from a closed access journal) https://www.sciencedirect.com/science/article/pii/S0090825820310921?dgcid=rss_sd_all
@TheLancet is more prone than most journals to political bias.
It has a whole subsite dedicated to "climate change"
https://www.thelancet.com/climate-and-health
It has a whole subsite dedicated to "climate change"
https://www.thelancet.com/climate-and-health
CORRECTION:
#hydroxychloroquine
I should have realised this thread would be of interest and checked my spelling on the first tweet!
#hydroxychloroquine
I should have realised this thread would be of interest and checked my spelling on the first tweet!
Thanks to this poster we now know that this data is fake.
#hydroxychloroquine #medicalfraud #researchfraud
Lancet article archived here: http://archive.is/8WEnF https://twitter.com/SajidJa23282468/status/1264938091313115136?s=20
#hydroxychloroquine #medicalfraud #researchfraud
Lancet article archived here: http://archive.is/8WEnF https://twitter.com/SajidJa23282468/status/1264938091313115136?s=20
Another proof that this is fake data.
PSM requires *by necessity* a subgroup of each group to be compared. It can't use 100% of the data. Maybe 50%, more likely 25%.
Not these people @Surgisphere - they don't lose a single patient!
@TheLancet has some real explaining to do
PSM requires *by necessity* a subgroup of each group to be compared. It can't use 100% of the data. Maybe 50%, more likely 25%.
Not these people @Surgisphere - they don't lose a single patient!
@TheLancet has some real explaining to do
Why am I getting an Elizabeth Holmes feeling about @surgisphere? https://www.pharmanewsdaily.com/rapid-coronavirus-test-developed-by-chicago-based-surgisphere/
