Read on Twitter
So @NickHudsonCT seems to think @Kevin_McKernan “DEMOLISHED” my recent thread on @ClareCraigPath ‘s pretty awful PANDA document https://twitter.com/dr_barrett/status/1378878486027309057?s=21
Let’s examine that proposition - I ask you to please share&retweet as this will be fun ..
#COVID19 #Medtwitter #scicoms
Let’s examine that proposition - I ask you to please share&retweet as this will be fun ..
#COVID19 #Medtwitter #scicoms
This is Kevin seen in video by “vaccines can make you time travel” @naomirwolf (senior fellow at AEIR based at #GreatBarrington) - he is showing her a fungus on agar plate because that has something to do with sars-CoV2 diagnostic PCRs ... 
Kevin dropped a number of “clangers” trying to rebut my thread some worthy of LONG threads... (those that follow me know I do LONG 
s)- for now let’s look at this one - I might end up doing a series on Kevin’s @BadCOVID19Takes for now let’s look at this criticism 
s)- for now let’s look at this one - I might end up doing a series on Kevin’s @BadCOVID19Takes for now let’s look at this criticism
It relates to this tweet in my - based on this Nat Comms paper https://www.nature.com/articles/s41467-020-20235-8 Kevin takes issue with my DENOMINATOR in Postive Predictive Value calculation ... https://twitter.com/dr_barrett/status/1378878562418184193
- based on this Nat Comms paper https://www.nature.com/articles/s41467-020-20235-8 Kevin takes issue with my DENOMINATOR in Postive Predictive Value calculation ... https://twitter.com/dr_barrett/status/1378878562418184193
Refresher for benefit of @Kevin_McKernan - Positive Predictive Value is probability that a positive test result is a TRUE positive. The BASIC formula for PPV is this
So let’s look at Kevin’s criticism again ... one thing Kevin wants is me to include the probable cases (false negatives) in my PPV calculation but False negatives means TESTED NEGATIVE - therefore they have NO PLACE in a PPV calculation #Kevinfail ... so there is that ... 
Now let’s get this the next claim “only 773 had enough DNA to be sent for sequencing” - is that a true claim ... well no ... only 773 samples were sent by District Health Boards - as retention and sending was not mandatory at time TY for confirming that @Jemma_Geoghegan
Importantly we can see a further follow up of the August Auckland outbreak which had a total of 179 cases ... the same team also carried out sequencing of these samples ... I saw the pre-print but now these results are published .. https://twitter.com/jemma_geoghegan/status/1378047894767378432?s=12
This time they received 172/179 samples or 96% of samples from the outbreak - resulting in 145 genomes recovered ; this suggests a PPV of at Least 84% (145/172) or EVEN assuming samples not sent would be unsuccessful in sequencing 81% (145/179)
This is in the context of LOWER prevalence* in test population during the August outbreak now what did we learn about disease prevalence & PPV ; we expect PPV lower (FDR higher) ... but MINIMUM PPV is relatively preserved despite FAR LOWER prevalence
https://twitter.com/dr_barrett/status/1378878555044601858?s=21
now what did we learn about disease prevalence & PPV ; we expect PPV lower (FDR higher) ... but MINIMUM PPV is relatively preserved despite FAR LOWER prevalencehttps://twitter.com/dr_barrett/status/1378878555044601858?s=21
In short the PPV based on % of samples in which genomes could be recovered is broadly similar to the prior study -it argues very much against any premise that Kevin is making that the samples never sent were or poorer quality / lower Ct / lower viral loads/ lower PPV etc etc
I have already pointed out that only 4% of NZ cases don’t have a infection source identified - so the PPV is likely much higher than the already high % (given low prevalence) suggested by the viral genomic studies ... https://twitter.com/dr_barrett/status/1378878555044601858
Even taking this bottom estimate of PPV - a middle of road sensitivity 75% (this genomic study suggests) suggests a specificity of the PCR of greater than 99.99% or less than 1 in 10,000 false positive rate ... In short the FPR is very very small ...
I do know this a very approximate calculation as sensitivity is a estimate (which means true prevalence is an estimate) - this is meant to just illustrate the point only @igdstrachan could probably do a better job at explaining the likely specificity from these numbers ...
So in short I don’t think @Kevin_McKernan “DESTROYED” my debunk of @ClareCraigPath’s PANDA document but rather exposed his own lack of understanding of the literature and how to evaluate characteristics of diagnostic tests ...
So I hope this retort did not disappoint .. @andrew_croxford ... for the record Kevin I didn’t need @MackayIM’s help on this - I am no high level expert but I can read a paper and understand basic concepts . https://twitter.com/andrew_croxford/status/1380264527602802694
Retweet and share if you enjoyed. ... @PanData19 a terrible organisation and needs to be challenged in their claims. @SiouxsieW @jneill @BadCOVID19Takes @DrDomPimenta @killeen_gerry @s8mb @cjsnowdon @Kevin_Fong @TakethatCt
