12. In the United States, for example, they compare 10.4 deaths (from all causes!!) per million Pfizer vaccine recipients with 7.5 per million J&J recipients. Pfizer were approved in December. J&J in February. Pfizer recipients have had a lot more time to die!

11. One straight-up falsehood on the Sputnik slide is that these numbers are being presented as death *rates*. Rates involve count per time. These don't take time into consideration. And that matters a LOT because some vaccines have been available longer than others.

10. Similarly, the reporting—both of deaths and of who has been vaccinated—can matter a great deal. How confident are you in the numbers coming out of those countries that have most heavily used Sputnik?

9. The age distribution can make a huge difference. If, for example. Pfizer is more often preferentially given to the elderly than is Sputnik, we would expect much higher overall death rates among those vaccinated with Pfizer compared to Sputnik.

8. The countries in question differ in factors including Age distribution of those vaccinated COVID incidence and mortality rate Background mortality rate Accuracy of vaccine logging and death reporting

7. Given that vaccine-associated deaths would be very low even if there *were* a problem, we would expect any causal differences by vaccine type to be swamped my demographic differences in who is receiving in the vaccine.

6. In the right column, we are given weighted averages. But these are nonsense as well, because the data are not collected in a way that allow for meaningful comparisons. Most critically, THESE ARE TOTAL DEATHS, NOT VACCINE ASSOCIATED DEATHS.

5. Taking the straight average of the death rate in Norway, which has vaccinated 1.2 million people, and the US, which has vaccinated 140 million people (not all with Pfizer, in either case), is bizarre. Why even show these numbers? Because they make Pfizer look bad.

4. Here are a few of the things wrong with these figures. The numbers in the column at left represent country-level averages, unweighted by number of recipients. No one with an iota of quantitative understanding would do this, unless trying to deceive.

3. But statistics (1) are only as good as the methods used to derive them, and (2) are only useful when they allow you to make fair and meaningful comparisons. The Sputnik V numbers fail spectacularly on both accounts.

2. Their unfounded claim is that we are observing higher death rates among Pfizer recipients. This is rubbish. In our book, we address the way in which people will try to bamboozle you with the unwarranted authority of numbers by throwing lots of stats at you.

1. Today’s antivax propaganda comes from a….vaccine manufacturer? Unfortunately, yes. The manufacturer of the Sputnik V vaccine is tweeting absolutely nonsense statistics in an effort to question the safety record of its competitors.

We need to be very honest with ourselves about what we know is true and what we hope is true.

@geoffrey_hurst @YouAreLobbyLud @DrDayaSharma @drvyom @drajm @enenbee @profmiketoole @PMGPSC @kiai @kenpcg @ketaminh @NjbBari3

We'd do well to consider Karl Popper's philosophy on falsifiability. We ignore black swans at our peril.

We're in a global emergency, we can't afford to assume best case and get it wrong. The caveat on superannuation returns applies "past performance is no guarantee of future performance".

Until we saw the emergence of: MRSA USA-300 and USA-400, and now (in my area of research) the emergence of hypervirulent (ST23 with K1 CPS) Klebsiella pneumoniae that produce ESBLs and carbapenemase. https://ecdc.europa.eu/sites/default/files/documents/Emergence-of-hypervirulent-Klebsiella-pneumoniae-ST23-carrying-carbapenemase-genes.pdf

There’s a lot of speculation that COVID may become less virulent. This is still an old concept in the field, akin to aerosol vs droplet. In my own field it was assumed that resistance came at the expense of virulence.

(PS, this is *even more* true if you replace “die” with outcomes like “hospitalized”, “long covid”, “have to spend several days in bed feeling like total crap”, etc. The more likely an outcome is, the MORE likely it is SOMEONE will experience it because you were infected.)

I’m seeing a lot of “these people are over-estimating risk” chatter that doesn’t acknowledge that the probability you die if you get covid is always less than the probability *anyone* dies if you get covid. It’s not “over-estimation” to consider community impacts.

Latest data from CDC on breakthrough cases (out of 115 million people fully vaccinated): - 1,359 hospitalized or fatal cases - 52% female, 79% over 65, 21% asymptomatic - 794 COVID-related hospitalizations - 181 COVID-related deaths https://cdc.gov/vaccines/covid-19/health-departments/breakthrough-cases.html…

this is important because it means that lab based studies looking at this tell you meaningful things about real world efficacy - and the lab based studies are much quicker than real world ones.Quote Tweet

President Biden announces the US will share doses of Pfizer, Moderna and Johnson & Johnson Covid-19 vaccines with other countries, in addition to 60 million doses of the AstraZeneca vaccine https://cnn.it/2RtpXwB

Yep. Also the crux of the controversial Swedish Strategy ...

4) Vaccinations are reducing covid case counts and are overwhelming the impact of the mask mandate being lifted

3) A social influence theory: Liberals were waiting on Fauci/CDC/NYT for permission to de-mask, while conservatives had long ago ditched theirs. Abbott took his cues from the latter, but that meant his edict responded to conservative behavior rather than guide liberal behavior.

Some possible interpretations: 1) Individual behavior is more important than state mandates: TX policy change didnt get pro-mask ppl to ditch their mask, and anti-maskers had already ditched theirs 2) Warm weather (& luck) made it less consequential to abolish mask mandates

Weeks ago, Gov. Abbott made Texas the first state to abolish its mask mandate and lift capacity constraints for all businesses. So, what changed? Nothing. There was ~no effect on COVID cases, employment, mobility, or retail foot traffic, in either liberal or conservative areas.

Fixed version: here's how the Imperial College model of Neil Ferguson performed over 1 year. I used their most conservative R0 assumption, so this is actually generous to them.

I do get why people want to shout at me. It totally sucks. But I think (again, personal opinion) we just know too little about B.1.617.2 except that it's spreading rapidly & the risks are quite high. Below is from this week's SAGE paper from Warwick. https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/984533/S1229_Warwick_Road_Map_Scenarios_and_Sensitivity_Steps_3_and_4.pdf… 5/5

Public Health England just released an updated report on B.1.671.2 . Cases more than doubled again in the last week (from 520 -> 1313). Looking at "S gene" detection as a proxy, B.1.617.2 might already be dominant in London & NW (SW is mostly traveller cases). 4/5

@markaustintv pointed out that people would be shouting at the telly hearing me suggest delaying Monday's opening... this is what I said about that. 3/5

I then also said what I, personally, thought that meant for next steps. Added to these must be much more support for local teams to beat outbreaks *and* proper financial & practical support for those who test positive & contacts. Once in place, could enable safer opening. 2/5

SHORT THREAD: I was on Sky News earlier where I explained why I thought test 4 (new variant test) for the next stage of the roadmap had not been met, because of B.1.617.2 (the so called "Indian" variant of concern). 1/5

What to do about it? Govt needs to keep communication why vaccines are important and that they are safe. AND we need to make it as easy as possible for people to get them - esp people unable to get during working day or unable to travel far. 5/5

Another general trend - all groups seeing lower coverage at younger ages. Even the highest group (least deprived) is not over 90% for the 50-54 group. Is this trend likely to continue as we vaccinate younger cohorts - particularly with worried about side effects? 4/5

By ethnicity: Much larger gaps for all age groups by ethnicity but less impact by age. Different but overlapping reasons driving ethnicity gaps compared to deprivation gaps? 3/5

By deprivation: Vax coverage gaps *widen* markedly as we move to young ager groups. This is not just a time effect - coverage has flattened off for all these age groups. Access (able to leave work to get vaccinated, travel, internet access) & communication likely issues 2/5

THREAD on VACCINATION & EQUITY in ENGLAND: I know I've tweeted about this before, but now we can look at how gaps by deprivation and ethnicity change with age groups and what that might mean... TLDR: widening gaps but access and communication will be key I suspect 1/5

NEW POLL: The J&J pause makes people *more confident* in vaccines, not less. M-O-R-E C-O-N-F-I-D-E-N-T

Insightful talk by @doctorsoumya @P4HR webinar on #VaccinePassports @WHO is developing smart Int'l Vaccine Certificates * Proof of vac * Confidential & Secure * Open Access * Interoperable But @WHO doesn't support requiring vacs for int'l travel until the world is more equal

What does 95% effective mean for a vaccine? We head to a stadium to learn! Warning: There are seagulls overhead! (Big thanks to @mariasundaram for help with this video!) Learn more about the #vaccines find family-friendly #coronavirus explainers at http://BrainsOn.org/Coronavirus

I cannot believe this headline is from May 7, 2021.

am I the only one discomfited by the fact that US teenagers are being vaccinated while an out of control pandemic rages in India and Nepal? I would have expected more discussion of this!

One year ago today: The WHO reports 81,454 new cases of COVID-19, and raises €7.4 billion for the Access to COVID-19 Tools Accelerator, an initiative to support development, production, and distribution for COVID-19 treatments and vaccines.

Downgrading the concern on B.1.617, the poorly named "double mutant" —98% effectiveness of mRNA vaccine in an Israeli outbreak @CT_Bergstrom https://twitter.com/CT_Bergstrom/status/1387941641395179524… —Lab studies: minimal immune evasion, expected full protection from vaccine @GuptaR_lab

Globally, the end of the pandemic isn't near. More than a million lives depend on improving our response quickly. Don’t be blinded by the light at the end of the tunnel. There isn't enough vaccine and the virus is gathering strength & speed. Global cooperation is crucial. 1/

A few people suggesting the answer is so obvious that the question may have been disingenuous. These sorts of answers say otherwise!

Ask any doctor or nurse in the NHS Why is this even a question?

Just hypothetically, if it were possible, if you could go back to January 2020 would you prefer to do what New Zealand did or what the U.K. did to ‘manage the pandemic’?

Given the dire situation in India and questions regarding the new variant B.1.617, the so called ‘Double Mutant’ we are sharing some prelim analyses on viruses with either or both of the mutations E484Q and L452R in the critical receptor binding domain that our antibodies target

I was very pleased to see Levitt resign yesterday from the science advisory board of the anti-vaxx group PANDA. Previously Sikora had resigned. This press release mentions other resignations. Anyone know if the 3 GBD authors finally resigned? Here's PANDA's views on vaccines:

this seems inevitable when a country accepts doses manufactured abroad while banning the export of doses (and raw materials!) from its own shores... I wonder how long international tolerance of this US policy will last

Meanwhile, in Europe...

In short: Bayes rule is very useful, and case/vaccine patterns in highly vaccinated populations don't always do what you may assume. 9/9

We can also flip the above equation around, which allows us to use data on % cases vaccinated and % vaccinated to get a rough estimate of vaccine effectiveness: https://twitter.com/AdamJKucharski/status/1382006630997323776?s=20… 8/Quote Tweet

This is an important result, because if cases appear among vaccinated individuals, many people's intuitive response is to ask 'surely the vaccine can't be that effective?' The answer: it may well be effective, just in a highly vaccinated population e.g. https://twitter.com/AdamJKucharski/status/1200329736544759808?s=20… 7/

Now we have something we can apply to real-life situations, because can measure many of these things. For example, if 60% of a population have been vaccinated, and vaccine is 80% effective, above means we'd expect (1-0.8)x0.6/(1-0.6x0.8)= 23% of cases to have been vaccinated. 6/

If we write out the ways in which we could get P(not protected), we end up with below equation (I've labelled the terms on the bottom of the fraction to make it clearer where these come from): 5/

Applying the above to our vaccine question, we therefore have: P(vaccinated | case) = P(case | vaccinated) x P(vaccinated) / P(case) which is equivalent to P(vaccinated | case) = (1–V) x P(vaccinated)/P(not protected) where V is vaccine effectiveness. 4/

If we want to know the probability of event A given event B, or P(A|B) for short, we can calculate this as P(A|B) = P(B|A) P(A)/ P(B) There are a couple more mathsy tweets coming up, so hold on as then we'll get back to the real-life implications. 3/

In above question, there are a lot of things happening conditional on other things happening (e.g. probability cases have been vaccinated), which means we can use Bayes rule (https://en.wikipedia.org/wiki/Bayes%27_theorem…) to work out the proportion of cases that we'd expect to have been vaccinated. 2/

If populations are highly vaccinated, we'd expect a higher proportion of future cases to have been previously vaccinated (because by definition, there aren't as many non-vaccinated people around to be infected). But what sort of numbers should we expect? A short thread... 1/

1/ To delay 2nd doses of the Pfizer and Moderna COVID vaccines (like the UK, Germany, & Canada)? https://usatoday.com/story/opinion/2021/04/08/covid-surge-deliver-first-vaccine-shots-delay-second-doses-column/7122747002/… Or not to delay? https://thinkglobalhealth.org/article/dont-be-seduced-two-shots-are-better-one… That is the question.

Seychelles is most vaccinated country in the world...but right now has more COVID cases per capita than India (!). How is that possible? Most vaccines used are Sinopharm. Which significantly reduces serious illness/death...but doesn’t do much to reduce transmission.

Boris's world beating for you.

@newscientistHmm this is strange, what could be the cause?

@newscientistWell, he's been aspiring to emulate a wartime prime minister. Looks like he's getting there.

@newscientistArticle seems lightweight on the stats. Where are the graphs with 5 year average. Deaths broken out by cause of death, age etc.

The UK has recorded the largest increase in excess deaths in the country since 1940 during the second world war https://bit.ly/3i4Y6vE

But isn’t it right that if you use age-standardised mortality which takes into account population growth and age the death rate is no higher than experienced in the 2000s?

@newscientistYes but Dave from Dorking says its a hoax

I think (hope) there might be a bit more possible, though: tools and standards to help us move *toward* greater objectivity - there's a continuum there

So true - so much so that there's a whole discipline devoted to the study of it, the Sociology of Scientific Knowledge. Watching it all play out on Twitter during Covid has been like watching SSK on speed. (Kudos to the social scis who address this as standard: "positionality")

The COVID Corps YouTube channel is live! Here's who we are and what we're about. New videos every Wednesday.

Apparently that's not the least of it. Now they've screwed up the correction as well. If 0.02% are diagnosed with COVID, that means 99.98%, not 92% like they claim, are COVID-free. The correction, shown below, is wrong in the story as well as in their mea culpa tweet.

Two days ago, Bloomberg @business @bopinion ran a story that overstated the risk of COVID to vaccinated people by something like 400-fold. Today they're tweeting this precise piece of misinformation. Vaccine skeptics are delighted. How can a reputable news agency be so sloppy?

Third time is the charm. Thank you Bloomberg @business @bopinion.

And unsurprisingly, the disinformers have also been busy. But don't worry our myth-busting page is calling out the bad stuff: https://hackmd.io/@scibehC19vax/misinfo_myths… 6/n

The fraught politics of COVID-19 and vaccinations. More details on our page: https://hackmd.io/@scibehC19vax/misinfo_politics… 4/n

The cultural sensitivities surrounding vaccinations have been explored further--with an additional section on the role of religion. Good news: all major faiths support vax https://hackmd.io/@scibehC19vax/vaxculture… 3/n

COVID-19 Vaccination Communication Handbook https://sks.to/c19vax : Multiple new updates to underlying wiki pages 1/n

Updates to public opinion about COVID-19 vaccinations: Good news, Europe has generally become more vax-positive https://hackmd.io/@scibehC19vax/publicattitudes… 2/n

And of course more and more facts roll in about the COVID-19 vaccines, for example real-life effectiveness (good news: it's very high!) https://hackmd.io/@scibehC19vax/c19vaxfacts… 5/n

In addition to successfully reducing shares by 24%, our intervention also reduced likes by 7%, and views by 5%. 6/7

We put a short prompt on videos that reminded people to think about the accuracy of the content they were watching. And then - when people went to share the video - we reminded them again that the video was flagged & asked them if they were sure they wanted to share. 3/7

Replying to @IrrationalLabsCool! Just out of curiosity, is the accuracy content what decreases sharing or is it the fact that you have increased the difficulty to share?

ooh this is actually really cool! Now when do we get to downgrade fake psychology facts

We designed an intervention that reduced shares of flagged content on TikTok by 24% via a large scale RCT, thread 1/7

Also, we may have succeeded in slowing people down and moving them from a 'hot' to 'cold' state. The extra question may have gotten people to pause for just long enough to reconsider their actions. 5/7

Congrats to the TikTok team for running an experiment and publishing the results publicly. The whole field benefits when we share learnings. 7/7

Replying to @IrrationalLabs@steverathje2 there’s hope

This intervention was inspired by previous research from @DG_Rand & @GordPennycook. People do value truth. An accuracy prompt has been shown to work because it reminds people about their own personal values of truth - at the critical point they are about to share something. 4/7

See article: https://techcrunch.com/2021/02/03/tiktok-to-flag-and-downrank-unsubstantiated-claims-fact-checkers-cant-verify/… 2/7

alongside dubious relationships with parties that in other contexts would require declarations of interest or that have independent hallmarks of being bad faith actors

Replying to @SciBehgood question, though I think anyone who genuinely believes they are acting for the greater good should welcome a public inquiry to make their case.

indeed! I suspect also, though, that for the most egregious cases of harm caused such an inquiry will be able to identify what are clear failings by *scientific standards* - such as cherry-picked data, selective reporting, unwillingness to admit error etc...

Dank der tollen Unterstützung von @TheRealTweetmo, Thomas Traill & Ulrike Hahn gibt es jetzt die deutsche Übersetzung des Anfang Januar erschienenen "COVID-19 Vaccine Communication Handbook" als "Kommunikationshandbuch zum COVID-19-Impfstoff" - https://sks.to/c19vax-de #COVID-19

Un petit livret sur les vaccins, très bien écrit par l'équipe @SciBeh, est téléchargeable ici : https://c19vax.scibeh.org/fr

experience feelings of vulnerability 3) Supervisory teams + institutional research culture is very influential. Recommendations include 1) Proper training, education +resources for ECRs on how to do OS is needed, 2) PIs, supervisors need to take holistic approach to ECR training

Replying to @ElaineToomey1 @johncoxnuig and 9 othersMight help. I am conducting an evaluation of the use of such a CV among researchers, panel reviewers and peer reviewers.

3) Institutions, funders, PIs need to review hiring/promotion/progression criteria, move away from impact factor/no. of pubs to more holistic criteria 4) Institutions, funders, PIs need to address culture of transparency across entire research cycle, not just data sharing.

Thanks to @jrpolanin and @ArdenClose for excellent #OpenPeerReview who made our paper so much stronger, to all participants and co-authors Ksenija Zecevic @houghtoncath @Chris_Noone_ @hopinlee @KarenMSikar

ECRs can 1) Get active - join/establish grassroots initiatives eg @ReproducibiliT @OpenSciUtrecht 2) Get vocal - use those platforms to engage w/ institutional management/ethics committees etc and 3) Get inquisitive - do research on research to find evidence-based solutions

Really thrilled our #mixedmethods study on factors influencing #openscience behaviours in early career researchers #ECRS has passed peer review! We found 1) ECRs Value OS but feel they are not fully supported to practice it 2) ECRs fear visibility of potential errors and

at the same time, I can see "average values" (say, estimated from a population) being used to make population level allocation decisions. BUT I cannot see how this could be used to *tell another individual* how much their life was worth. =>

Replying to @SciBeh and @STWorgand, I maintain that under, say, Art.1 of the German constitution it would be illegal to do so

Replying to @STWorgso one immediate response to that is that in expected utility theory, the standard normative framework for decision-making, utility is a *subjective quantity*, so it would ultimately be up to individuals to determine QALY values for themselves. =>

A while back I queried the wisdom of reputable academics publishing on the Toby Young created website http://lockdownsceptics.org. This tweet this morning by its creator, I feel, sheds further light on that issue

Replying to @SciBehThere is nothing wrong with making an error. But if your beliefs are based on a tenfold underestimation of risk they should change when that underestimation is corrected.

the rubric was developed in a hackathon from SciBeh's November 2020 conference (which you can revisit in videos and summaries here: https://scibeh.org/events/workshop2020/…) 2/3

new post on Scibeh's meta-science reddit describing the new rubric for peer review of preprints aimed at broadening the pool of potential 'reviewers' so that students could provide evaluations as well! https://reddit.com/r/BehSciMeta/comments/l64y1l/reviewing_peer_review_does_the_process_need_to/… please take a look and provide feedback! 1/3

Interested in the latest approaches to modelling the pandemic? Want to see what digital traces tell us about the public's response? Join us Tues, 16 Feb (5pm GMT) to hear Nicola Perra (@net_science) at the Tech + Democracy Seminar Series. Register here: https://cccmseminars.com/attend

interesting idea, but I think not applicable to what was described: certain types of posts are "being demoted" not blocked... - users still have the 'freedom' to seek out that content. They are just not being preferentially *served* that content.

Some might argue the moral dilemma is between choosing what is seen as good for society (limiting spread of disinformation that harms people) and allowing people freedom of choice to say and see what they want. I'm on the side of making good for society decisions.

"Well it's not a *moral* dilemma!" cry the academics as the leopard eats their faces

Replying to @Quayle @STWorg and 5 othersquite possibly! ;-) ...but they are invested in the idea that correct diagnosis raises the chance of successful intervention..... and, while the leopard might win on this occasion, on average that premise (which seems to me what science is all about) is likely true

Level up @zotero with these 2 plugins: - @scite: gives you 3 new cols: Mentioning, Supporting, Disputing (https://github.com/scitedotai/scite-zotero-plugin/releases…) - @PubPeer: Gives you a col with # of Pubpeer comments (https://github.com/PubPeerFoundation/pubpeer_zotero_plugin/releases…) HT @adam42smith

This week’s video is a little bit behind schedule, but we’re excited about it. Sneak peek:

Replying to @SciBeh @jayvanbavel and 5 othersindeed, that is the definition I am familiar with. Tobacco/Facebook may face a choice between profits vs. lives/democracy, but that is not a moral dilemma. The reason this is important is because good-faith actors can face agonizing moral dilemmas, but here we have something else

By definition "A moral dilemma is a situation in which a person is torn between right and wrong." In some cases, it can include competing moral virtues (e.g., deontology vs. utilitarianism). But that needn't be the case.

because we're all scientists, and precision matters: "Moral dilemmas, at the very least, involve conflicts between moral requirements. Consider the cases given below." https://plato.stanford.edu/entries/moral-dilemmas/… => definition of moral dilemma is a conflict between two moral imperatives

Extraordinary data from Scotland on excess deaths by cause and location in 2020 https://nrscotland.gov.uk/files/statistics/covid19/covid-deaths-data-week-53.xlsx… 6,686 deaths involving COVID-19 closely match 6,704 excess over 5-year average. But 570 *fewer* deaths than normal in hospitals, even with over 3,000 Covid deaths! ?????

so sorry about the typo Koudenburg, Koudenburg, Koudenburg!!!!!

Now speaking #scibeh2020: N. Koudenberg from Univ. of Groningen on her recent work with C. Roos and T. Postmes on comparing the navigation of disagreements in text-based online and face-to-face discussions

Translations are beginning to roll in. We have the French translation of the policy summary live here: https://hackmd.io/@scibehC19vax/lang-fr… @SciBeh @mariejuanchich List of forthcoming languages here: https://hackmd.io/@scibehC19vax/lang

NYU too I believe. What is happening in Canada and Australia, especially regarding international students.

this is utterly bizarre : how would one conceptually even begin to determine a number by which the model *overestimated unmitigated deaths*. What is the comparison unmitigated "prediction" to what actually happened supposed to mean?

"Unmitigated" appears to mean "if the government didn't do anything", compared to lockdowns.

so, given that no one can know the "unmitigated number" what they seem to be calculating is in difference deaths given lockdown and model prediction without lockdown and calling that the "overestimate" - which seems truly bizarre

@SciBeh and @STWorgat the same time, I can see "average values" (say, estimated from a population) being used to make population level allocation decisions. BUT I cannot see how this could be used to *tell another individual* how much their life was worth. =>

Very good question. I think that economists may have something to say about that (or to answer for?) https://en.wikipedia.org/wiki/Quality-adjusted_life_year… QALYs are a quantitative manifestation of this AFAICT. Eugenics dressed up as cost benefit analysis? Concerning but not entirely clear.Quote Tweet

so one immediate response to that is that in expected utility theory, the standard normative framework for decision-making, utility is a *subjective quantity*, so it would ultimately be up to individuals to determine QALY values for themselves. =>

Third presentation in our "Open Science and Crisis Knowledge Management" panel - Iratxe Puebla from ASAPbio about preprints during the pandemic #scibeh2020

Our first panel session #scibeh2020 of Day 2 will focus on the challenges of building sustainable, transparent, and constructive online discourse among researchers as well as between researchers and the wider public.

A nice insight piece from @RTI_Intl about lessons learnt estimating Rt for #covid19. https://rti.org/insights/pandemic-metrics-estimating-rt… Helpful to read with my scientist hat on in order to understand the problems with current tools.

Summary of severity estimates for B.1.1.7 Important to distinguish whether we're talking about risk of death/ICU/hospitalisation following a positive test (which is estimated to be higher in many studies), or risk of death/ICU following hospitalisation (which is not).

Vacancy for a behavioural scientist in our team at the Joint Research Centre of the European Commission https://recruitment.jrc.ec.europa.eu/?site=BRU&type=AX… #behaviouralscience #behaviouralinsights

great presentation by Michele Starnini just now which proposes a radical overhaul for publishing and reviewing focussing on creating new incentive structure- thanks Michele!

New book "Urban Informatics", #OpenAccess https://lnkd.in/gXUQecE Edited by @jmichaelbatty @CUHKofficial Michael Goodchild, Wenzhong Shi and Anshu Zhang Chap. in Urban Mobility Patterns: A Case Study of @MexicoCity with @FahadAlhasoun @ @jlmateos et. al

I wrote about the outdoor mask mandates. https://theatlantic.com/ideas/archive/2021/04/are-outdoor-mask-mandates-still-necessary/618626/… Masks are good; and inside, among non-vaccinated ppl, they're great. But governments need to give Americans an off-ramp to the post-pandemic world. Ending outdoor mask mandates would be a good place to start.

Are you still hesitant about AstraZeneca vaccine? Check the below video! #publichealth #vaccine #sciencework

it could be meaningful only vis a vis certain qualitative constraints: e.g., "look, model predicts fewer deaths for unmitigated than observed even *with* lockdown" => model underpredicts.... but that's very much not the scenario here

.@noUpside: U.S. disinformation researchers have focused on the election and COVID-19 vaccines because of the potential for significant public harm in these areas.

#COVID19: Known Unknowns (Facing up to scientific uncertainty during a pandemic) Join a webinar hosted by @fgodlee & @mendel_random with @mrc_ieu Find out more and register here: http://ow.ly/RDRi50CdSON

Important comment by @_nickdavies here: new studies are compatible with B.1.1.7 having higher severity because new studies only consider in-hospital severity. If variant has higher hospitalisation rate, but same outcome within hospital, variant is more severe overall

In today's much-discussed @nytimes story from @apoorva_nyc (https://nytimes.com/2021/05/03/health/covid-herd-immunity-vaccine.html…) there is a graph that I find quite problematic. It purports to show county-level data about vaccine hesitancy.

Report that "SA" variant considerably reduces the Oxford/AstraZeneca vaccine's efficacy https://twitter.com/janinegibson/status/1358151228203683854

...Latest data for standard doses: 63% (52-72%), with some possible lowering from "UK" variant (maybe "SA" too?). So many moving parts! US trial could be similar, but with much greater certainty - or lower. Could be higher. 8/8 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3777268…, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3779160

..Then there's placebo control. A trial in Cuba is testing hypothesis that meningococcal vaccine boosts innate immunity & protects against Covid-19 https://rpcec.sld.cu/en/trials/RPCEC00000314-En… *Long* shot, but if it does at all, that control in the other trials may have reduced apparent efficacy..

...The diversity should be higher - eg the UK trial had <1% Black people & it's 10% in the US trial: that could influence efficacy too. (Everyone calls it "the US trial", but there are some participants from Latin America too, though not Brazil)...

...UK/Brazil 90% healthcare & social workers: won't be in US. https://thelancet.com/journals/lancet/article/PIIS0140-6736(20)32661-1/fulltext… Could have an influence, eg even increasing efficacy: hypothesis for CoronaVac was HCWs depressed efficacy by diagnosing many very mild infections: perhaps? (see https://absolutelymaybe.plos.org/2021/01/31/variants-3-new-covid-vaccines-and-contested-efficacy-claims-a-month-of-seismic-shifts-and-confusion/#sinovac…)...

I fear that AZ resistance will grow following the decision by the Fr govt and medical advisory board that under 55s who have had a dose of AZ should receive a 2nd dose of another vax. This goes against WHO guidelines but the French board says there should be no problem.

Stats. Figures for use of doses in France do now show a clear resistance to AZ – and to a lesser extent Moderna. As of Thurs, by my calculations, Pfizer doses were 85% used, Moderna 69% and AZ 61%. Overall French utilisation rate is now 78%, slightly down on last week.

Thirdly, this is not just a French success. Most other EU countries are also vaxxing at high rates now that large supplies of Pfizer and to a less extent Moderna are available. Johnson and Johnson starts its one-jab rollout in the EU on 19 April.

Secondly, the vax rollout is beginning to encompass high proportions of older age groups. Over 62% of over 75s have had at least 1 jab and 34.2% have had both. For 65-75’s the figures are 38% and 6%. This should begin to reduce acute cases soon.

Now good news…. There are signs that the UK-variant fuelled 3rd wave of Covid in France is peaking. Some stats have been muddled by glitches but a plateau of 700 cases/100,000 people/7 days seems to have been reached in the worst-afflicted areas, Gr. Paris and the N-west.

We halved all public health funding between 2012 and 2020, to a paltry £3 billion per year, even though a pandemic has consistently been the biggest threat to the country, higher even than war. We spend 13 times more on defence than PH. And 400x more than on infection control

We could have done what they did in one province of China (which is much poorer than us) if we had invested in strong public health, included independent PH advisers on SAGE, and mobilised districts and communities. We still aren't doing it.

The UK wasted £22 billion on a system not run by public health experts. We had to borrow £373 billion even before the second surge. The loss of GDP and future unemployment suggest the impact of the pandemic could cost us a sum of more than a trillion pounds.

Now look at the outcomes in Anhui province, the same size as the UK. They achieved elimination of new cases in six weeks.

Now read the interventions they arranged within a week of identifying the first case. Then read it again and consider how many we did here in the first wave, and how many we are doing now a year later? Yes, its embarrassing

The devolved administration identified 991 cases in January and February last year. They also tracked down 29,399 contacts through local community based workers. Note the ratio: they detected 30 contacts for every case, far higher than in the UK.

We halved all public health funding between 2012 and 2020, to a paltry £3 billion per year, even though a pandemic has consistently been the biggest threat to the country, higher even than war. We spend 13 times more on defence than PH. And 400x more than on infection control

We could have done what they did in one province of China (which is much poorer than us) if we had invested in strong public health, included independent PH advisers on SAGE, and mobilised districts and communities. We still aren't doing it.

The UK wasted £22 billion on a system not run by public health experts. We had to borrow £373 billion even before the second surge. The loss of GDP and future unemployment suggest the impact of the pandemic could cost us a sum of more than a trillion pounds.

Now look at the outcomes in Anhui province, the same size as the UK. They achieved elimination of new cases in six weeks.

Now read the interventions they arranged within a week of identifying the first case. Then read it again and consider how many we did here in the first wave, and how many we are doing now a year later? Yes, its embarrassing.

Ioannidis' exclusion fits with him under-estimating IFR by using non-representative samples in areas that under-estimate COVID deaths. The WHO + the USA's CDC know better, and so rely on Levin et al.: https://web.archive.org/web/20210324195745/https://www.cdc.gov/coronavirus/2019-ncov/hcp/planning-scenarios.html… https://twitter.com/AtomsksSanakan/status/1374617361194565634… https://link.springer.com/article/10.1007/s10654-020-00698-1

With that framework in place, let's start with the page-by-page review of Ioannidis' paper: Ioannidis excludes @GidMK + @BillHanage's paper Levin et al., because it focused on specific countries. https://twitter.com/AtomsksSanakan/status/1336442679689965570… https://link.springer.com/article/10.1007/s10654-020-00698-1… https://onlinelibrary.wiley.com/doi/10.1111/eci.13554

So in this thread, *keep this in mind*: Ioannidis has to keep non-representative samples in, because representative samples show an IFR incompatible with his position. That's his main game, + what he often distracts from https://twitter.com/AtomsksSanakan/status/1375343648129359880…

- he could just wait for representative sampling in less hard hit areas - areas often looked less hard hit because they under-estimated COVID-19 deaths, so including them under-estimates IFR etc. https://bmj.com/content/370/bmj.m2859… https://twitter.com/AtomsksSanakan/status/1369873446642069504… https://medrxiv.org/content/10.1101/2021.01.27.21250604v1.full-text…

Ioannidis defends his use of non-representative samples. But his defense fails. For example: - non-representative samples are still unreliable - he uses non-representative samples even in hard hit areas https://twitter.com/AtomsksSanakan/status/1341285257644027904… https://who.int/bulletin/volumes/99/1/20-265892/en/

8. Gambler’s fallacy Arguably the odd one in the list, but cognitive biases about probabilities of recurrent events are very real and relevant https://en.wikipedia.org/wiki/Gambler%27s_fallacy

7. Prosecutor's fallacy Pr(B|A) is not Pr(A|B). Confusing sensitivity/specificity for predictive values, p-values for probabilities about the hypothesis,.... the prosecutor's fallacy list is long

6. Berkson’s paradox Also known as collider bias, something we have seen plenty in the COVID-19 literature

5. Simpson’s paradox Perhaps one of the most famous paradoxes in statistics. Reversal of the direction of effect by simply combining two groups is something that may keep awake at night

4. Lord’s paradox Between group comparisons with baseline and follow-up: analysis of change scores or ANCOVA? Doesn't matter? Well it does...

He spends quite a bit of time on my and @LeaMerone's paper, arguing that we “cherry-picked” evidence to suit our conclusions and that our analysis methods are “overtly implausible”

The author looks at each review and discusses his view on their limitations and successes, then concludes that the best estimate is his own

So I don't know if the primary purpose of this paper makes sense But what is it exactly? Well, it’s mostly a review of systematic reviews

In this thread, the researcher in question, @GidMK, offers his thoughts on the whole affair.

Honestly, the biggest takeaway in all of this is that when you're reviewing a paper, you can't afford to merely skim the appendix.

On the other side of the coin, there’s evidence that in some countries that the death figures from COVID-19 may underestimate the true toll by an order of magnitude (or more!) https://bmj.com/content/372/bmj.n334

For example, this recent systematic review of seroprevalence studies found that even after including more than 400 pieces of research total there was insufficient evidence to infer a truly global estimate

Anyway, you can read it for yourself. It's published in the journal for which Ioannidis previously served as Editor in Chief. https://onlinelibrary.wiley.com/doi/10.1111/eci.13554… Therein John claims the IFR for COVID is 0.15%. By official counts, 0.166% of the US population has already died of COVID.

They go on to explain that these aforementioned points led to yesterday's decision by the @PEI_Germany to recommend suspension of the #AstraZeneca COVID-19 vaccine as a precautionary measure until further assessment can be completed.

The PEI states that they consulted other experts in thrombosis, hematology, and an adenovirus specialist about this issue & all unanimously agreed that a pattern could be recognized here & a connection between the reported cases and the AZ vaccination "was not implausible".

The younger to middle-aged population in which these severe cerebral venous thromboses+platelet deficiency was observed is not the group at highest risk for a severe/fatal COVID-19. (Sidenote: this is why so many of us wanted information on age group yesterday)

The PEI ran an "observed-versus-expected analysis", comparing the # of such cases expected *without* vaccination in a 14-day period with the # of cases reported in those vaccinated with #AstraZeneca (1.6 million) in Germany. About 1 case would be expected, 7 were observed.

The PEI writes that this number of cases following #AstraZeneca vaccination was found to be statistically significantly higher than the number of cerebral venous thromboses that normally occur in the general population.

and "search" is just one way of conceptualising/pinning down the likelihood P(no evidence|Hyp_false). See e.g., here https://books.google.de/books?hl=en&lr=&id=6Q-NS7CUTF0C&oi=fnd&pg=PA121&dq=Hahn+u+inference+from+absence&ots=bTWmnEjvL3&sig=xyqByJJAQy61MZEiq8BvdylGmlU&redir_esc=y#v=onepage&q=Hahn%20u%20inference%20from%20absence&f=false… similarly, calculating BFs, it's the prob. of *actual data* under null that we are using, not some further new evidence item "search"

I now see what you are aiming at! I disagree, though: it seems odd to take the search itself to be the evidence, given that it is *result* of the search that determines the evidential impact, not search per se. I can search throughly and fail, or search and find something.

But if your focus is on evidence instead of the research object than @richarddmorey is right, but misses that his search already is evidence and that he only refers to the evidence and not the research object.

I can't agree with the premise that it is fals: Take the first assumption stating Corona virus. There was no evidence yet that it exists just an assumption, but this did'nt meant it does not exist. I am no nativ english speaker so maybe I miss something in translation.

At this point in the pandemic, can't believe we're still talking about cases rising due to testing That's not what's happening in America Dispelling disinformation is getting boring I know But that's where we are We're seeing more infections Mask up Fin

17 states are testing LESS than 2 wks ago In 16 of those 17 states, cases are up! So in 16 states, fewer tests but more cases Could not happen if this was about testing Finally, remember the lag? Hospitalizations up in 44 states, deaths in 34 states

while I agree with that sentiment in principle, I think there are really interesting and important questions about what counts as "an expert" and expertise in a crisis like this. Outside/fresh perspectives can sometimes be valuable when coupled with willingness to learn 1/2

I don't think we should be paying attention to dilettantes when there are experts.

How has #COVID19 changed clinical research in global health? The final paper in @LancetGH’s new Series discusses the fundamental issues in clinical trial research exposed by the pandemic and provides recommendations. Read http://hubs.li/H0LlS5Y0