1,274 Matching Annotations
  1. Last 7 days
    1. COVID-19 test positive and died in hospitals in England (up to 2nd Sep) - Under 60 with no pre-existing conditions 305 deaths in a popn of 56 Million.
    1. 9) Household studies can be very useful here, where we follow the household contacts of infected vaccinated individuals and infected unvaccinated individuals, and compare how frequently they are infected. Pioneering work from my mentor @betzhallo. http://courses.washington.edu/b578a/readings/PreziosiHalloranVaccine2003.pdf…
    2. 10) Finally, vaccine efficacy vs. effectiveness? We like to reserve "efficacy" for estimates from randomized trials, where everyone receives the vaccine as intended (proper cold chain, no missed doses). We distinguish this idealized measure from real-world "effectiveness." END
    3. 8) A vaccine that prevents infection entirely provides indirect protection to others. If I can't get infected, I can't infect you. But it is possible to have a vaccine that prevents disease but individuals can still be infectious.
    4. 7) So far I have talked about how well a vaccine directly protects the vaccinated individual. Another important type of vaccine effect is the ability of a vaccine to reduce infectiousness to others. This is known as indirect protection, and is related to herd immunity.
    5. 6) Preventing infection entirely is the hardest to achieve. And of course a vaccine that prevents infection will also prevent disease and severe disease. But we can have vaccines where people are still infected but their disease severity is lessened.Quote Tweet
    6. 5) Most Phase 3 trials are measuring efficacy to prevent disease as the primary analysis, with efficacy against infection and against severe disease as secondary analyses.Quote Tweet
    7. 4) Though we talk about vaccine efficacy as a single number, there are actually several different types of vaccine efficacy, such as: - Efficacy to prevent infection (sterilizing immunity) - Efficacy to prevent disease - Efficacy to prevent severe disease
    8. 3) Vaccine efficacy of 50% roughly means you have a 50% reduced risk of becoming sick compared to an otherwise similar unvaccinated person. Or you have a 50% chance of becoming sick given that you were exposed to enough infectious virus to make an unvaccinated person sick.
    9. 2) Vaccine efficacy (VE) measures the relative reduction in infection/disease for the vaccinated arm versus the unvaccinated arm. A perfect vaccine would eliminate risk entirely, so VE = 1 or 100%. This can be calculated from the risk ratio, incidence rate ratio, or hazard ratio.
    10. VACCINE EFFICACY 101: A biostatistician's primer Ten tweets to cover: - How is vaccine efficacy calculated? - Distinguishing between infection, disease, & severe disease. - Measuring reduced infectiousness. - Vaccine efficacy vs. effectiveness!
    1. Members of host communities are also part of this experiment, of course. (We should have mentioned this above.) University leaders haven't sought their consent either.
    2. But university leaders, including at @uniofleicester, are neither informing us of the risks, nor seeking our consent to participate in this experiment. (This important point was just made by a member in our emergency general meeting.)
    3. Universities are conducting an experiment, an experiment that involves human beings (university staff and students) and a life-threatening virus. But experimental subjects must give informed consent. (That's basic research ethics.)
    1. Finally, I’d like to correct the ByLine article where it mentions me (in relation to the March claims by Prof Gupta above): I’m not the head of the UCL Institute for Global Health, though I do work there. 8/8
    2. I am a signatory of the other letter: https://twitter.com/martinmckee/status/1308043860652830721?s=20… One the positive side both letters do agree that the NHS should be kept open for all The disagreement is in how much of a risk Covid is and how we deal with it. 7/8
    3. b)It doesn’t acknowledge the risks of long-Covid c)It claims no link between restrictions and mortality 6/8
    4. So I question the scientific judgement of the Sikora-Gupta-Heneghan letter, and probably should just stick to that: a)The letter offers no practical way to separate low and high risk given the reality of multi-generation households, the workforce and society more broadly 5/8
    5. With others I challenged that at the time as dangerous: https://ft.com/content/ebab9fcc-6e8d-11ea-9bca-bf503995cd6f… 4/8
    6. There is relevant history with many of the authors of the letter including Prof Gupta who claimed in March without any evidence that 50% of the UK had already been infected: https://ft.com/content/5ff6469a-6dd8-11ea-89df-41bea055720b… 3/8
    7. I think the issues raised in the article are worth discussing and not comparable with Gates “mind-control vaccines” and am glad that some people have engaged with them. 2/8
    8. Some reflections on this: Firstly, I can see the article uses inflammatory language so apologise for that and any offence caused. They were genuine questions in my Tweet though and I was not expecting to be called deranged for posting it. 1/8
    1. Below is a 3 group WAIFW with different Pr of connections in each group (by color) and between groups. Essentially WAIFW is a stratified random mixing model
    2. Because I did a bad job of defining WAIFW: it is the Who Acquires Infection From Whom matrix. Rather than random edges between persons, WAIFW generates edges random at \beta_w for connections in the group and \beta_o for outside
    3. Ultimately, the (possibly) unobserved network matters. If your model only assumes some random mixing parameter for broad groups of people, your model is probably way too simple to say anything meaningful about what policies we should actually consider
    4. If your strategy is instead to vaccinate the persons with household-only contacts then you will have to go above the threshold
    5. If your strategy is to vaccinate the between-household contacts, you can be lower than the threshold
    6. Network 4: Household Network This last network is variation on the previous clustered network. It instead places people into households. Depending on the vaccination strategy used, we can go above or below the 'threshold'
    7. This network is closer to how human interactions are actually structured. This is also closer to WAIFW setups (but WAIFW won't capture the power-law for degree within clusters that the above network has)
    8. Network 3: Clustered Network In a clustered network, we can go above the herd immunity threshold but still have outbreaks. The following network has 75% vaccinated but an outbreak would occur
    9. "Paul, all you have shown is that you can effectively go below the threshold. That doesn't negate the whole concept. We might instead say we need *at least* the threshold" Bad news my imaginary interlocutor
    10. Network 2: Mesh Network Well it looks like a big ole net. We can guarantee no transmission with only 50% vaccinated. Less than the supposed threshold
    11. Network 1: Random Mixing When someone talks about a threshold for herd immunity, this is the underlying network of what they are generally talking about (setting aside WAIFW for the moment). The threshold calculation applies normally
    12. In all the following networks, the elements of R_0 are held constant. The network structure will change; but the \beta look the same despite that Our R_0 will be 3, meaning the herd immunity threshold is 0.67 (but everything applies to other R_0 > 1)
    13. To show this, let's talk about a perfect vaccine. If you get this vaccine you are perfectly protected from the infection and thus cannot transmit it (everything also applies to imperfect vaccines but it's messier) Blue circles are vaccinated individuals and red are unvaccinated
    14. where \beta is the effective contact rate, N is the number of individuals, and r is the inverse of the duration The threshold says if are above that level the disease will disappear / we expect no outbreaks of disease. However, that threshold is neither sufficient nor necessary
    15. Herd immunity is a far squishier concept then many seem to be describing in their "shielding" or "stratified herd immunity" plans. Here is the formula for herd immunity threshold for a SIR model
    1. SciBeh is organising a workshop on "Building an online information environment for policy relevant science" Mark the date, Nov. 9/10, 2020, join us, contact us with thoughts and suggestions, and RT!
    1. The U.K. map gets redder as daily cases reach 14000 and R=1.4 . Our data are over double govt figures based solely on testing. Worrying to see London cases accelerating for the first time since April and student mobility not going to help. Keep logging!
    1. RT @AlecStapp: New research finds that fake news comprises "only 0.15% of Americans' daily media diet." Researchers argue that ordinary ne…
  2. Sep 2020
    1. This thread is based on our new paper with @JuliaLMarcus Caroline Buckee and @aetiology Accepted in CID - preprint version can be accessed here: (https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3692807…)
    2. Environment: Contact pattern also depends on the setting of the encounter. Contact tracing studies suggest an almost 20x higher risk of transmission indoors compared with outdoor environments. (10/n) (https://ft.com/content/2418ff87-1d41-41b5-b638-38f5164a2e94…)
    3. void crowded indoor poorly ventilated environments. Spend more time outdoors. Maintain your distance (more is better but 2 metre is not a panacea). Improve ventilation: open windows/doors. Wear a mask indoors. Wash hands. (27/n) (https://vox.com/science-and-health/2020/5/22/21265180/cdc-coronavirus-surfaces-social-distancing-guidelines-covid-19-risks…)
    4. Policymakers and health experts can help the public differentiate between lower-risk and higher-risk activities and environments and public health messages could convey a spectrum of risk to the public to support engagement in alternatives for safer interaction (26/n)
    5. There are many things that could be done within families to decrease transmission. We need to provide clear instructions, and means of support to enable those with symptoms/positive test and their contacts to isolate. (25/n) https://abc.net.au/news/2020-09-15/coronavirus-swept-through-jos-house.-heres-how-he-dodged-it/12660218
    6. Early viral load peak in the disease course indicates that preventing onward transmission requires immediate self-isolation with symptom onset (for a min of 5 days). Messages should prioritise isolation practices, and policies should include supported isolation. (24/n)
    7. In summary: The disproportionate impact of COVID-19 on households living in poverty, and the racial and ethnic disparities observed in many countries, emphasize the need to urgently update our definition of "vulnerable" populations for COVID-19 & address these inequities. (22/n)
    8. A real overlap in the causes of mortality and deprivation can be seen here. The age-standardised rate of deaths involving COVID-19 in the most deprived quintile was more than double (2.3 times higher) than in the least deprived quintile in Scotland. (21/n) https://nrscotland.gov.uk/files/statistics/covid19/covid-deaths-report-week-19.pdf
    9. Covid-19 could now be endemic in some parts of England that combine severe deprivation, poor housing and large BAME communities, national lockdown in these parts of the north of England had little effect in reducing the level of infections (20/n) (https://theguardian.com/world/2020/sep/05/covid-19-could-be-endemic-in-deprived-parts-of-england//…)
    10. Previous research suggests that although social distancing during the 2009 H1N1 pandemic was effective in reducing infections, this was most pronounced in households w greater socioeconomic advantage. Similar findings are emerging for COVID-19. (19/n) (https://pnas.org/content/117/33/19658//…)
    11. In Madrid, 37 neighbourhoods are seeing the highest incidence, 4 x the Spanish average. Common factors: these areas are poorer, denser and have a high proportion of immigrant population. (18/n) (https://elpais.com/sociedad/2020-09-19/como-son-las-zonas-restringidas-en-madrid-mas-densas-con-mas-inmigrantes-y-sobre-todo-mas-pobres.html…)
    12. PHE surveillance report shows that while the number of infections is increasing mainly in 20-29, 30-39 ages in England, SARS-CoV-2 is spreading most in highly deprived areas - where people are in poorly paid work and can't afford to isolate. (17/n) https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/919676/Weekly_COVID19_Surveillance_Report_week_38_FINAL_UPDATED.pdf
    13. Households in socioeconomically deprived areas are more likely to be overcrowded, increasing the risk of transmission within the household. These disparities also shape the strong geographic heterogeneities observed in the burden of cases and deaths. (16/n)
    14. People in lower-paid occupations are often classified as essential workers who must work outside the home and may travel to work on public transport. These occupations often involve greater social mixing, exposure risk due to prolonged working hours and job insecurity. (15/n)
    15. The largest clusters of cases observed in the USA have all been associated with prisons or jails. In the largest meat packing plant in Germany, while the common point of potential contact was workplace, risk was higher for a single shared apartment, bedroom and carpool. (13/n)
    16. Much worryingly the largest outbreaks from across the world are reported in long term care facilities such as nursing homes, homeless shelters, prisons, and meat-packing plants where many people spend several hours working, living together, and share communal spaces. (12/n)
    17. Prolonged indoor contact in a crowded and poorly ventilated environment increases the risk of transmission substantially. But decreasing occupancy and improving ventilation through opening windows/doors can lower the risk. (11/n)
    18. Environment: Contact pattern also depends on the setting of the encounter. Contact tracing studies suggest an almost 20x higher risk of transmission indoors compared with outdoor environments. (10/n) (https://ft.com/content/2418ff87-1d41-41b5-b638-38f5164a2e94…)
    19. Prolonged indoor contact in a crowded and poorly ventilated environment increases the risk of transmission substantially. But decreasing occupancy and improving ventilation through opening windows/doors can lower the risk. (11/n)
    20. hen we look at the viral load dynamics & contact tracing studies, those who are infected are very infectious for a short window, likely 1-2 days before and 5 days following symptom onset. No transmission documented so far after the first week of symptom onset. (7/n)
    21. Individual factors: Many ppl either do not infect anyone or infect a single person, and a large number of secondary cases are caused by a small # of infected ppl. Although this also is related to other factors, individual variation in infectiousness plays a major role.(6/n)
    22. This encapsulates the problem nicely. Sure, there’s a paper. But actually read it & what do you find? p-values mostly juuuust under .05 (a red flag) and a sample size that’s FAR less than “25m”. If you think this is in any way compelling evidence, you’ve totally been sold a pup.
    23. One amazing way to do this would be to control the pandemic; Venmo me.
    24. universities and everything else, it becomes more and likely. 13. Yes, most of this is my own opinion, I genuinely hope I'm wrong about the winter...
    25. is good for shielders, the elderly or the economy. 12. No, we can't say for certain that there will be resurgence of infections in the winter. However, without minimising community transmission over the summer combined with increased social mixing involving schools, pubs, work
    26. protect you better. 10. Yes, you CAN get on top of this with non-pharmaceutical interventions combined with rigorous test, trace, isolate, test again... 11. If we don't get on top of it, we'll have cyclical limbo of local lockdowns and confused unlocking. I can't see how this
    27. they are not a conspiracy and yes you should have it when its available. Antivax is an act of self-harm for humanity. 9. Yes, face masks worn properly do help limit spread. No, they won't protect you personally very well, but better than nothing. Surgical masks and respirators
    28. 7. Yes, kids can catch SARS2, no they don't often get very unwell, but yes they can certainly spread it. 8. No, we don't know whether any vaccine will work, but there are some amazing efforts ongoing. No, herd immunity is NOT a viable option and it hasn't happened anywhere. No,
    29. pick up "dead RNA" as I've seen mentioned. Honestly... 6. Yes, the NHS is struggling to catch back up after the peak. This is NOT to do with being distracted, wrongly tasked, or overly cautious. It is due to being chronically underfunded and understaffed for a decade. Simple.
    30. lower numbers in parts of UK keep it around 1 (or slightly higher in some parts). R0 is only 3 when you don't intervene! 4. No, the virus is not getting "weaker". It is infecting younger, healthier people better able to cope. 5. No, tests are not wildly inaccurate, they don't
    31. 2. Yes, there are fewer hospital cases and fatalities. This is proportionate to infections, plus more younger people infected. Care homes are better protected (finally) and most shielders did NOT pause, I suspect. 3. Cases are increasing, as is R0, but regional variation and
    32. Upsetting to see so many half-truths, dismissive crap and bizarre media conspiracies floating around...sorry, have to get this off my chest. 1. No, of course there aren't as many infections as in spring. We had a lockdown, albeit truncated, and most people still distance...
    1. some of them are just kids being kids, making stuff up and playing into taboos. but unfortunately deep state conspiracy theories have become so centralized and weaponized, that intentions don’t matter, the content is all one big soup of viral misinformation
    2. Even where the stuff they are sharing is found somewhere else, they jump on it fast. I think I first saw the topographical map sea floor Epstein tunnels conspiracy on it, b/c some young conspiracy entrepreneur was just on top of that.
    3. One thing to put things in perspective on this -- I dig through a lot of conspiracy professionally. It's hard to introduce me to new ones. But I flipped through tiktok for an hour (w/ a little bit of algo training via bookmarking) and was finding stuff I'd never seen.
    4. Amidst many college outbreaks are a slew of very successful schools. My employer, @Yale, is among them: I hope they stay that way. Congrats to the students, faculty, staff, & our community for working together to achieve ZERO positive results in last 7 days. (9,425 tests).
    5. Amidst many college outbreaks are a slew of very successful schools. My employer, @Yale, is among them: I hope they stay that way. Congrats to the students, faculty, staff, & our community for working together to achieve ZERO positive results in last 7 days. (9,425 tests).
    6. Diagnosis of COVID at admission is made by clinicians, now experienced in what the disease looks like. PCR tests are used to support (or otherwise) their clinical diagnosis. The rise is new cases coming into hospital is very real and being seen across the country.
    7. This needs to be accounted for in some reporting figures, particularly when patients are readmitted to hospital (see , for example, our piece on this from June http://bit.ly/3hSmlvc). But, this is NOT the reason for increasing cases or admissions. 2/3
    8. There's a fair bit of misinformation doing the rounds today. One relates to a clear misunderstanding of PCR testing and false positives. It is true that some of those with COIVD can have positive PCR tests for weeks, sometimes at low levels, long after they're better. 1/3
    1. older generations spreading conspiracy theories and misinformation on facebook has been reported for years. what people don’t realize is the same issues persist with kids today on tiktok, despite the widespread narrative that they’re immune to fake news from growing up online
    2. i wrote a longform article about how conspiracy culture became so viral today. no demographic is immune. we all have a little conspiratorial thinking in us—that’s what makes ideological cults about the “deep state” so appealing and terrifying
    3. gen z will make a conspiracy theory out of ANYTHING
    4. the state of conspiracy theories on tiktok cannot be understated. these kids believe ANYTHING and spread misinformation like wildfire
    1. Without a deeper understanding of the connection between contradiction and truth, and some basic appreciation of reasonable discourse rules, we cannot possibly have meaningful critical exchange between scientists and general public. and that won't happen over night... 3/3
    2. and, in fact, seem often to view the fact that an exchange about their views has backed them into a contradiction as indication that one is just "interested in winning the argument" as opposed to getting at truth. 2/3
    3. having spent a few days looking at 'debate' about COVID policy on lay twitter (not the conspiracy stuff, just the "we should all be Sweden" discussions), the single most jarring (and worrying) thing I noticed is that posters seem completely undeterred by self contradiction 1/3
    1. Because Science is a horrible journal that doesn't even let you read the abstract, here it is.
    1. No Imperial model points to 85,000 deaths in Sweden. This cites work by other researchers. Imperial had no role in developing this model, its assumptions or its parameters.
    2. Imperial’s COVID-19 Response Team reports and tools, including models, can be accessed on our website
    3. We’ve dealt with similar misleading claims before
    4. Neil Ferguson and Imperial did not produce a model for Sweden pointing to 85,000 deaths
    1. The pandemic proves we all should know ‘psychological first aid.’ Here are the basics. /lifestyle/wellness/pandemic-psychological-first-aid-anxiety/2020/09/21/7c68d746-fc23-11ea-9ceb-061d646d9c67_story.html?tid=ss_tw
    1. Additional note on French data. The challenge (and one that's always an issue in real time) is that the lags resulting in updates to earlier data also mean that recent numbers may be revised upwards in future.
    2. Quick clarification (thanks for flagging @marbin2050): the report of 239 deaths in Spain was from a day earlier (https://english.elpais.com/society/2020-09-17/spain-reports-239-new-coronavirus-deaths-the-highest-figure-so-far-during-the-second-wave-of-the-epidemic.html…) – 162 were reported yesterday (https://english.elpais.com/society/2020-09-18/spain-reports-11291-new-coronavirus-cases-and-162-deaths.html…)
    3. I presume model is based on this recent paper, which predicted a second peak at ~30 deaths per day (https://medrxiv.org/content/10.1101/2020.09.01.20185876v1…). For context, there were 27 COVID deaths reported in UK yesterday: https://coronavirus.data.gov.uk/deaths
    4. Spain reported 239 COVID deaths yesterday & France 154. So seems odd for this model to predict that if UK continues on same trajectory, then adds some not-particularly-good TTI in 8 weeks, deaths will peak at ~50 per day (unless I've misunderstood plot...)
    1. I tracked down the original source of the graph: https://mdedge.com/pediatrics/article/228682/coronavirus-updates/children-and-covid-19-new-cases-may-be-leveling… Coupled with the headline, this is atrocious data presentation from @AmerAcadPeds / @AAPNews. (Cases may be indeed leveling off somewhat, but this graph sure doesn't let you make that assessment.)
    2. To illustrate what is wrong with the original, @JasonSalemi overlays percentage change in cumulative plot with the cumulative plot itself, for Florida data on pediatric cases. The red line falls away precipitously even during the duration of peak growth in cumulative cases.
    3. Plotting the percentage increase in the cumulative total: for all but a few very specialized applications, this is a remarkably misleading form of data visualization because the ever-increasing denominator masks changes in the numerator.
    4. Not so much. Here's a regular cumulative plot of my time spent on zoom.
    5. Maybe it's coincidence, but it looks a lot the way that I've been able to bring my time spent on zoom calls under control over the course of the pandemic. Pretty impressive!
    6. Here's a remarkably misleading #dataviz via @BethPathak. Sure looks like the pediatric COVID situation is getting better, right?
    1. furthermore, there is no risk or probability linked to the 5 points on the Likert scale - any interpretation that assigns lower risk to lower numbers is pretty much as good as any other in terms of what I personally am entitled to think "won't happen to me" means... 5/5
    2. To be biased, judgments must be *at odds with some rational standard* but someone who is hyper-vigilantly not leaving the house, avoiding social interactions + washing their home delivered groceries can entirely realistically indicate they don't think they will be infected 4/5
    3. "The items were rated in a 5-point Likert scale (1=strongly disagree,5= strongly agree) and a mean score was computed with higher values indicating more optimism bias" 3/5
    4. Here the measure: "The participants completed three items to assess optimism bias: “I don’t think it’s going to happen to me”, “I don’t think it’s going to happen to my loved ones”, “I am afraid that someone I love may get infected” (reverse coded) => 2/4
    5. I cannot retweet this without noting that the measure of bias used is not a measure of "bias": 1/4
    1. For those who might think this issue isn't settled yet, the piece include below has further graphs indicating just how much "protecting the economy" is associated with "keeping the virus under control"
    1. I cannot attest to the accuracy of the underlying science/model but the idea of the tool is very cool and seems extremely useful!
    1. Here you go, debunking debunked - though I'm not wasting any more of my time on this twaddle!
    1. RT @KFF: .@DrewAltman discusses two fundamental policy decisions made by the Trump administration that set the U.S. on the controversial an…
    1. We can't predict the future. But we know only fools repeat errors of the past over & over without giving consideration to the best evidence available. The next 2 weeks will fill in a lot of details. Be patient. Be vigilant. Be careful. #MaskUp #SocialDistance #TestTestTest 8/end
    2. For those who think we are reopening too slow, look to Sweden: they are not relaxing measures until October. Their Universities remain in hybrid format. Or look to Israel, where they are on the cusp of a full lockdown after throwing caution to the wind. 7/8
    3. College testing skews data. In NYS, Tompkins county(Ithaca/Cornell) is responsible for 0.5% of the population but 4% of the total testing. 0.2% positive rate. OTOH - Oswego (SUNY) -0.5% of pop is responsible for 0.5% of total testing with 5% positive rate. 6/8
    4. We KNOW we have more college testing. We also know testing, overall, is trending lower = non-college-related testing is dropping sharply. Over the last 2 weeks many hot-spots have either already re-opened bars or have announced plans to do so. 5/8
    5. Case counts continue to drift lower. But we have recent experience where average age of cases drifted lower when bars reopened & severe cases showed up much later. Florida, for instance, hit a low for positive rate in late May - 2 weeks later cases started to take off. 4/8
    6. This is reducing the positive rate, giving us a less reliable indicator. We know from hospitalization and death data a lot about infections that happened in early to mid August, and that news is good. But what is happening NOW? I don't think we know enough. 3/8
    7. The national figures that we have come to rely on are more reliable but subject to MUCH more challenging interpretation than at any time. A LARGE percent (20% of Illinois in this article) of current tests are from colleges doing MASSIVE testing, in some cases. 2/8
    8. Thread/ Colleges, testing, & why we need caution. I am FAR LESS worried about the students. I worry about this - "One major risk is that infections could spread to at-risk faculty and staff and those in the surrounding community." 1/8
    9. Our daily update is published. States reported 763k tests, 37k cases, and 663 COVID-19 deaths. Important caveat: this update does not include Texas’s daily data, which is still not in today.
    1. Tried to reproduce your graph with Estonian data - curiously similar (despite of the >100x smaller scale on the X-axis)
    1. Some additional info: We're looking whether approach-training changes behaviour to real spiders (2 groups) and my supervisor says they will ask us to check if a change in behaviour is mediated by a change in approach-bias (on a variation of the training task).
    2. We're writing up a RR We have some analyses that are secondary to the RQs, but are likely to be conducted because reviewers will ask for them. We cannot afford enough participants for these to have sufficient power. Should we mention in Stage 1 anything about these analyses?
    3. Hi @TwitterSupport--I'm an ID epi that does a lot of science communication. Most of the other amazing scientists I do this with are verified. It would be great if the people I'm communicating with could know my account is really me & really trustworthy. Can I get some help?
    1. Testing is broken. So many stories in last 24 hrs of medical colleagues trying to get tests because kids have a fever or cough. They can’t get one. They can’t work until they get one. How long is this farce going to continue. This was entirely predictable.
    1. I’ve again updated my chart of COVID-19 hospital admissions to include the latest numbers. Admissions in England are rising. The 7-day moving average has increased by 70% since its low point on 26 August. This is the first sustained increase in admissions since March/April.
    1. Matt Hancock tells Radio 4 that on average people only have to travel ten miles to get a nasal swab test for Covid using his failing private system. Yet every GP surgery can do nasal swabs...but are not allowed to for Covid.
    1. A powerful image. These girls are sitting their university entrance exams in Afghanistan ...under a heat of 37C...on the ground...social distant...during a pandemic...because they know that only education will make them free
    1. 18. Disclosure: There are IHME doubters and disciples. I’m neither. I am a long-time ally of the GBD, which is vital work that's improved by transparency and honest criticism. In my view the IHME COVID-19 models don't yet meet that standard, but I'd like to see them to get there.
    2. 17. Bottom line: This model is getting lots of attention, from the public and (presumably) policy-makers. Trust in science and public health has been badly damaged during this pandemic. Transparency and good communication are essential to any hope for a science-driven response.
    3. 16. Provide more scenario analyses that allow some unpacking of the forecasts. What if the expected effects of seasonality are less pronounced than in the base model? How about running the best-fitting predictive model that excludes seasonal effects, for comparison?
    4. 15. Make it easier to find these methods. They should be one click away from the forecasts here: https://covid19.healthdata.org/united-states-of-america… Publish the forward projections of all predictors. The forecasted trends are driven by these. People need to see them.
    5. 14. So, hoping @IHME can confirm whether I've got all this right. I’ll leave it to others to weigh in on the strength of evidence for seasonal effects on SARS-CoV-2 transmission. Meantime, I have some suggestions for things that @IHME could do to elicit constructive input.
    6. 13. Which brings us to seasonality. Seasonality is captured using weekly, state-specific vital statistics data on pneumonia mortality from 2013 to 2019. So, as far as I can surmise, the estimated rise from 900 to 2900 daily deaths derives entirely from this seasonal effect.
    7. 12. It’s not behavior. The main projection expects distancing behavior to improve in response to a worsening epidemic. It’s the purple dotted line diving down in the figure, meaning reduced contacts. It’s not mask use, which is held constant ... …nor testing, which goes up.
    8. 11. So, this is the key: projections for the next four months depend on projections of the independent variables. In the main projection, daily deaths rise from <900 now to almost 2900 by December 1. What in the model drives things to get much worse?
    9. 10. Step 5: Make forward projections of the predictors, which leads to forward projections of beta. These are then plugged into the SEIR model to produce forecasts of all the other outcomes including cases and deaths.
    10. 9. Step 4: It’s all about that beta. Here’s the heart of the projection model: a linear regression of beta on a bunch of covariates, including several time varying ones: - social distancing mandates - changes in mobility - testing per capita - mask use - pneumonia seasonality
    11. 8. Step 3: More back-calculation! Here, a pretty standard deterministic SEIR model is fit to the estimated series on new infections. Key output in this step is estimated time series on beta, the transmission rate. [Aside: yep, this is a LOT of estimates on estimates ...]
    12. 7. Step 2: Back-calculate a time series on new infections from the smoothed death time series, based on … - assumed age pattern of mortality - assumed lag from infection to death - assumed age-specific infection fatality rates
    13. 6. Step 1: Estimate smoothed mortality time series using splines. First, cases and hospitalizations used as leading indicators to predict deaths. Then, a 2nd model synthesizes deaths from direct observation with the death series predicted from cases & hospitalizations.
    14. 5. The current approach is described as a hybrid ‘mortality spline + SEIR’ model. Let’s have a look at the components. Here’s the schematic from the July preprint. I believe it’s pretty self-explanatory. [Narrator: It isn’t.]
    15. 4. So, what’s in the model? There have been a few major renovations. The last one, I believe, was rolled out in July. All remnants of the much-maligned mortality CurveFit were excised in the July release. This CurveFit-ectomy was a welcome advance.
    16. 3. After some digging, I think the current IHME forecast model is described here: https://medrxiv.org/content/10.1101/2020.07.12.20151191v1… @IHME: if this is the current model, every estimates update and the FAQ need to point here, and not just to the March and April preprints.
    17. 2. First observation: the methods really need to be easier to find and vet. Optimally, publish all code. At least, have every update point clearly to the technical document with full model details. Right now, the FAQ and results updates point to old model versions.
    18. 1. Did the latest @IHME mortality forecasts making the media rounds - 410,000 deaths - seem high to you? Yeah, me too. I wanted to understand what’s driving projections of 220K more deaths by New Years. So, I tried to peek under the hood, as best I could. Buckle in. A thread.
    19. The majority (68%) in ICU care had one or more underlying condition considered as one of the risk groups, most prevalent being hypertension (37%), diabetes (25%), chronic pulmonary heart disease (24%), chronic respiratory disease (14%) and chronic cardiovascular disease (11%).
    20. Interesting numbers when I used the death certificate audit data from two Swedish regions where C-19 was established as a direct cause of death and excluded C-19 as a minor contributing factor only.
    1. A lot has gone wrong in the response to COVID. I would have done things differently in economic policy too. But overall the massive response from the Federal Reserve and even more importantly from Congress has been working and should be continued as long as needed.
    2. Action should be based on circumstances. The $600 a week boost to weekly unemployment checks may have made sense when the economy has shutdown but with an UR of 8.4% it should change. The President's $400 is reasonable--but the Senate needs to actually pass it for it to be real.
    3. Further action is needed. We're still in a bad recession. State/local job creation has been weak/negative in recent months & will get worse without aid. And households will run through their cushions soon, consumption growth will start to slow, and that will take a toll on jobs.
    4. The fiscal response has now ended. The "cliff" was at the end of July but given that most households saved a lot in the spring & had healthier balance sheets than pre-crisis they have some ability to smooth. So I would not expect bad macro impacts until Sep or Oct.
    5. The shock to the economy from COVID has in many ways been much larger than the shock that precipitated previous recessions. At the same time, the policy response has also been MUCH larger this time than previous times with a discretionary fiscal stimulus 5X the previous record.
    6. In the double dip recession in the early 1980s the unemployment rate was 8.5% or above for 24 straight months. In the financial crisis it was 34 straight months. This time it will have been 4 months (although it could tick up again).
    7. An unemployment rate of 8.4% is much lower than most anyone would have thought it a few months ago. It is still a bad recession but not a historically unprecedented event or one we need to go back to the Great Depression for comparison.
    1. Employment in industries where people can work from home is still down 3.9%, with little recovery. That's equal to peak-to-trough drop in jobs in those industries during Great Recession. Big red flag for longer-term recovery.
    1. Iowa COVID numbers are spiking. Jodi Ernst says she doesn't believe them, suggests doctors may be inflating figures to get higher reimbursement.
    1. 7 #COVID19 deaths on my reservation as of today. Nobody knows how many have the virus bc the numbers are different depending on who you ask. Tribal, IHS, state & federal data systems are spitting out different damn #s. This is ridiculous! Frustrated & angry af. We are dying!
    1. I apologise for any misunderstanding. What I meant was we need to get back to “work as usual” - face to face contact where it is safe to do so as well as online teaching. Buckingham and all universities have been working exceptionally hard all summer to prepare for the new term.
    2. I have always said I think a second wave quite likely. But I believe my own University in particular, but also the sector at large, is very well placed to manage if it occurs. We need to get back to work. The show goes on.
    1. It shouldn’t be hard to know who to trust on coronavirus information. Trust people who’ve been consistently right over the last nine months. Don’t trust people who’ve been repeatedly, unapologetically wrong. That’s it.
  3. Aug 2020
    1. The return to workplaces issue, just like the schools, is a perfect example of how we currently have to turn *everything* into a fight. When there are valid & powerful arguments on both ‘sides’, clarity & leadership are crucial. Sadly, our leaders prefer to foment the fighting.
    1. here's the quote from the piece itself: [An example of confirmation bias is...] "...our propensity to interpret declining infection rates as a confirmation that the lockdowns “worked”, when in fact this is a textbook example of the post hoc fallacy.".
    2. why? because it pretends that the temporal cue (falling numbers after lockdown) literally is *the only* evidence being used. Whereas, in actuality, a huge body of evidence *predicts* that lockdowns will bring down numbers and explains *why*
    3. this kind of piece behavioural scientists need to reject! A shallow understanding of the bias literature in an even shallower application to the pandemic- the idea that believing lockdowns brought down infection rates is an example of the "post hoc fallacy" is bizarre 1/3
    1. Of course, our plans may still go to hell in a hand basket. From the @CSBS_Illinois vantage point, behavioral and social science could be critical to the situation, but we may need to shift our priorities and consider other perspectives in order to wield more influence 23/
    2. Combining constant testing with comprehensive exposure notification, mask use, and curtailing large gatherings, there is hope that we can keep the outbreaks, which will happen, to a minimum (thanks South Korea and other countries for showing us how it is done, btw). 22/
    3. In the case of @Illinois_Alma, our ability and motivation to test twice a week is a great example of creating conditions where individual differences will hopefully not matter enough to close us down. 21/
    4. The efforts so far have focused on individuals changing their own behavior. While that is 1 solution, another “social science” solution is to create conditions where the individual differences don’t matter. This idea does not seem to be considered as much as one would hope. 20/
    5. Pay students not to party? Gift cards for wearing masks? TikTok threads supporting social distancing? They probably wouldn't hurt, but they are not magic bullets. 19/
    6. Punishment--threatening expulsion and the like--is simple behaviorism. Of course, it is only the stick part of behaviorism. While adding in some carrots would not hurt, I am hard pressed to identify any proactive incentive structure that would magically fix the situation. 18/
    7. The futility of a situationist position is no better demonstrated by the repeated cycle happening at universities where administrators wish their undergrad would simply behave differently and they don’t. I mean how much more of a strong situation do you need than a pandemic? 17/
    8. Situationism assumes that the overriding cause of human behavior is the incentives in any given situation. Change the situation, change the behavior. There is no need to consider prior standing on any attribute or population characteristics. 16/
    9. The idea that college students would miraculously and immediately overcome their well known propensities to be a bit more cavalier than their older patrons is the hallmark of situationism. 15/
    10. The two dominant approaches taken by universities to the need for self-control have been magical thinking or punishment. Interestingly, both magical thinking and punishment do reflect established social and behavioral science paradigms. 14/
    11. Relatedly, many folks, like @dynarski, are appropriately taking universities to task for expecting 19-year olds to get on the COVID-19 behavioral bandwagon and toe the self-control line. HT @sTeamTraen 13/
    12. So, until we start providing data that can help in applied settings, like the one we are facing, I think we should refrain from saying that we should be heard or have more influence. We need to have something to say of value first. 12/
    13. Knowing the compliance % would be rather useful to the modelers right now, but that is applied research. Having hard science envy, we over value “basic science” which is designed to be as useless as possible for applied issues like these. (maybe we can change that too) 11/
    14. Partially, it is because we don’t have much usable knowledge to provide. I know that undergrads are more impulsive and anxious than older populations, but I don’t know how that translates into something useful like the % of students will shirk our requests to comply. 10/
    15. The second reason for our lack of influence is well-worn preference for biological and technical answers to our pandemic problems--the old bias toward the hard sciences issue. While I could complain about this bias, I won’t. Why? 9/
    16. Over time, the student affairs people, given our myopic focus on our own research needs, appropriately began to rely on other social scientists, but not the ones employed as researchers at their own university (maybe we can do things differently going forward...) 8/
    17. The only reasons for us to work with student affairs in the past was to plead with them to selfishly use their data or gain access to the students so we could collect our own data. 7/
    18. First, front line social and behavioral science researchers have seldom helped the units that were charged with the day-to-day activities of undergraduates, like the Vice Chancellor for Student Affairs office. 6/
    19. While I feel like my university has afforded the @CSBS_Illinois the chance to inform the process, for which I am grateful, for the most part, our efforts have not been systematically incorporated into the pandemic planning. I see at least two reasons for that, IMHO. 5/
    20. We posted information on how social and behavioral science knowledge could help. https://csbs.research.illinois.edu/news-events/social-and-behavioral-science-and-covid-19/… Conducted workshops: https://csbs.research.illinois.edu/understandingcontemporarychallenges/… Ran studies: https://csbs.research.illinois.edu/news-events/blog/… and wrote blogs: https://csbs.research.illinois.edu/2020/08/16/what-we-know-about-college-students-to-help-manage-covid-19/… 4/
    21. At the @CSBS_Illinois, we started curating social and behavioral science insights from day 1 in an attempt to provide information to our community. 3/
    22. Being at a University (@Illinois_Alma) that has, to my knowledge, created the most informed and comprehensive system to open safely, and being the director of our social and behavioral science unit (@CSBS_Illinois) I have some thoughts I’d like to share. 2/
    23. Thread alert: There is a lot of back and forth about the use and abuse of behavioral and social science knowledge to help universities and colleges open up to in-person education, or not, in the time of COVID-19. 1/
    1. LIVE now: Independent SAGE's weekly briefing. Please join us for latest analysis & questions from the press & public. All welcome!