COVID-19 projections for hospital resource usage
April 2, 2020 6:26 PM   Subscribe

COVID-19 projections for the US, state by state The Institute for Health Metrics and Evaluation from the University of Washington has offered this tool to help hospitals predict what resource usage will be like over the next few weeks. Numbers are updated daily.
posted by Quietgal (49 comments total) 24 users marked this as a favorite
Go KY.

Go Banana.
posted by Young Kullervo at 6:32 PM on April 2 [3 favorites]

I live in Washington and everyone keeps telling me that the good-looking numbers here is because "we're doing stuff right" but we're still not testing boatloads of fucking people who probably have it and I am very, very skeptical that things are getting that much better that much quickly for our state, no matter how much I hope.

You still can't get tested unless you're in a high risk group, symptoms be damned. Truly hard for me to believe its been enough to flatten the curve as much as UW is predicting its been flattened.
posted by deadaluspark at 6:40 PM on April 2 [11 favorites]

The IHME projections were discussed in another thread.

The TL;DR seems to be that these projections make pretty optimistic assumptions.
posted by aneel at 6:42 PM on April 2 [9 favorites]

There's also Covid Act Now which has a different set of predictions. I am not certain how many podcasts endorse their statistical model.
posted by RobotVoodooPower at 6:42 PM on April 2 [1 favorite]

Important caveat: This is strictly a curve-fitting exercise. It doesn't incorporate any actual mechanistic epidemiological model:

"What the method is really doing is using the second derivative of the time trend on the log scale to estimate where we are on the curve. Once that second derivative goes negative, so the exponential growth is slowing, the model takes this as evidence that the rate of growth on the log scale will rapidly continue to go toward zero and then go negative." -Andrew Gelman (March 29th, 2020)

This means that the uncertainty reflected here literally can't imagine the possibility of a second wave (if, for example, precautions are relaxed prematurely), which means the long-term downturn error bars are way, way too overconfident. So this is a useful tool for the time being to make approximate projections, but the worst case scenarios are much, much worse than this model implies.
posted by belarius at 6:43 PM on April 2 [35 favorites]

belarius – apologies. But can you dummy-down the Andrew Gelman quote for us, and what you mean as a "curve-fitting exercise" and "mechanistic epidemiological model"? I like to think I'm a smart guy, but I think compared to your understanding of the science, I'm probably not. And I'm a little concerned as I had been suggesting this site to others. Also, do you have some predictive site/reference you would recommend something as superior?
posted by WCityMike at 6:59 PM on April 2

For a smart critique of this model, I recommend the Twitter feed of Carl Bergstrom, particularly this thread.

An important point he makes is that the range of values shown in the model are not a best case scenario vs. a worst case scenario. Instead, it shows what the range of possible situations could be even under a best case scenario. In this case, a best case scenario is that every state quickly implements very tight restrictions on movement (if they haven't already) and keeps them in place through the end of the pandemic, that people follow them, that health systems won't get overwhelmed and supplies run out, and that things will progress like they did in Wuhan according to the Chinese government's official statistics (which may be markedly underestimating things).
posted by Tsuga at 6:59 PM on April 2 [8 favorites]

A curve-fitting model is much simpler (but not necessarily worse because of that). It takes the raw data (in this case, the number of deaths per day), and finds a simple equation that estimates how that changed over time, without caring about why things happen that way. A mechanistic model looks under the hood, trying to understand how different factors affect the data, and makes predictions from there, but is freaking hard to do.
posted by Tsuga at 7:08 PM on April 2 [4 favorites]

belarius – apologies. But can you dummy-down the Andrew Gelman quote for us

The incredibly simple explanation using what I learned in AP Calc 20 years ago: If you throw a baseball, at lets say a 20 degree angle, it goes up for a bit, levels off, and then it starts to fall back to earth. Using the type of math you would learn in the beginning Calculus, you can actually map out the whole curve pretty quickly since the acceleration up will go to zero at the top of the curve and then fall. We can do this because (especially in the "assume a vacuum" universe math homework happens in) the rules that govern the way up also govern the way down.

This kind of thinking breaks down in real life because all kinds of crazy shit can happen: Trump could force everyone to go back to work at gunpoint, running out of ICU beds could cause people to die of Covid-19 that would have lived, a meteors could hit every major hospital in California, etc. In my baseball example, it would be like if the ball started falling back to earth and then just shot straight up into space.
posted by sideshow at 7:10 PM on April 2 [7 favorites]

I’ve been refreshing this yesterday and today. For my home state it has oscillated between 14 and 22 days from now as the peak death rate. I feel like little changes in the current data are having a huge impact on its projections.

It also estimates that we will have enough hospital beds here and honestly I feel like that’s a dangerous message to put out there without some big flashing disclaimers.
posted by beandip at 7:24 PM on April 2 [3 favorites]

Ugh. Moratorium, please, on all unqualified curve fitters. Engineer's disease in a new form.

I'm glad Americans are learning math, but
posted by eustatic at 7:30 PM on April 2 [38 favorites]

The IHME site's prediction for Iowa on Monday was 138 deaths; on Wednesday it was 1367. Today it's 1488. Covid Act Now's social-distancing-only scenario went from 46,000 to 10,000 in about five days. So what I've learned from watching the models is that somewhere between 138 and 46,000 of Iowa's 3 million people will die. Which I feel like I could have told you before I saw either prediction, with no expertise or mathematics involved at all.

I understand that it's a difficult problem, with way less information available than one would want, and conditions changing more or less continuously, and that people want some kind of prediction that would help them make decisions about what to do. But -- perhaps those conditions are exactly the sorts of conditions in which you should keep your wildly-fluctuating model to yourself, and not stick it up on the internet for anyone and everyone to look at?
posted by Spathe Cadet at 7:31 PM on April 2 [14 favorites]

There's also Covid Act Now which has a different set of predictions. I am not certain how many podcasts endorse their statistical model

I haven't gotten a chance to look at this model in depth, but it looks they attempt to model the effect of different levels of intervention. Looking at their predictions for several states, it looks like New York is the only state that they project to peak before the end of summer given current intervention. The rest of the states look like they will just see a rebound of cases once social distancing restrictions are relaxed.

I mention this because during the 1918 influenza pandemic because Philadelphia famously had an extremely lethal fall wave when the city neglected to impose social distancing measures before it was too late. Cities such as St. Louis and San Francisco escaped the worst of the fall wave because they imposed more stringent social distancing measures, but then they just got hit by a bad second wave in the spring.

That pattern has been weighing on me since I read about it. I'm honestly not sure what the endgame here is. And yeah, those projections are a lot scarier than the IHME model.
posted by eagles123 at 7:33 PM on April 2 [6 favorites]

In my baseball example, it would be like if the ball started falling back to earth and then just shot straight up into space.
I think it's more like if the baseball started falling back, and then halfway down someone hit it up again with a bat. The model assumed no other force would affect the ball besides gravity.
posted by floomp at 7:39 PM on April 2 [7 favorites]

I don’t put a lot of faith in these models because unless your state is really small, like DC small it depends on WHERE the cases are to see if the system overloads. You city could be overloaded, while the city over is fine. These just aren’t fine grained enough.
posted by jmauro at 7:42 PM on April 2 [9 favorites]

There are states with lousy healthcare systems, very late shutdown orders, and probably rapidly escalating numbers of infected people. When things get bad in those places do you think people are going to stay in those states? Many people won't have a job any longer to tie them to Florida, Mississippi, Louisiana, or whatever ACA rejecting state they're currently living in. No, they'll run to California or New York or some other state with a semi functioning, that is after the surges pass, health care system. That migration will start a second wave. I don't see how any model could account for that type of internal movement.
posted by rdr at 7:45 PM on April 2 [12 favorites]

Ditto for someone hearing Hospital A is full and, once symptoms appear/escalate, simply driving an hour or two (or three or four) away to get to a hospital that doesn't have the reputation of being full/swamped. So, yea, combined with what rdr is saying, I don't see anywhere getting a pass on this.
posted by RolandOfEld at 7:51 PM on April 2 [6 favorites]

So, in a nutshell, April is going to be a really rough month for everyone.
posted by LilithSilver at 7:55 PM on April 2 [4 favorites]

Ugh. Moratorium, please, on all unqualified curve fitters. Engineer's disease in a new form.

I believe these are professional public health curve fitters, but some criticism from other public health people does apply.
posted by atoxyl at 8:06 PM on April 2 [4 favorites]

Yeah, I just finished a class on healthcare informatics at none other than the University of Washington.

Don’t get me wrong, I think there is potentially a lot of power in using Big Data to drive predictive analytics in healthcare. BUT, these models are created by IT geeks with zero direct healthcare experience, and the validity of these models depends entirely on the quality of the data that goes into them. Only front line health care workers seem to understand exactly how flawed the data inputs are. The reliability of data harvesting in the very complicated real world of health care delivery is a major problem yet to be overcome before this approach can really be useful. And most of us front line workers who are collecting this data have much more immediate competing priorities than making sure our coding is 100% complete and accurate.
posted by Slarty Bartfast at 9:35 PM on April 2 [23 favorites]

This thread is reminding me that my governor (Kim Reynolds) got her BLS five months before she was elevated to the office of governor. I doubt her studies gave her a foundation to understand these numbers or where they come from.
posted by Big Al 8000 at 11:02 PM on April 2

The Hammer and the Dance (via Medium) is an informative analysis of possible trends.
posted by fairmettle at 12:07 AM on April 3 [5 favorites]

Even assuming that the curve-fitting stuff is 100% accurate (ha) this doesn't really say anything about where those hospital beds are located. E.g. maybe Missouri won't have a shortage but St. Louis will, or maybe some rural-ish area with a nursing home will.
posted by Foosnark at 4:49 AM on April 3 [2 favorites]

The WaPo this morning has a long and very well-written piece about the various models [including a few paragraphs on the distinction between the epidemiological SIR models and the IHME curve-fitting], and how those models are being used to set policy:

Experts and Trump’s advisers doubt White House’s 240,000 coronavirus deaths estimate

posted by Westringia F. at 6:21 AM on April 3 [1 favorite]

California, one of the largest economies in the world, is just over the peak. Hunkered down in my home office, I laugh at COVID-19. Laugh, I say, HAHA! (I need to wash my hands now.)
posted by SPrintF at 6:42 AM on April 3 [3 favorites]

I’m super worried about reports I’ve heard that say the test only has maybe 70% sensitivity. What the heck! How can a test be approved for public use with sensitivity that low? Am I missing a trick here? Am I naive?

Anyway, I have to imagine that misfeature is affecting our models, too.
posted by eirias at 6:53 AM on April 3 [1 favorite]

SPrintF, that's not at all what I'm seeing in the IHME model, which projects the peak for CA to occur on 26 Apr. Is there indication from any other model that CA is "just over the peak"?
posted by Westringia F. at 7:05 AM on April 3

Uh, yeah, California is not just over the peak. That is not how this works. That's incredibly dangerous thinking right now if it encourages people to stop sheltering.
posted by OnTheLastCastle at 7:06 AM on April 3 [4 favorites]

I was reliably informed that a crisis of this magnitude would turn us into a society of scavengers, and we would all be issued with studded leather and dune-buggies. Instead everyone seems to have become statisticians - my twitter feed is constantly filled with graphs and charts.

My armchair sociologist theory is that all the people that were posting climate change graphs 2 months ago pivoted very easily to epidemiology - it's all just numbers and trends. The problem is that climate is well studied and has lots of easily obtainable measurements (relatively speaking). This disease is new and understudied - we simply have little idea of how it is affecting the population. There are just too many variables, especially when estimating how many people are already infected and whether immunity is conferred.

These graphs are pretty but useless, the equivalent of graphing the number of points your team scores in the first weeks of the season then drawing out a trend line with error bars for the entire year.
posted by AndrewStephens at 7:07 AM on April 3 [7 favorites]

Something to keep in mind about the confidence bands around the IHME curves: Any curve that fits within those shaded regions also falls within their 95% confidence estimate, not only those that peak at the same time as the main curve. (You can see this pretty clearly if you compare the curve of the lower bound of the band to that of the main curve; other curves are also possible.)

That is, there is not only uncertainty in the height of the IHME curve, there is also uncertainty in the timing of the IHME curve.

And that's leaving aside all the other [major, IMO*] issues with the IHME projections.

* I do mathematical & statistical modeling of living systems, but not these living systems.
posted by Westringia F. at 7:16 AM on April 3 [6 favorites]

That's California dreaming, I'm afraid.
posted by Too-Ticky at 7:24 AM on April 3 [2 favorites]

I just spent 11 weeks learning to do statistical analysis from decision trees, hypothesis testing and regressions and you can make data say pretty much whatever you want if you futz with it. The actual techniques aren't often that hard, but the experience seems to be in using the principle of parsimony (mmm) to remove the chaff and get to an actual model for something. Which is one sentence that makes it sound like it's not insanely complicated.

But it is.

Too bad I guess I'm done with grad school for now.
posted by OnTheLastCastle at 7:44 AM on April 3 [2 favorites]

Any chart has to allow for so much complex localized information in a patchwork system like the US or even Canada, where resources are unequally distributed across different governments and large regions, that it hard to believe anything now we see is capable of a good prediction of much.

Any model for British Columbia, for example, that doesn't model what would happen if the virus gets into remote first nation communities and the dense populations of the homeless in many towns and cities, is going to be effectively useless. Those communities have huge complicating factors from lack of clean water, health care access, housing,.etc. that an outbreak among them would demand far more of a central system than an outbreak in a place with a very different set up. A lot of charts also don't model for endemic health issues in certain groups and locations, where people will also need a lot more resources thrown at them to survive.

(Bitter note: BC is basically going to get a lot of people killed in poorer communities with its current approach. I am currently having to scour shops for certain supplies for a place I used to volunteer out, because they are not getting what they need in a very wealthy city.)
posted by lesbiassparrow at 8:32 AM on April 3 [6 favorites]

Yeah, consider "daily" death totals just as an example. America has 4 time zones. It's a new day when it's still 8pm in the Pacific time zone. If someone was compiling but missing part of it on either end or it was being reported to them off... well, you see how even something "simple" like a daily total can get muddled.
posted by OnTheLastCastle at 8:44 AM on April 3

For the *maybe* simpler explanation check out: Simulating an epidemic by 3Blue1Brown. Or The Coronavirus Curve - Numberphile. These are simplistic models based on SIR. The Susceptible, the Infected, and the Recovered (or Dead) and a few numers... How many S people does and I person meet and infect? How long does it take for an Infected to become Recovered.

You can draw pretty curves with these equations. What I gather this sit is doing is going backwards and taking the Number of New Cases, the Number Currently Infected, the rough Recovery Time... And then saying that if everything was perfect, this is how it would go. The 3Blue1Brown video goes into a bunch of limiting travel, increasing social distance, going to a common place (store/church), etc.

It's just the beginning of actually trying to model the actual situation. It's more like "if everybody stayed home and didn't leave for a month, it would be over. Everybody Infected would have Recovered (or Died) and there would be almost Zero transmission". Sadly it's the real world,not the spherical cow in a vacuum world.
posted by zengargoyle at 10:16 AM on April 3 [6 favorites]

I am wondering how these models deal with sampling bias? That is, what is the true number of cases given limited testing, changes in testing protocols, testing as a function of perceived/apparent risk etc...
posted by piyushnz at 11:37 AM on April 3 [1 favorite]

I don't watch the US govt conferences but I wouldn't be surprised if they're choosing specific numbers from specific models for political benefit:

Pres: 2.5 million people were going to die, I reduced it to only 250K, reelect me

And the population will pat themselves on the back for doing such a great job, and none of the fundamental problems in our health system will be fixed. I'm already seeing this in my own very-liberal state. No doubt some communities will declare the pandemic is over and drop all restrictions, setting them up for a second wave of deaths, like in the 1918 pandemic.
posted by meowzilla at 12:28 PM on April 3

To be clear, I live in California, and I see these headlines both highly rated at the same time:

- California dropping the curve better because it acted so early
- California is the second-worst state for testing by capita

So just like any other emerging situation, the data is incredibly fuzzy and people are choosing what they want to see.
posted by meowzilla at 12:32 PM on April 3 [2 favorites]

Both of those things can be true.

No one is doing sufficient testing to track the outbreak in terms of cases. There is not enough materials in production to do it. Though I hope that no one is underreporting COVID deaths. Which means you can get a pretty good idea of the effectiveness of responses, if a little more delayed.

The last two weeks of data for America as a whole in terms of deaths looks, there is no nice way to put it, really bad. Like death total* doubling every 3 or 4 days or less really bad.

For reference, the initial doubling time used in the model in the Imperial College paper that pretty much said "everyone has to self isolate or the healthcare system will get overwhelmed" was 5 days. So it is reasonable to expect things will be worse than those already dire predictions.

I've only been keeping track of Washington, but here it's been more like doubling every 8 days, so measures are clearly having some effect here, and California's still got fewer deaths than we do. I hope California isn't catching up and is in the same boat. So we're clearly doing better, but they still don't have enough tests to give them out for milder symptoms.

Of course the hope of flattening the curve is for the growth rate to stop being exponential sooner rather than later. But I wouldn't unclench until there's a really solid deviation from exponential growth on the death count. Maybe then someone can get a good guess of when and how high we'll hit a peak.

For now, it's not actually possible for the health care system to over prepare and everyone else needs to just stay TF at home (NSFW: profanity) for a good long while.

*Note: for an exponential curve, the value and the rate of change double at the same rate, and the death count is going to be less noisy than the deaths per day.


It feels so... messed up to take this bloodlessly about people dying, but I'm scared y'all, and I don't know how to phrase it otherwise without it going straight in the rant thread. Just... please, please, please encourage everyone you know that this is serious and they shouldn't cheat on self-isolation, no "just one visitor, it will be okay." Because then that visitor might think it's okay to have one visitor and so on and soon you have a network of transmission again. Stay safe out there as best you can. For those who are high risk even if you aren't.
posted by Zalzidrax at 3:32 PM on April 3 [5 favorites]

I wish the serological testing could come online soon. I had a cold that hit me like a ton of bricks last month and I would really like to know if it was COVID-19 and if I'm immune. If so I can get out and start helping instead of hunkering down trying to flatten the curve.
posted by Your Childhood Pet Rock at 4:40 PM on April 3 [1 favorite]

Serological testing is key definitely key. We can know which healthcare workers are immune. We can know who isn't an asymptomatic carrier and who is immune, which will provide flexibility with reopening the economy. It might expand the pool of plasma donors if that is needed and proves medically efficacious. And, it will help us understand how many people become infected and develop antibodies to the virus without developing symptoms, which help with modeling.

As for current data, it's impossible to know the true case rate right now. My brother and his wife live in Brooklyn. They both developed a fever that lasted about a week or so. They got through it fine, thank god, but they'll never show up on any official case count. Did they have COVID? It's impossible to to know right now.
posted by eagles123 at 5:14 PM on April 3 [1 favorite]

My wife put up a post Stay In Your Bubble about the need to remain isolated and stay at home. The post has an excellent gif illustrating the consequences if lockdown directives are not adhered to. I hope this okay to post as my immediate relative.

Isolation is a directive here, not advice and the army are on standby to back up the police where anyone refuses to comply.

Graphic is from Siouxsie Wiles & Toby Morris who are cartooning about Covid here.
posted by unearthed at 5:38 PM on April 3

I wish the serological testing could come online soon. I had a cold that hit me like a ton of bricks last month and I would really like to know if it was COVID-19 and if I'm immune. If so I can get out and start helping instead of hunkering down trying to flatten the curve.
posted by Your Childhood Pet Rock at 4:40 PM on April 3

I'm not sure I really understand how immunity works for this disease.

First question: is there any solid evidence yet (or is it far too soon to tell) that a person who has recovered from COVID-19 has any immunity from reinfection? I'm not sure where medical researchers are with that question.

Second question: can a person who is immune still act a disease vector if precautions are not taken? For instance, if you shake the hand of an infected person and then shake the hand of someone else without having washed your hands in between. That would still be an issue, correct? Also, can an immune person still carry the virus itself (say, in their lungs) even if they are otherwise unaffected, like Typhoid Mary? Or is that more or less the same thing as being asymptomatic?
posted by acidnova at 10:33 PM on April 3 [1 favorite]

First question: is there any solid evidence yet (or is it far too soon to tell) that a person who has recovered from COVID-19 has any immunity from reinfection? I'm not sure where medical researchers are with that question.

With a normal immune system, yes with a BUT. Any infection with a functioning immune system will confer at the minimum a short term immunity to COVID-19.

Here's the "BUT". It's all based on probabilities. The key thing to confer immunity is whether the antibodies on your memory cells match the antigen presented by the pathogen. The reason it confers immunity is because your body is constantly cranking out B cells with essentially random antibodies on them, looking for antigens. You basically have an immune systems cranking out keys looking for a lock. This means it involves random chance and a random time for your body to find the right key for the SARS-CoV-2 lock. This is why some people make it through with a minimum of symptoms. If your body is lucky enough to find a good enough key quickly you can start flooding the body with antibodies which stop the virus before it has the chance to gain a beachhead.

When you have recently had the disease, there are memory B cells. They hang out mostly in your spleen but some of these float around in your blood and what happens is that if the virus comes in again, your immune system is primed. You have a key ready for that lock in your blood and can immediately start producing antibodies before the virus has even the slightest chance to replicate.

The reason we can still get things like colds or flus is because they change their antigen. They accumulate mutations that change the proteins they use for entering cells enough so that the prevalent antibodies don't match the antigen. This, again, is an entirely random process based on how quickly mutations accrue and how good the person's antibodies are at conforming to new mutations. Since the antibodies are random, each person has a slightly different antibody for an antigen. This is why people like AbCellera are conducting assays of antibodies to try and find the best one to turn into a monoclonal antibody for mass production.

This is also why they're currently calling for blood donations from COVID-19 survivors. The antibodies in the blood can be turned into a convalescent serum. You're using antibodies from someone else to bind to the virus which can give the body a fighting chance while it ramps up its own B cell and antibody production to keep the virus at bay. Since there's a limited number of antibodies it can only work for so long but it'll give a house edge needed to make an infection less severe. The severity of an infection is all a product of how many times the virus can replicate and how many cells and organs it can infect. Lowering replication numbers at the start can be insanely helpful for keeping the infection to a minimum.

In the case of SARS-CoV-2 the antigen drift has been negligible to the point where some SARS-CoV antibodies (circa 2003) have been effective against it, which probably means one infection is most likely going to confer immunity. If there's a new strain coming through that presents a slightly different spike protein that the antibody doesn't fit, at that point your memory cells will be useless and your body has to go through the whole process of finding an antibody, ramping up B cell production, and then fighting it again while the virus has had a chance to gain a beachhead. This is why some people can get reinfected by chance even with a functioning immune system.

COVID-19 is a very large virus which lowers the chance of antigen shift, as well as having a 3’-5’ exoribonuclease which proofreads the RNA that SARS-CoV-2 uses, also minimizing the number of mutations. The lower the number of mutations, the less chance the spike protein will appreciably change. Again, all based on probabilities.

Second question: can a person who is immune still act a disease vector if precautions are not taken? For instance, if you shake the hand of an infected person and then shake the hand of someone else without having washed your hands in between. That would still be an issue, correct?

Sort of? A person who's immune isn't going to be spreading droplets like an infected person and the droplets are what causes most of the infection because they have a large amount of viral copies in them. The key thing is whether a virion is able to make it from a surface to an orifice to something it can infect. Again, this is probabilities. You have mucus linings which trap viruses and other defenses that are physical which reduce but don't eliminate the chance of infection. In the case of an immune person, there's going to be massive attenuation of probabilities between an infected person, that immune person, and another person who is not immune.

So while it's not zero, it's low enough that we can restart the economy with proper serological surveillance and isolation.
posted by Your Childhood Pet Rock at 5:17 AM on April 4 [18 favorites]

Ditto for someone hearing Hospital A is full and, once symptoms appear/escalate, simply driving an hour or two (or three or four) away to get to a hospital that doesn't have the reputation of being full/swamped. So, yea, combined with what rdr is saying, I don't see anywhere getting a pass on this.

So it begins.
posted by RolandOfEld at 9:16 PM on April 4 [1 favorite]

J. B. Pritzker, the Illinois governor, is challenging Mike Pence’s calculations that his state needs only 1,400 ventilators.

“I pray that the vice-president was right, but let me tell you where I think he got his numbers. He looked at a University of Washington model1 that’s out there that people are accessing to look at every state,” Pritzker said on CNN’s State of the Union.

“The problem is they didn’t put that model out there for resource allocation reasons. If you look at the model there’s a central point that shows Illinois needs only 1,400 or fewer… but there’s also a worst-case scenario that shows we would need five, or six, or seven thousand more.

“We’re looking at all of the numbers, and everybody is taking an educated guess because we really don’t know.”

He added: “This virus is unpredictable. Our guess is we need a few thousand more than we have now, up to 4,000. That’s what we’ve asked the federal government for, and over time they’ve given us 450 total.

“We’re looking everywhere and anywhere across the world to get ventilators. Here’s the problem, the president didn’t enact or use the defense production act until just recently.”

Pritzker said he was worried it would lead to a shortage of ventilators nationwide, because he thought New York would still be at its peak at a time the need became greatest elsewhere and unable to share its supply.
1 The IHSE is at the University of Washington.
posted by WCityMike at 1:21 PM on April 5 [3 favorites]

On the latest update from my hospital I see that they are using this prediction to help determine when the peak will hit here in GA. I generally feel our hospital is preparing well for this, and given that we are a academic medical center affiliated with a university there are plenty of experts available to critique and choose which information to use; certainly people far more knowledgeable than I. That makes me think this model is pretty good. Of course, as has been pointed out, there is a lot of uncertainty in any prediction given that we are not testing nearly as much as we should and we really don't know how well people will follow the advice to shelter in place, wear masks, and so on over the next few weeks. So it looks like the next two weeks will be crucial, at least here.
posted by TedW at 10:09 AM on April 6

They review the accuracy of the model so far in their updates. It doesn't matter that much how accurate these models actually are. These types of models have already gotten Boris Johnson and Donald Trump to take some action rather than being entirely passive.
posted by rdr at 2:29 PM on April 6

Not to be overly glib, but I don’t think Boris Johnson is acting on much of anything right now. That guy is in rough shape.
posted by Big Al 8000 at 10:08 PM on April 6

« Older This post has no dreams.   |   Which character are you? Newer »

This thread has been archived and is closed to new comments