Home > Uncategorized > A better proxy than confirmed cases for the US is hospital beds

A better proxy than confirmed cases for the US is hospital beds

March 14, 2020

If you’re anything like me, you’re driving yourself a bit nuts looking for information on the COVID-19 situation in the U.S. as well as internationally.

And, if you’re like me, you will have been tearing your hair out frustrated by the bad quality of data. For terrible political and cultural and of course capitalistic reasons, we are not getting tested at reasonable rates as charts like this demonstrate:


That means, as cool as websites like this or this are for up-to-date data on country and state level confirmed positive test results and deaths, we are not actually seeing enough to know how bad things really are. Note to the people who run those sites: please add columns for “number of tests administered” so we have some idea of how incomplete the data really is.

All of the above is nothing new, I’m sure you’ve already complained at length to your friends about this very topic.

The only things I really want to say is that, considering that we don’t know who is infected, we should use as a proxy of our problem not the “confirmed case” count but rather the “hospital beds used above normal” count.

It’s not perfect metric either, primarily because the virus takes days to spread and days to get people quite sick, and indeed once the hospital beds really start filling up, it’s already a much bigger problem. But I’d still argue that we should try to use this proxy, for the following reasons:

  1. Hospitals definitely keep track of their bed counts, you can be sure of that. So the data is available.
  2. It doesn’t depend on the availability of testing, which as we’ve discussed above, is very problematic. Our death count is most assuredly way too low because there have likely been plenty of people dying of coronavirus who just never got tested.
  3. In other words, counting beds is relatively free of political manipulation.

There are also problems with this proxy, both of which make it underestimate the problem. First, because people who have elective surgery are rescheduling for a later date. Second, because our social distancing has probably made other illnesses less common, leading to less hospital stays for other reasons.

Even so, I’d love an intrepid journalist to try to collect this statistic from many national hospital chains, and compare it to last year’s bed count, as well as yesterday’s and last week’s, to see how things compare.

Categories: Uncategorized
  1. sarah_cook@tds.net
    March 14, 2020 at 5:42 pm

    Interesting idea. My husband just had a scheduled doctor appointment cancelled because they’re on such overload. As you say, elective surgery won’t happen.

    Even if we had ample testing instead of the current fiasco, there would still be a lot we can’t know yet. Specifically, what’s the rate of asymptomatic carriers, which is really important for preventive community measures. Given that some countries are doing a quite good job with testing (though still presumably just people with symptoms or at special risk), this kind of info about the epidemiology of the disease would exist if there were enough data anywhere – even in Italy or China – to know it yet.


  2. Fiana
    March 14, 2020 at 5:54 pm

    OECD has the listing of the number of total hospital beds per 1000 people until 2017:

    Sadly US only has 2.77 per 1000 people, which is behind Italy. Grim picture ahead.


  3. March 14, 2020 at 5:59 pm

    First, thanks for the useful links.
    Personally I assume all of us who live normal lives in any metropolitan area have or will be exposed to this virus, and that perhaps a little more emphasis ought be applied to maintaining a strong immune system, for those who can (and aren’t already compromised)… i.e., good diet, exercise, sunshine, fresh outdoor air, adequate sleep, perhaps various vitamins/supplements, etc. and rapid attention to any coughs, chills, fever, aches, etc. that do arise — I’m a bit bemused at the number of grocery carts I see piling up with chips, ice cream, candy, soda, and other unhealthy comfort food (although if you have a large family and intend using it up slowly over a long period it’s okay). And outdoor surfaces can actually be a lot more ‘sterile’ of harmful agents than indoor surfaces — the sun being a great sterilizer — while indoor air can be truly dreadful. In short, I know I experience better health (particularly in winter months) when I spend lots of time outdoors and active, than when isolated and sedentary indoors. Just my two cents.


  4. Hal Davis
    March 14, 2020 at 6:00 pm

    There are some comparative numbers here, but you have to scroll for a bit: https://delano.lu/d/detail/news/countries-most-critical-care-beds-capita/209781 Luxembourg in English

    The countries with the most critical care beds per capita


  5. March 14, 2020 at 6:04 pm

    The numbers seem to be speculative at this point.


    Washington Post: This is the coronavirus math that has experts so worried: Running out of ventilators, hospital beds

    The United States has roughly 2.8 hospital beds per 1,000 people.

    The United States has an estimated 924,100 hospital beds, according to a 2018 American Hospital Association survey, but many are already occupied by patients at any one time. And the United States has 46,800 to 64,000 medical intensive-care unit (ICU) beds, according to the AHA. (There are an additional 51,000 ICU beds specialized for cardiology, pediatrics, neonatal, burn patients and others.)

    A moderate pandemic would mean 1 million people needing hospitalization and 200,000 needing intensive care, according to a Johns Hopkins Center for Health Security report last month. A severe pandemic would mean 9.6 million hospitalizations and 2.9 million people needing intensive care.

    Now, factor in how stretched-thin U.S. hospitals already are during a normal, coronavirus-free week handling usual illnesses: patients with cancer and chronic diseases, those walking in with blunt-force trauma, suicide attempts and assaults. It’s easy to see why experts are warning that if the pandemic spreads too widely, clinicians could be forced to ration care and choose which patients to save.

    Brussels Times: ‘Exponential’ coronavirus cases put hospital bed count under stress


    “The number of [coronavirus] infections is increasing exponentially,” Geert Meyfroidt, the head of Belgium’s Association for Intensive Care Medicine, told De Morgen.

    Belgian hospitals currently have 1,900 intensive care unit beds, putting the country among the EU’s top five in terms of equipment availability, Meyfroidt said.

    “We know that 20% of people with symptoms end up in the hospital, and 5% end up in intensive care,” Meyfroidt said. “That would mean that we would need 50,000 beds on intensive care — we could never handle that.”
    There are some comparative numbers here, but you have to scroll for a bit:


    Delano: Luxembourg in English

    The countries with the most critical care beds per capita

    Hal Davis


  6. Christina Sormani
    March 14, 2020 at 6:27 pm

    How about a count of patients arriving at hospitals with difficulty breathing and requiring oxygen rather than counting beds?

    It seems all beds are used at all times in NYC hospitals with many people sitting while receiving IVs or oxygen even before COVID-19. This is just based on visiting NYC hospitals frequently with my father in law who had COPD over the past three years and waiting 6-8 hours for them to find a bed for him. Not only will there be people with elective surgery told not to come in in the first place, but more people will be sent home with oxygen tanks than are already sent home with such tanks.


  7. March 14, 2020 at 6:52 pm

    Kathy O’Neill is a brilliant former finance quant and present occupier.

    I have had this same thought about why we are using as our primary metric a data set that is NOT AVAILABLE. Until we have something better, we might be better off relying on numbers of reports to MDs; etc of respiratory distress and fevers, over and above the usual range of the past 50 years, and factoring out increased reporting, due to publicity. The tests are mostly for understanding epidemiology. They don’t determine treatment. We’re about past the point of knowing who can be presumed to be unexposed. Assume everyone is, and should stay away from the vulnerable.

    Demand for beds is another great, though also rough, measure.

    Vince Gay



  8. JV
    March 14, 2020 at 7:48 pm

    If I were to be tested, I would also want to know the actual results of the “tests”.
    I believe PCR is the only valid test for viral varmints. We need actual data quantified…how many millions of virus particles are viewed in the sample. I believe it has to be many millions to be viable.
    So, what KIND of test is being used and…
    What are the quantified results of the test and…
    How about a viewable graphic/photo of the COVID-19


  9. jacksartspace
    March 14, 2020 at 8:00 pm

    Agree with you, Cathy about the data we are seeing. I’ve been writing quite a bit on my blog about coronavirus. A good friend of mine sent this web article that I’ve found useful, and provides some of data you ask about…
    View at Medium.com



  10. Ben
    March 14, 2020 at 11:32 pm

    One interesting resource is the ILInet information published as part of the CDC fluview report: https://www.cdc.gov/flu/weekly/index.htm . If you scroll about half way down the page, there’s an interesting chart that shows an uptick in hospital visits for “Influenza-like illness” in the week ending Mar. 7. Part of the explanation given below the graph is is “Regions 2, 7, and 10 reported the greatest increases in ILI relative to their baselines. Clinical laboratories in regions 2 and 10 reported a decrease in influenza virus circulation; however, these are areas of the country where COVID-19 is most prevalent and more people may be seeking care for respiratory illness than usual at this time. The ILI increase in region 7 appears most likely due to low reporting.” Regions 2 and 10 include New York and Washington.


  11. jim
    March 15, 2020 at 5:24 am

    Johns Hopkins has a well designed summary page that seems to update frequently: https://coronavirus.jhu.edu/map.html but there’s no testing data there. I think the CDC site is poorly designed and infrequently updated: https://www.cdc.gov/coronavirus/2019-ncov/cases-in-us.html. The CDC also has another (equally poor) page with testing numbers: https://www.cdc.gov/coronavirus/2019-ncov/cases-updates/testing-in-us.html


  12. March 15, 2020 at 1:42 pm

    Using the data that Ben mentioned, Abe Stanway put this together showing flu like symptoms at NYC hospitals: https://medium.com/@astanway/real-time-covid-19-tracking-e6b92001edb


  13. Tech17
    March 18, 2020 at 12:07 pm

    One of the ways people in the Seattle area have managed to get themselves tested is to enroll in the University of Washington flu virus study. The first thing that happens is you are sent a sample kit which you send back and they automatically test for what you have, including the virus that causes COVID-19. It’s a backdoor that worked in this particular “hot spot”.

    I’ve encouraged local leaders to start thinking of a major goal being accurate stats about this infection. That means broad testing, not just testing people who are in critical need (the policy to date). I’ve also suggested they deploy the flu tests that are already plentiful and available so you can eliminate people who simply have a flu that looks/feels like COVID-19 but in all likelihood they do not have COVID-19 … not 100% certainty but saves some resources obviously.

    My recommendations have apparently been passed on to the medical leadership now and perhaps will have a positive effect. (I learned a long time ago never to underestimate the negative amplification factor public administration can bring to any situation …)


  14. March 21, 2020 at 2:34 pm

    Reblogged this on Managementpublic.


  15. tucsonSteve
    March 25, 2020 at 10:26 pm

    what are false neg and false pos of pcr tests that could be rolled out quickly and in large numbers in the US?
    how is SKorea able to test 10k people per day and US is not? What are false neg/pos of Skorea pcr test? Is it a good test or is it garbage?
    Why is no one asking these questions?


  16. M.
    March 26, 2020 at 4:35 am

    Google and Wikipedia will NOT help you here.

    I know *that* there was an agency that tracked this when I was in grad school, but I forget what it is off the top of my head. (I want to say Centers for Medicare and Medicaid Services or Agency for Healthcare Research and Quality?) I also forget whether it is (1) publicly available information, (2) information you have to pay for (i.e., access through your university’s LexisNexis/PubMed/etc. subscription), and/or (3) information you have to FOIA yourself.

    Option #3 is the “wait, what?!” one. You used to be able to FOIA private commercial information like this from the relevant federal agency for the purposes of research or reporting. So when I was in grad school, I FOIA’d the FDA to get all the research done by pharmacetical firms so that I could avoid publication bias in my metaanalyses. The trick was knowing which federal agency or agencies had the information you were looking for.

    That said, the Supreme Court recently ruled in Food Marketing Institute v. Argus Leader Media that business records like these are “confidential” and thus exempeted from federal FOIA requests. I don’t know what researchers are doing now, but I can say that if I were still in grad school or academia, I’d start by FOIA-ing the equivalent EU or Australian agencies. But I’ve been out of the game for a while now so they may be having the same FOIA issues…


  1. March 16, 2020 at 4:42 pm
Comments are closed.
%d bloggers like this: