Skip to Main Content

Much of the current discourse on — and dismissal of — the Covid-19 outbreak focuses on comparisons of the total case load and total deaths with those caused by seasonal influenza. But these comparisons can be deceiving, especially in the early stages of an exponential curve as a novel virus tears through an immunologically naïve population.

Perhaps more important is the disproportionate number of severe Covid-19 cases, many requiring hospitalization or weekslong ICU stays. What does an avalanche of uncharacteristically severe respiratory viral illness cases mean for our health care system? How much excess capacity currently exists, and how quickly could Covid-19 cases saturate and overwhelm the number of available hospital beds, face masks, and other resources?

This threat to the health care system as a whole poses the greatest challenge.


As I initially described in a Twitter thread, simple mathematics can derive rough estimates for how this might play out.

This exercise can inform our level of urgency and equip us to anticipate non-obvious, second-order effects, some of which can be mitigated with proper preparation.


As of March 8, about 500 cases of Covid-19 had been diagnosed in the U.S. Given the substantial underdiagnosis at present due to limitations in testing for the coronavirus, let’s say there are 2,000 current cases, a conservative starting bet.

We can expect a doubling of cases every six days, according to several epidemiological studies. Confirmed cases may appear to rise faster (or slower) in the short term as diagnostic capabilities are ramped up (or not), but this is how fast we can expect actual new cases to rise in the absence of substantial mitigation measures.

That means we are looking at about 1 million U.S. cases by the end of April; 2 million by May 7; 4 million by May 13; and so on.

As the health care system becomes saturated with cases, it will become increasingly difficult to detect, track, and contain new transmission chains. In the absence of extreme interventions like those implemented in China, this trend likely won’t slow significantly until hitting at least 1% of the population, or about 3.3 million Americans.

What does a case load of this size mean for health care system? That’s a big question, but just two facets — hospital beds and masks — can gauge how Covid-19 will affect resources.

The U.S. has about 2.8 hospital beds per 1,000 people (South Korea and Japan, two countries that have seemingly thwarted the exponential case growth trajectory, have more than 12 hospital beds per 1,000 people; even China has 4.3 per 1,000). With a population of 330 million, this is about 1 million hospital beds. At any given time, about 68% of them are occupied. That leaves about 300,000 beds available nationwide.

The majority of people with Covid-19 can be managed at home. But among 44,000 cases in China, about 15% required hospitalization and 5% ended up in critical care. In Italy, the statistics so far are even more dismal: More than half of infected individuals require hospitalization and about 10% need treatment in the ICU.

For this exercise, I’m conservatively assuming that only 10% of cases warrant hospitalization, in part because the U.S. population is younger than Italy’s, and has lower rates of smoking — which may compromise lung health and contribute to poorer prognosis — than both Italy and China. Yet the U.S. also has high rates of chronic conditions like cardiovascular disease and diabetes, which are also associated with the severity of Covid-19.

Support STAT: If you value our coronavirus coverage, please consider making a one-time contribution to support our journalism.

At a 10% hospitalization rate, all hospital beds in the U.S. will be filled by about May 10. And with many patients requiring weeks of care, turnover will slow to a crawl as beds fill with Covid-19 patients.

If I’m wrong by a factor of two regarding the fraction of severe cases, that only changes the timeline of bed saturation by six days (one doubling time) in either direction. If 20% of cases require hospitalization, we run out of beds by about May 4. If only 5% of cases require it, we can make it until about May 16, and a 2.5% rate gets us to May 22.

But this presumes there is no uptick in demand for beds from non-Covid-19 causes, a dubious presumption. As the health care system becomes increasingly burdened and prescription medication shortages kick in, people with chronic conditions that are normally well-managed may find themselves slipping into states of medical distress requiring hospitalization and even intensive care. For the sake of this exercise, though, let’s assume that all other causes of hospitalization remain constant.

Let me now turn to masks. The U.S. has a national stockpile of 12 million N95 masks and 30 million surgical masks for a health care workforce of about 18 million. As Covid-19 cases saturate nearly every state and county, virtually all health care workers will be expected to wear masks. If only 6 million of them are working on any given day (certainly an underestimate) they would burn through the national N95 stockpile in two days if each worker only got one mask per day, which is neither sanitary nor pragmatic.

It’s unlikely we’d be able to ramp up domestic production or importation of new masks to keep pace with this level of demand, especially since most countries will be simultaneously experiencing the same crises and shortages.

Shortages of these two resources — beds and masks — don’t stand in isolation but compound each other’s severity. Even with full personal protective equipment, health care workers are becoming infected while treating patients with Covid-19. As masks become a scarce resource, doctors and nurses will start dropping from the workforce for weeks at a time, leading to profound staffing shortages that further compound the challenges.

The same analysis applied to thousands of medical devices, supplies, and services — from complex equipment like ventilators or extracorporeal membrane oxygenation devices to hospital staples like saline drip bags — shows how these limitations compound one another while reducing the number of options available to clinicians.

Importantly — and I cannot stress this enough — even if some of the core assumptions I’m making, like the fraction of severe cases or the number of current cases, are off even by several-fold, it changes the overall timeline only by days or weeks.

Unwarranted panic does no one any good, but neither does ill-informed complacency. It’s inappropriate to assuage the public with misleading comparisons to the seasonal flu or by assuring people that there’s “only” a 2% fatality rate. The fraction of cases that are severe really sets Covid-19 apart from more familiar respiratory illnesses, compounded by the fact that it’s whipping through a population without natural immune protection at lightning speed.

Individuals and governments seem not to be fully grasping the magnitude and near-inevitability of the national and global systemic burden we’re facing. We’re witnessing the abject refusal of many countries to adequately respond or prepare. Even if the risk of death for healthy individuals is very low, it’s insensible to mock decisions like canceling events, closing workplaces, or stocking up on prescription medications as panicked overreaction. These measures are the bare minimum we should be doing to try to shift the peak — to slow the rise in cases so health care systems are less overwhelmed.

The doubling time will naturally start to slow once a sizable fraction of the population has been infected due to the emergence of herd immunity and a dwindling susceptible population. And yes, societal measures like closing schools, implementing work-from-home policies, and canceling events may start to slow the spread before reaching infection saturation.

But considering that the scenarios described earlier — overflowing hospitals, mask shortages, infected health care workers — manifest when infections reach a mere 1% of the U.S. population, these interventions can only marginally slow the rate at which our health care system becomes swamped. They are unlikely to prevent overload altogether, at least in the absence of exceedingly swift and austere measures.

Each passing day is a missed opportunity to mitigate the wave of severe cases that we know is coming, and the lack of widespread surveillance testing is simply unacceptable. The best time to act is already in the past. The second-best time is right now.

Liz Specht is the associate director of science and technology at The Good Food Institute.

  • I’m not sure that math is that good. Italy has a total of 21,000 cases today and a population of 60,000,000. If I assume the total infected population in Italy reaches 3x the current number or 60,000. That gives a 0.1% infection rate. If that same rate holds true in USA then roughly 320,000 people will get infected. And 10 to 50% will need hospitalization. So 160,000 max. We can handle that in our hospitals – if evenly distributed.. Also don’t forget it is essentially over in China and their Total number of infected people is 81,000.. This isn’t as bad as the article indicates by a long shot. It is bad, but it is not the end of the world. For some it will be much worse than for others, no doubt.

    I certainly hope my math is more on the correct side. Please help to flatten the curve. This will be a difficult time, but we will make it.

  • I am pretty sure that your math using doubling times of 6 days is not right or meaningful. Initially, when cases are few, doubling times might be a reasonable rough estimate. However, as the total number of cases becomes much larger over many increments of 6 days, the doubling time will not remain the same as many of the previously infected persons are recovered and are no longer able to transmit the virus. Hopefully someone with better math skills can suggest what better than doubling times can provide a better projection

    • When you “describe previously infected persons are recovered and are no longer able to transmit the virus”, this is herd immunity and will start to slow the spread at 60-70% infection of the total population.

      It’s evident from progression in China, South Korea, etc. that lockdown really works in slowing this down. The exponential curve is likely to continue unless these steps are taken.

  • I appreciate the expertise of the author of the article. I would like to subscribe, but there is no indication of how much it costs. I hate being forced to sign up for your 30 days free in order to find out how much it will cost. It’s a trap. You are hoping I’ll sign up and forget to unsubscribe. I hate you for that. I don’t have much money, so it’s a big decision for me. Please be upfront about the cost. You seem to be people of integrity.

  • Hospital beds are nowhere as bad as you are indicating. Doctors are already postponing elective surgery due to Corona virus. And, many procedures can be done on an outpatient basis even if the preferred protocol is to have a day or two in the hospital. We also have military and temporary hospitals available if needed.

    That said, the number of doctors, P.A.s, nurses, and orderlies is somewhat fixed. The goal of the governmental agencies is to reduce the “surge” of patients and spread out the need for hospitalization to a manageable rate.

    • Additionally, we are now coming out of the flu season. At this time of year there should be thousands of beds coming available each week. Not sure what the actual numbers are, but there are millions of hospitalizations annually for the flu.

  • Early on and we are in EARLY on in a pandemics course, that rate of doubling will hold true. But it will level off/decrease over time. It can’t double every 6 days until it exceeds the entire population of earth. Rough guess it will infect 40% of the world’s population according to CDC. No point in debating that number. It is a highly educated guess in the first place. And they don’t have a lot of data yet to make more accurate predictions. And don’t forget “You can’t fool Mother Nature”. She is doing her thing and we are just along for the ride.

    There is enough information out there to help blunt the impact. Governments are becoming very proactive. Ours finally decided it was not totally contained and a hoax, The delay is proving to be costly.

  • Late April to early May we will start getting a more accurate description of the damage this virus can do. It has an incubation period of 3-14 days, typically you start showing symptoms in 3-5 days. On average it took about a month to kill those who have died from it. The mortality rate will never be truly known unless everyone gets tested, and that will never happen. It will all be speculation. In my opinion, if we do not implement a nationwide shutdown immediately then millions will die in this country alone, either directly or indirectly before it’s all said and done. If 100 million people are infected in the U. S., which is ultra conservative, and even 10% require medical care that’s 10 million people. It will peak at different times depending on location. So let’s say Cook County Illinois peaks with 1 million people infected simultaneously in various stages, 10 percent requiring a hospital bed, respirators, breathing apparatus, ICU equipment, etc. So 100,000 people in serious to critical condition. There are 33,000 hospital beds in Illinois. My example is of one county in Illinois, though it is the largest. And my numbers are very conservative. There are 5.2 million people in Cook County. My point – If a nationwide lockdown is not compulsorily implemented immediately then the U.S. is in dire straits indeed. The 3.5, alleged, mortality rate or the 1 percent predicted final mortality rate is a farce. It will be exponentially higher depending on how long we wait to do what is necessary.

  • The homeless and those in limited crowed housing will at this point end up in hospital beds, California already has a this problem.

    Replacement home-care workers don’t cross between agencies. Locally I demand in Philadelphia that the recently closed Hahnemann Hospital reopen for flu patients, see Silly Panic Silly Panic Coronavirus

    It is possible that this virus will slow down in hot weather

  • Perhaps you are familiar with Bayes’ Theorem. If you are, you’d be wise not to encourage “widespread surveillance testing.” When testing indiscriminately, you make the test virtually worthless because you end up with more false positives than true positives.

  • Guess you didn’t notice the last part of paragraph about doubling every 6 days” ” this is how fast we can expect actual new cases to rise in the absence of substantial mitigation measures.”

    I found this to be a very helpful “what-if” projection that of course will need to be revised as more actual data come in. It gives us a sense of what we might be facing if we don’t both individually don’t exercise caution and don’t buy time for the health system to marshal their resources and plans.

  • This is a really well written article. I’d love to see an update of her projections, taking into account that most state/communities have now put in place serious “social distancing” efforts since it was written.

Comments are closed.