Let me preface this by saying first that I am not a pandemic specialist. I do have a PhD, but in anthropology, not anything useful. As a result, I am leaning heavily on the work of others who have estimated the capacity for the US health care system. While we don’t yet know everything about the spread of coronavirus, we do know a fair bit about ourselves. And that is one of the reasons you are starting to see real panic this week.
The problem isn’t coronavirus so much as our health infrastructure. Because we have a private health care system in the United States, it has been shaped by the free market. Where demand for health services has been high, there is high capacity. Where demand has been low, there is low capacity. As a result, different regions of the country will be better or worse in their response due to historical chance. But first, let’s run some numbers. As of 2018, there were 327.2 million people in the USA. Let’s assume 70% will get the disease. So for the purposes of these calculations, 229 million Americans will be considered. This number will be higher or lower based on our choices in the coming days.
From Wuhan China, we know that roughly 15% of coronavirus cases require care, and that 5% can become critical. We also know that susceptibility to this disease is not evenly distributed — younger people tend to have better outcomes, older people tend to have worse.
The Hubei numbers, which to date remain the best documented, present a good picture on the face of it, a majority of cases (80.9%) were mild. So for the sake of our simple model, let’s assume this is accurate, and only 19.1% of cases will require some level of medical attention, or about 1 out of every 5 people you know right now.
229 million *19.1% = 43.7 million instances of medical attention
Can US hospitals handle this number of cases?
Americans use hospitals less per capita than other developed nations; we have about 2.8 hospital beds per 1,000 people.
If we calculate these numbers, at any given time, America has room for 916,160 people in its hospital on any given day. As a result, America can only, in theory, handle about 1 million cases of coronavirus at any given time, assuming no one has been hit by a car, has cardiac arrest, has cancer, needs surgery, etc. When you add these up, about 61% of America’s hospital beds are being used at any given time. As a result, for the US healthcare system to maintain pace with coronavirus, there can only be 357,302 cases which require hospitalization at any given time. If we add in the the 80.9% of cases which did not require hospitalization, that brings us to our first threshold number:
Threshold 1: America has the capacity to handle 1,870,693 cases (5.7% of the population) of coronavirus at any given time based on hospital beds
Now, by “handle”, I don’t mean satisfactoraly. I just mean that this is roughly the flux of people who can have the disease at the same time without blowing up the US healthcare system. This does not mean every hospital has enough equipment to handle each case — coronavirus is a respiratory illness, and the capacity for that is different. This leads us to our second threshold: ventilators.
According to the CDC, about 20–30% of patients needed respiratory support. This means they needed some kind of ventilator to assist their breathing. Like hospital beds, this is a finite resource, there are only so many that are available. Numbers on these are harder to come by, but not impossible. In 2013,a survey found that there were 52,118 full-service ventilators in the USA, and 98,738 less-than-full-service ventilators. At any given time then, the US has the capacity to assist respiration in 150,856 cases.
A quick note: at this point our assumptions begin to compound, and the potential for under- or over-estimating becomes high. Be skeptical of these numbers, but they should be somewhat ballpark.
Let’s first assume that respirator use closely matches hospital capacity — e.g. 61% are used. That would mean that America can handle 58,833 additional cases of assisted breathing due to coronavirus at any given time.
Remember, 20–30% of patients who needed hospitalization (themselves 19.1% of coronavirus cases) in Hubei needed respirators. That would mean 3.8% to 5.7% of coronavirus cases need a respirator. This gives us a range for threshold 2:
Threshold 2: America has the capability to handle 1,548,236 to 1,032,157 cases of coronavirus at any given time based on ventilators.
This is a bit narrower than our estimate based on hospital beds. In our 70% infection example, that means 8.7 to 13.1 million people will need this respiration assistance. This is the most significant bottleneck, as ventilators will only be available to a small fraction of those people. The good news is that respirator capacity is built with flu season in mind, and we are coming out of it now. As a result, the respirator capacity could be double this number, but no more than triple.
We’ve now got a rough estimate of the capacity the US healthcare system has for coronavirus right now — we can handle a little bit over 1 million simultaneous cases, but not much more than that. This assumes the basic contours of how the virus affects people is accurately represented in Hubei province in China.
But how many cases will America have?
Because of testing failures, it is impossible to estimate how many cases of coronavirus are active today in the United States. What we do know is that the disease is growing exponentially in a similar way around the world.
This growth rate is based on European countries, and all show a similar when starting dates are aligned (e.g. each exponential growth curve starts on a different day). Note that the y-axis is on a log scale (e.g. each major line tick is 10X more than the previous). Data so far indicates a doubling rate somewhere between 4.4 to 7.5 days, with modeling supporting a doubling time of 6.4 days.
This then supports the following model for COVID-19 cases in the United States, which suggests that America will cross 1 million cases by the end of April. Based on threshold 2, that will be when America’s capacity for assisted ventilation will be reached. Assuming doubling continues, by the first week of May threshold 1 will be reached, in which hospitals reach their maximum capacity.
Italy has already reached this point, and ventilators are reserved for younger patients who have a higher chance of surviving with treatment.
This links into the conversation regarding case fatality rate — what % of people will die because they contract COVID-19? Numbers range from 0.5% to 5%, but that ultimately depends on how overwhelmed a health care system is. In South Korea, which runs 10,000 tests a day, the case fatality rate is closer to 0.8%. In Italy, whose health care system is collapsing, the case fatality rate is closer to 6.7% today.
Based on this, one expects that the case fatality ratio will look like South Korea (0.8%) until the capacity of the American health care system to treat patients is breached at the end of April, in which case it will be somewhere between 5–10%
It is impossible to predict the future, but we can estimate best-case and worst-case scenarios. Let’s look at timeline 1, where drastic intervention halts the spread of the disease, and timeline 2, where COVID-19 doubles about every week. Let’s assume cases miraculously disappear in July for the sake of simplicity (spoiler alert: they won’t).
Timeline 1: Drastic measures keep contemporaneous COVID-19 caseloads under 1 million at any time, and doubling does not occur: 8,000 deaths per month for the duration of the epidemic till July: 3 million cases and 24,000 deaths
Timeline 2: Doubling occurs every week, with the capacity of the health care system collapsing in late April/early May. In this scenario, 8,000 people die through the end of April, with the rate exploding in May and June. Calculated by month (very roughly):
April: 1 million cases, 8,000 deaths
May: 30 million cases, 1.8 million deaths
June: 224 million cases, 13.4 million deaths
Again, this assumes the problem goes away by July. These numbers are shocking, and I’ve run them a couple times. If you think I am in error, please let me know and I will happily edit. But this analysis should underscore how absolutely essential it is to Flatten the Curve, e.g. adopt behaviors which minimize the spread of the virus to prevent our health care system from being overwhelmed.
The virus is not the most important variable in the case fatality ratio; the capacity of the health care system is. We must keep the total disease load under 1 million cases per month.
Thank you for reading, if you find that I am in error in any of my calculations please let me know.
