UPDATED March 24, 2020: As the number of confirmed coronavirus cases continues to grow exponentially, the capacity of the hospital system to treat these cases is becoming a concern. The goal of “flattening the curve” is that testing, isolation, and social distancing will slow the increase of new cases. Hopefully, these efforts reduce the numbers of new patients who require hospitalization to a rate that hospitals can handle.
In this post, I’ll identify the top 10 states in the United States that have the greatest likelihood of experiencing hospital capacity problems if coronavirus cases continue to grow exponentially. To recognize these states, I’ll assess per capita rates for both coronavirus infections and hospital beds. I’m looking for states that have a relatively large number of coronavirus cases given the size of their population and have a relatively low number of hospital beds.
The Exponential Growth of Coronavirus Cases and Hospitals
As you may have heard, most people infected by the coronavirus have mild symptoms. However, 10-15% of the infected require hospitalization. At the moment, U.S. hospitals are experiencing an unusually high usage of their resources, but they’re not currently overwhelmed. However, the CDC is warning hospitals to prepare for shortages.
Unfortunately, given the exponential growth of new cases, the levels we see today can increase to shockingly high numbers in a short timeframe. A study in Lancet suggests that the coronavirus doubles every six days. Some reports indicate a quicker doubling rate, but they likely represent data skewed by insufficient testing.
When I wrote this on March 19, there were 14,250 confirmed cases in the United States. If new cases double every six days for two months, we’ll have 15 million cases in late May!
Exponential growth of coronavirus cases requires the following conditions:
- Regular contact between people.
- Large numbers of uninfected people.
- No effective vaccination or natural immunity.
The goal of quarantines, lockdowns, and social distancing is to reduce the “regular contact” requirement. With these measures, we’ll hopefully be able to slow the growth and not have 15 million cases in 60 days. How many less depends on how well people follow the protocols. Governor Newsom of California expects that 56% of Californians will be infected within that span of time, which equates to 22 million people. However, that seems to be based on a doubling rate of every four days. There are still large uncertainties surrounding the virus and the effectiveness of the efforts to slow it down.
Exponential growth highlights why taking action early, before the number of infections is unmanageably large, is crucial.
Related post: Coronavirus Curves and Different Outcomes
Comparing Coronavirus Cases to Hospital Beds by State
To identify states with the highest potential for hospital shortages, I’ll assess coronavirus cases per capita and hospital beds per capita. We’re looking for states that, given the size of their population, have relatively more coronavirus cases and fewer hospital beds than you’d expect.
Overall, states have a median infection rate of 5.3 confirmed cases per 100,000 people, which is up from 1.9 when I originally wrote this post four days of ago. Of course, that rate will continue to grow with time. States have a median of 250 hospital beds per 100,000. Typically, two-thirds of those hospital beds are occupied. To identify the 10 states with the greatest potential for shortages, I’ll assess the ratio of coronavirus cases to hospital beds. States with potential problems will have higher ratios—in other words, more cases to beds. The list below starts with the worst ratios. The median ratio for all 50 states is 0.0215.
State | Cases / 100,000 | Beds / 100,000 | Ratio | |
1 | New York | 107.4 | 270 | 0.398 |
2 | Washington | 28.5 | 170 | 0.168 |
3 | New Jersey | 31.8 | 240 | 0.133 |
4 | Louisiana | 25.2 | 330 | 0.076 |
5 | Colorado | 12.0 | 190 | 0.063 |
6 | Connecticut | 11.6 | 200 | 0.058 |
7 | Vermont | 11.9 | 210 | 0.057 |
8 | Michigan | 13.2 | 250 | 0.053 |
9 | Massachusetts | 11.1 | 230 | 0.048 |
10 | Rhode Island | 10.0 | 210 | 0.048 |
Click here for my full CSV dataset with the complete set of states: Coronavirus_hospitals_03232020
Why Are These States on the List?
This list of states has some obvious candidates that the media has discussed frequently. Washington and New York have both been widely reported as being walloped. Washington has fewer cases of the virus per capita compared to New York, but it has notably fewer hospital beds than NY. In the original post, New York and Washington had virtually equal ratios, but now NY has increased far beyond Washington with this update. New York by far has the largest ratio and has the greatest potential for overwhelming its hospital system.
Washington and New Jersey have nearly equal ratios whereas there was a larger gap between them previously. Notably, the Northeast claims 6 of the 10 spots.
Louisiana is a surprise. It was #4 before and remains in fourth place. When I originally wrote this post, I hadn’t heard much about that state in the media. However, today I just started hearing that it has fastest growth rate for new cases in the United States. While it has an above average number of hospital beds, Louisiana has FIVE times the median infection rate! Colorado was also a surprise. Colorado’s infection rate is more than double the median and it has a relatively low number of hospital beds per 100,000 people.
I expected to see California in the top 10 given how much we’ve heard about it on the news. However, it falls at #21. California’s infection rate equals the median. This state has a large total number of cases (2108), but it also has a very large population at 40 million, which brings the per capita value down.
With this update, Michigan appears on this list. It was previously at #17 but it is now #8. Maine was previously #10 on the list but falls to #17.
Closing Thoughts
We’re not experiencing hospital capacity shortages now. However, it is instructive to look at where more cases are occurring and seeing how hospital beds stack up. Additionally, the hope is that social distancing and quarantines will prevent hospitals from being overwhelmed.
There are several caveats about this analysis. First, we’re looking at the current number of cases by state. Our assumption is that states with more cases per capita now will have higher rates later. Additionally, the number of cases confirmed by testing in the U.S. is underreported due to inadequate testing. So, we don’t have perfect numbers, but they’re the best we have. Finally, hospital beds aren’t a perfect measure either. Hospitals depend on other resources such as masks, gowns, gloves, and ventilators, to name a few. These resources might run out faster than hospital beds. However, the assumption is that states with more hospital beds have a higher capacity overall.
This analysis provides a first take on the states that have a relatively high amount of cases and a low number of hospital beds.
Johns Hopkins University provided the coronavirus data. The Kaiser Family Foundation provided the hospital bed data. World Population Review provided the population data.
biju says
thanks sir u helped for my project
Johnson says
Hi Jim,
Thanks for the insight. I have a challenge, how can we calculate rate of infection (DV) as against personal hygiene (IV)
Rohit says
Hi Jim,
Thanks you for the article- simple and objective.
I am wondering if there’s a way we can predict how likely a person is to be infected besides a simple #infected/total population w.r.t states/countries.
Thanks,
Rohit
Jim Frost says
Hi Rohit,
That would be really difficult to calculate. With COVID, we don’t really know the true number who are infected. We just know the number of confirmed cases, which is different. The risk also varies by region and even by the exact place you live. For example, here in the U.S., about one-third of COVID deaths are in nursing homes. You’d also need to measure how much people moving around an interacting in your area. Are people in quarantine overall? Or, are they moving around freely? So, there are a number of very local factors to consider. And unknowns about the true number of cases.
There might be very advanced epidemiological models that can estimate this risk, but I don’t know enough about them. But those are the types of factors you’d need to consider: number of susceptible, number of cases, amount of movement, population density, living conditions, and so on.
Michael O'Hare says
I find your statistial explanations straight forward and easy to follow. However I can’t help thinking in this article you are making a couple of incorrect assumptions, Firstly the expotential growth as you point out requires regular contact between people, but it also requires that contact to be random ie each infected person has an even chance of meeting an unaffected person and that seems unlikely. People associate in quite discrete groups. Herd immunity also comes into play this varies depending on infection rates. Currently numbers anywhere between 20-70% of population. If these factors are taken into account the growth should slow regardless of any mitigation measures. A person becomes infected and all their group is vunerable, as each new person becomes infected the numbers they can infect decreases. At some point one group member affects a non goup member and the process starts again, repeatedly. A consequence of these factors would appear to be that as the pandemic continues the doubling rate must inevitably slow. Surely the lock down measures are only an extension of this slowing the process by making the discrete groups smaller.
Jim Frost says
Hi Michael,
Exponential growth requires that there are many uninfected individuals near an infected individual. The vast majority of the population still has not been infected, so the potential for exponential growth still exists. Experts estimate that herd immunity might kick in when around 60-70% of the population has been infected. However, we want to avoid depending on herd immunity. If we assume that we need only 60% of the population for herd immunity and that the mortality rate is on the low end of the estimates at 0.5%, we talking a million deaths in the United States alone. And that’s using the safer numbers for both measures.
As I mention in this post, herd immunity (the virus can’t easily access uninfected people) will slow the rate. But, we’re nowhere near that point yet. There have been several good studies into the overall infection rates. They range from 2-5% in Santa Clara County, CA to 13% in New York state. New York is the hottest of hot spots in the U.S., so their percentage is going to be higher than the national number. Clearly, we’re nowhere near herd immunity yet.
Lockdowns work by reducing the R0, which is the number of individuals that an infected person will infect on average. As you say, it does this by limiting the number of interactions. The lockdowns have been effective. If it weren’t for them, we’d still be experiencing exponential growth. I show the unrestricted exponential growth rate to illustrate how effective the lockdowns have been. We avoided those severe numbers.
Mars says
Hello, thank you so much. Its comforting to see objective analysis, based on consistent methodology given that there are so many unknowns.
Could you possibly please kindly provide an update of this post?
Jim Frost says
Hi Mars,
Thanks so much for your kind words. I’ll update this post within 24 hours.
Jerry says
Good article, Jim. In this epidemic it is also useful to use age-adjusted rates, because the virus hits older age groups much harder and different states have different age structures. It’s also noteworthy that in my state (Texas) and presumably other states, ‘elective’ and ‘nonessential’ medical procedures are curtailed during the emergency. This will free up both hospital beds and staff. But don’t forget that other people have to be hospitalized too, for heart attacks, cancer, etc. Another idea, and maybe you did this already (i’m behind on reading these), is to incorporate estimates of the R0 (“R-nought) number —the number of people each infected person spreads the virus to— in these calculations for how fast the virus spreads. Unfortunately, the current situation provides good “real-world” exercises for these epidemiological calculations !
Adrian says
Great article, Jim. Thx for putting this together. What jumps out at me is the 7 of 10 northeastern states (I include NJ therein) already leading the pack in terms of nearest at risk for over-burden. It’s not like they’ll be able to “locally” share burden if it comes to that.
Jim Frost says
Hi Adrian,
I noticed that too. Number 11 is New Hampshire. So, that region seems like it could face difficulties.
Hopefully, the social distancing measures are effective!
Sean Saunders says
Hi Jim – it’s actually 15,000,000 by May, not 1,500,000.
Furthermore, many amateur forecasts we are seeing assume unbounded exponential growth, but obviously, we would expect some sort of carrying capacity effect, as the proportion of infected (past or current) increases in the population. How and when do you see that impacting the growth model for the number of reported cases?
To that end, what would the data need to look like in order to signal that we are indeed succeeding at “flattening the curve”?
Many thanks for your analysis!
Jim Frost says
Hi Sean,
Decimal places and the wee morning hours don’t mix well!
I agree, clearly, the extreme forecasts assume that we’re doing nothing and the virus is spreading freely. Hopefully, the social distancing measures make an impact. As for when we might see an effect, the experts are saying that it might take several weeks. The problem is that the tests are detecting infections that already happened–they’re delayed detections. Plus, of course, we’re under testing. So, the latest numbers give us a fuzzy picture of infections that happened maybe a week or so ago.
I think we’ll be able see the effect, whatever the magnitude of it is, by comparing the curves to what we’d expect with unbounded exponential growth. That difference should represent the effectiveness of our social distancing, quarantining measures. At least roughly. As far as I know, there’s no good evidence from prior cases to predict how these measures will work on such a large scale. It’ll probably take much study after the fact to reconstruct what happened and how much the counter measures helped. Again, our picture is so fuzzy and incomplete right now!
shawnjanzen says
I appreciate your use of median rather than mean. I agree with the many caveats you list. Even using the state as the level of analysis can be problematic, even though it is arguably the best one for government policy implementation and data aggregation. Normalizing by population helps, but population density within a state and city permeability can make things far more fluid than state borders show. I’m thinking Mardi Gras and the high levels of tourism this time of the year accounts for the spike in Louisiana.
Jim Frost says
Hi Shawn,
Unfortunately, the best data I have is at the state level. I’m sure there might be localized hot spots might swamp local hospitals. Planners would undoubtedly use more granular data.
I did take a quick look at population density by state but didn’t see anything there at this point. Part of the problem is that confirmed cases by testing represent infections that occurred in the past. And, we’re under-testing, which gives us an incomplete picture. We have a fuzzy, incomplete picture of infections that occurred in the past! I think it’ll take much study after the fact to understand what has happened.
Great points about Louisiana!
Salwa Mousa says
It’s great article. I will try to modified sigmodial in this data.
Thanks
Anne Sydor says
Of the 2/3rds of hospital beds that are typically occupied, how many of those are for people whose hospital-based care could be postponed or redirected? Does Kaiser provide any data for that.
Thanks for the analysis. . . I always love your blog and especially love that you are using it so well to provide rationality in this time of crisis. Anne
Jim Frost says
Hi Anne,
Those are great questions. Unfortunately, I don’t know the answers. I do know that places like New York have asked patients to postpone elective surgeries to free up hospital beds and resources. But, I don’t know how much of an impact that will have.
Thank you so much for your kind words about my blog. They mean a lot to me!
Janardhan Mydam says
This is excellent analysis Jim .. you made the complex data so simple to understand .
Love your posts and the way u present and your teaching skills .. just excellent .
Thank you so much
Dr Mydam