Modeling when existing hospital beds fill with COVID-19 patients, by US state, as of November 23, 2020

About this model

This model attempts to determine at what point hospital beds will be filled by COVID-19 patients and hospitals will be above capacity.

We look at number of available hospital beds by US state and compare that to the number of COVID-19 patients that will need a bed. Assumptions are listed below. It takes time between when a patient gets infected, when they show symptoms, and when they go to the hospital, so the number of people who test positive and the death numbers we see in the news are lagging indicators. This model adjusts for that, and takes into account the case doubling rate of the epidemic.

See the underlying data

Note: Hospital beds vs. ICU beds & ventilators

This model estimates at what point hospitals in individual states will reach hospital bed capacity with COVID-19 case patients. The healthcare system needs regular beds, ICU beds, doctors, nurses, ventilators, medicines, masks, and other supplies. We use beds as a proxy for all these but understand that shortages of these items in any location will restrict treatment of the sick. And, in a crisis, beds can be set up in gymnasiums, supplies and machines can be shipped to critical hotspots, ventilators can be purchased, and retired doctors can be asked to help.

Results

Key Variables


Doubling rate for the epidemic
days for the total number of cases to double (November 23, 2020 and later)
The US, Spain, France, Germany, UK, Sweden, and Brazil were all doubling every 3 days until recently. Current social distancing measures in some countries are now increasing the doubling rate. Try adjusting the assumptions to calculate how much social distancing is needed. China, which has currently contained the outbreak, is at 33 days. Italy is currently at 5 days. Source: https://ourworldindata.org/coronavirus
days for the total number of cases to double (before November 23, 2020)
Estimate current cases (adjusts for lag between infection and test)
(Change switch to "Method 1" OR "Method 2")
Method 1

Estimate current cases based the number of deaths (if available)

If there are not deaths in a particular state, use Method 2 for that state

On average, a patient who died would have become infected 23.6 days ago (or more). From the inputs above, we use the 23.6 days as well as the death rate of 3.0% to estimate how many people were infected 23.6 days ago. Then we calculate the current number of cases using the doubling rate of 3.0%. In Washington state, when extrapolating from the number of deaths, the model intentionally does not add in the people from the same nursing home who died, providing a more conservative estimate.

OR

Method 2

Estimate current cases based on official testing, adjusted for time delays

To get tested in the US right now, most people (we assume 90%) must have pneumonia or other severe symptoms. That means that those people contracted the disease approximately 12.1 days prior to being tested. They were part of a cohort of infected people, some of whom got better. We use the hospitalization rate (12.7%) to estimate the size of that cohort and then grow that number to the present day at the doubling rate of 3.0%.

Method 3 (not recommended)

Estimate current cases based on the official testing numbers

This uses the official numbers of cases as published by Johns Hopkins CSSE and only includes people who have been tested. Keep in mind that in the US many people are currently not tested. Also the testing results are for people who are already symptomic, so those people were infected generall 12-23 days ago -- meaning this is a lagging indicator.

Hospital bed availability

% of normally operating hospital beds are usable for COVID-19 patients

34.1% was the average hospital vacancy rate in 2017. Hospitals can increase this % by deferring elective surgeries and adding temporary beds. We use 50% as the current surge capacity.

Source: https://www.statista.com/statistics/185904/hospital-occupancy-rate-in-the-us-since-2001/

Hospitalization and fatality rate
% of COVID-19 cases result in fatality

Estimates for this number range from 0.84% (S. Korea) - 3.5% (WHO est.)

Source: https://www.businessinsider.com/coronavirus-death-rate-south-korea-compared-to-flu-2020-3

Source: https://wwwnc.cdc.gov/eid/article/26/6/20-0320_article

% of those who enter the hospital (people with severe symptoms) will die

Estimates range from 28% (China study 1), 12.4% (China study 2), 15% (China study 3), 4.3% (China study 4), to 50% (Italy). We use the average of those, 23.7%.

China study 1 (Jan): https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30566-3/fulltext

China study 2: https://jamanetwork.com/journals/jama/fullarticle/2762130

China study 3: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30183-5/fulltext

China study 4: https://jamanetwork.com/journals/jama/fullarticle/2761044?guestAccessKey=f61bd430-07d8-4b86-a749-bec05bfffb65

Timeline for an individual case's disease progression
days after infection, people arrive at the hospital if they have severe symptoms

Median incubation time is 5.1 days. After the onset of symptoms, 1 week (7 days) to the development of severe disease and hospital admission.

5.1 + 7 = 12.1 days

Source: https://www.who.int/docs/default-source/coronaviruse/who-china-joint-mission-on-covid-19-final-report.pdf

Source: https://jamanetwork.com/journals/jama/fullarticle/2761044?guestAccessKey=f61bd430-07d8-4b86-a749-bec05bfffb65

days median hospital stay for a patient who dies

Average time to death from initial symptoms is 14 days or 18.5 days; subtracting 1 week to develop severe disease yields 7 days or 11.5 days in the hospital.

Source: https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30195-X/fulltext

Source: https://www.medscape.com/viewarticle/926504

days median hospital stay for a patient who survives

Median time to discharge after initial symptoms is 17 days or 22 days; subtracting 1 week to develop severe disease yields 10 days or 15 days in the hospital.

Source: https://www.medscape.com/viewarticle/926504

Source: https://jamanetwork.com/journals/jama/fullarticle/2761044?guestAccessKey=f61bd430-07d8-4b86-a749-bec05bfffb65