MUSC University Hospital Impact Modeling

Four model scenarios were developed based on the following pattern of growth rates of new infections. These projection are assuming that the growth rate in the coming weeks will decline in response to community behavioral change such as enhanced mask use and physical distancing.

Note: These projections are related only to MUSC University Hospital, and do not include patients hospitalized at other area hospitals.

  Worst Case Scenario Slow Decline in Growth Rate Moderate Decline in Growth Rate Fast Decline in Growth Rate
July 1 – July 15 9% 8% 7% 6%
July 16 – Aug 1 8% 7% 5% 4%
Aug 2 - Aug 15 6% 5% 3% 2%
Aug 16 – Aug 31 2% 1% 1% 1%

MUSC has developed a local impact model to estimate the demands on MUSC University Hospital from COVID-19. We used assumptions in the model as shown below. We began by specifying the demographic characteristics of each affected county. We then used DHEC daily reports on the number of confirmed cases in each county to map the estimated number of cases by age group. Then an impact matrix is used to transition these cases through the stages of illness, including hospitalization, ICU admission, and use of invasive ventilation, mortality, and recovery. Data used in this impact matrix are derived from published literature derived from other epidemics. We are updating the impact matrix with new data as they become available, and as we process more cases at MUSC we will transition to using our own internal data. The sources needed to provide care are also estimated over time for each possible scenario. Two key model parameters are (1) the prevalence of viral infection, and (2) how quickly the epidemic progresses over time. To quantify these two parameters we are using estimates of accumulated cases provided by DHEC, and fitting the epidemic curve to that seen in other locations (see graphic below for illustration). Adjustments are made for undiagnosed cases following DHEC estimation techniques. DHEC guidance is that estimated total cases is approximately 6 times the number of confirmed cases. The model is regularly updated based on calibration to what we see in terms of number of cases identified, and from data derived from our own health system over time. We would like to note that in this latest iteration of the model we have reduced the inflation factor of cases based on the DHEC guidance. This was done because the number of cases presenting in hospital was far less than would be predicted than with the inflation factor added. The current iteration of the model results shown for Charleston area now are more closely calibrated to the trends in hospitalization. We will monitor these values and recalibrate in future as needed.

Model Assumptions:

  1. Percent infected but are asymptomatic or have very mild symptoms = 33%
  2. PPE per day per patient = 21
  3. Average duration of hospitalization in standard bed = 7
  4. Average duration in ICU bed = 9
  5. MUSC’s market share of COVID-19 patients in Charleston Tri-County area:
    1. 34% Charleston
    2. 19% Berkeley
    3. 16% Dorchester
  6. Percent of ICU patients that will need a ventilator = 80%
  7. The following impact matrix is used for progression through illness:
Matrix model assumptions
Age   % Symptomatic who are hospitalized  % Admitted to ICU bed  % Fatal
 0 to 9  0.1%  5.0%  0.002%
 10 to 19  0.3%  5.0%  0.006%
 20 to 29  1.2%  5.0%  0.03%
 30 to 39  3.2%  5.0%  0.08%
 40 to 49  4.9%  6.3%  0.15%
 50 to 59  10.2%  12.2%  0.6%
 60 to 69  16.6%  27.4%  2.2%
 70 to 79  24.3%  43.2% 5.1% 
 80+  27.3%  70.9%  9.3%


Typical Duration of the COVID-19 Outbreaks

Average daily change in total cases over the previous 7 days.

What Does This Graph Show?

A question on nearly everyone’s minds has been how long the COVID-19 epidemic waves will last, and how much longer people will have to be under mobility restrictions and lockdowns. While it is too early to tell for the U.S. and for the Charleston area specifically, we can look to other countries who encountered an outbreak earlier than us. Probably the best signal for how long an epidemic wave will last is the experience of other epidemics. If they show similar temporal patterns in how they progress, that reinforces that this is a pattern that would be seen elsewhere. It should be noted that these trends in the epidemic progression are all taking place under a high degree of social distancing. Should that intervention be removed, there is a significant chance of additional waves of transmission.


This graph plots the daily number of new deaths over time for multiple cities with large and advanced or advancing epidemics. The number of new deaths is plotted as a 3-day rolling average, which is a statistical method of smoothing out a curve to account for sudden irregularities in the data. The horizontal axis indicates the number of days since 10 daily deaths was first recorded. Moving from right to left shows you the passage of time, so you can see that Mainland China is furthest to the right because it has experienced the epidemic the longest duration of any country. Looking to the situation in China for guidance, we can expect that the number of daily new deaths will become severely reduced after about 2 months. You can see from the graph that other countries are following a very similarly shaped steep curve compared with China. We can therefore anticipate that the curve for the U.S. (light blue line) will start to come back down in the next few weeks, and that the average duration of an epidemic wave seems to be approximately two months. As these other epidemics mature in the next few weeks we will gain more insight to how consistent this pattern of rapidly escalating epidemics waves of two-months duration.