Forecasting the impact of Covid-19 using an epidemiological model trained on real world data

The Modeling Covid-19 (MC19) project is an epidemiological model of Covid-19 fit to actual social distancing, testing, and fatality data. We use this data to project how Covid-19 might spread through a population for different locations and different social distancing scenarios.

A model is only as good as the data it’s based on, and we’re thankful for the many people and organizations who have worked to produce the data that powers the model. All data has its caveats and limitations: due to (still) limited testing, cases are undercounted to a large degree that varies by state. Fatality counts are generally seen as more reliable, but analysis of all-cause mortality indicates that they are often still underreported. We’ve tried to make the best of the available data and hope to continually improve the model as more data becomes available.

Questions? Feedback? Contact us at

Model run on
Recent & upcoming changes

We use data from the Covid Tracking Project, mobility data from Google and Unacast, hospital capacity data from Esri, and demographic data from Wolfram.
“All models are wrong, but some are useful.”

The impact of social distancing

To illustrate how social distancing can impact the spread of Covid-19, we’ll use an example location  —  Connecticut  —  and consider two scenarios: one without social distancing and one with significant social distancing.

What might happen if Connecticut stops social distancing and returns to normal?

The model projects that if the virus is allowed to spread uninhibited without social distancing measures, the current number of people who require hospitalization will quickly exceed the available hospital capacity in Connecticut, which will increase the total number of fatalities.

Now consider an alternate scenario, with a social distancing period:

What might happen if Connecticut enacts a policy that results in social distancing levels — the same amount as  — ?

In this example, the model predicts that the social distancing period slows the spread of the virus, which allows the current number of people who require hospitalization to remain below hospital capacity, and lessen the total number of fatalities.

The contrast between the outcome of the scenario with a social distancing period and the scenario without distancing measures illustrates how social distancing “flattens the curve.”

From here, we can zoom out. How might Covid-19 affect the people who live in Connecticut?

How could distancing affect the population?

Our model is based upon a standard epidemiological model called the SEIR model. The SEIR model is a compartmental model, which estimates the spread of a virus by dividing the population into different groups:

  • Susceptible people are healthy and at risk for contracting Covid-19.
  • Exposed people have Covid-19 and are in the incubation period; the model assumes most exposed people cannot infect others.
  • Infectious people have Covid-19 and can infect others.
  • Hospitalized people are currently in the hospital or ICU. As a simplifying assumption, we do not model susceptible healthcare workers. As a result, the model assumes hospitalized people cannot infect others.
  • Recovered people have had Covid-19 and are considered immune to re-infection. Our model assumes that the typical immune response will last “at least a year.”
  • Deceased people have passed away due to Covid-19.
Healthcare workers are at a higher risk of contracting Covid-19. The fraction of overall reported cases who are healthcare workers was observed as 10% in China , 10% in Italy , and 15.8% in Ontario. Modeling this scenario is complicated: healthcare workers are also part of the overall susceptible population and have a higher likelihood of being tested.

If Connecticut returns to normal without any social distancing, the model projects that Covid-19 will quickly spread through the population:

If Connecticut enacts a policy that resulted in social distancing , the model projects that Covid-19 cases would stabilize:

Comparing the two examples demonstrates how social distancing plays a significant role in how the virus might spread. The model projects that scenarios with little or no social distancing result in the virus spreading through the population quickly, while scenarios with significant social distancing suppress the virus and result in fewer cases overall.

However, we also need to consider what happens after our distancing period ends. What might happen if we return to normal after social distancing?

The second wave

Let’s look at our social distancing example on a longer timeframe. If we continue to model further into the future, we can project what might occur if we return to normal:

By the end of the social distancing period the majority of the population is projected to still be susceptible to Covid-19. While the model projects that the social distancing measures drive cases down to nearly zero, some cases remain. If left unchecked, the model predicts that these cases will cause another outbreak, creating a second wave of infections.

The second wave appears similar to the example without social distancing above. The model projects it will put a similar level of strain on Connecticut’s hospital system:

Is it possible to avoid the second wave? One option is to continue social distancing until a vaccine is developed, which experts estimate will take at least a year. A vaccine would allow the population to develop herd immunity without requiring mass infections.

“Test and trace” is another approach that involves tracking the virus to identify and suppress future outbreaks without establishing herd immunity. This approach could allow distancing restrictions to ease, but only if certain conditions are met.

Test and trace

A “test and trace” approach couples high testing rates with widely deployed contact tracing. Contact tracing attempts to reduce the spread of a disease by identifying, notifying, and testing people who have recently been in contact with an infected person and encouraging infected and at-risk people to quarantine.

Data from South Korea shows that with a “test and trace” strategy it’s possible to ease distancing restrictions and still prevent exponential growth for a period of time. Specifically, the data suggests that distancing restrictions can be reduced when the number of new cases per day falls below two cases per million people. To keep the virus contained, data suggests that for every 100 tests performed, only one should come back positive and 80% of positive tests should be able to be traced back to another known case.

Interact with the model

We've run the MC19 model state-by-state for different future social distancing scenarios and plan to add more states to the model as more data becomes available. Select a state below to view the model’s predictions and compare them to actual data:

Modeled states in the U.S.