Reuters built an epidemic model to simulate various aspects of the Covid-19 pandemic and how the virus might spread within a population. We ran the model thousands of times to compare when “herd immunity” would kick in for various reproductive numbers and under a range of scenarios.
The model enabled us to show readers what level of immunity would likely be required in order to stop the spread.
An interactive feature also allows parameters to be adjusted to show things like balancing vaccine distribution, effect of interventions like masks and distancing, and people “travelling” and possibly causing super-spreader
The project provides an evergreen resource for those looking to estimate herd immunity projections or to better understand the concept. It has attracted high readership figures and gathered a lot of attention on social media. The piece was also picked up by other news organisations.
The model is still shared and used widely to date, five months after publication.
We also batch processed 100,000 runs of the simulation locally to give a comprehensive sample of outcomes, allowing us to accurately show where the herd immunity range fell. We were then able to create some separate SVG static graphics from this data to present in the page as part of the explanation.
We also decided to host the full model at the foot of the page which let’s readers adjust parameters in real time for a reactive experience.
What was the hardest part of this project?
There was no easy part to this project. However, two main challenges stood out above the rest.
Accuracy of an epidemic model
There are a number of variables which can be inserted into an equation to allow an outcome. The challenge for us was to carefully select data which fairly represented the virus and strike a balance between a model which was unrealistically simple and overly complicated, potentially introducing a wider opportunity for error. We then had to make sure all of the behavioral elements between the variables were correct, including social interventions like wearing masks and social distancing.
We worked with epidemiologists and mathematical modeling academics for months to ensure our model was as accurate and realistic as possible, while also making clear the simulation is based on set parameters and assumptions. We also added the interactive at the bottom of the page to allow these parameters to be adjusted.
Running in the browser
Another challenge was ensuring smooth delivery in the browser while hundreds of thousands of calculations are being made live.
What can others learn from this project?
Sometimes if there is little or no hard data available on a subject, it may be an opportunity to make something completely new and provide a service to readers. It can be a daunting task to build something like this from scratch, especially when a team doesn’t have in-house expertise in epidemiology. Taking the extra time to work with experts and learn the theory of the subject can provide a solid foundation for an ambitious project.