How a family planning model was adapted to guide COVID-19 decision-making
In early 2020, as researchers looked to modeling to mitigate COVID-19, one group of scientists created a tool to pinpoint how specific interventions could change the course of the pandemic

In February 2020, as the world tried to both understand and predict the path of the SARS-CoV-2 virus, a small group of researchers at the Institute for Disease Modeling (IDM) looked for solutions. Their goal was clear: create a mathematical model – quickly – so policymakers could have a roadmap to fight the spreading disease.
“It was started at around midnight, and it was ready by 8:00 a.m., so it was basically just a single all-nighter for developing the first draft,” said IDM Senior Research Scientist Dr. Cliff Kerr, who together with team leader Dr. Daniel Klein burned the midnight oil to release the first iteration of the model they would call Covasim.
Today, Covasim, or COVID-19 Agent-based Simulator, has been through 80 releases. Like many other models, its simulations are based on data on COVID-19's spread, including things like reported cases, deaths, population and age distribution, and number of contacts between people. But its strength is that it can evaluate the effect of various kinds of interventions – non-pharmaceutical or pharmaceutical – down to the level of individuals.
“The reason we focus on interventions is because that is what matters to decision makers,” said Dr. Kerr. Covasim, he added, has helped those on the frontlines of health policy determine the effectiveness of things like lockdowns and mask mandates, and later, vaccinations and vaccine mandates.
Haron W. Gichuhi, an informatics researcher in Uganda who participated in a workshop IDM offered on Covasim in March, agreed: “The fact that Covasim allows one to simulate a ‘what if’ scenario, and test an intervention in this scenario, makes it a formidable tool for fighting pandemics.”
Creating a model that would address intervention outcomes relied on granularity. “Because it is agent-based, you can really look at the connections between people and networks in a lot of detail. You can use it to ask questions such as what would happen to transmission dynamics or case counts if you vaccinated schoolteachers, for example. To do that, the model is able to examine how the behaviors of schoolteachers inside a school, for example where they go and who their close contacts are, could influence the outcome of different interventions,” said Dr. Kerr.
Whereas most models can estimate things like how reducing transmission by a certain percentage in a given population affects disease spread, Covasim allows users to dive deeper, estimating the impact of things as specific as the effects of vaccinating 70 percent of people over 65 in a given population, where mask mandates are no longer in place. Being able to ask scenario-based questions like these aligned with how policymakers were already thinking about the pandemic response. The result? More effective public health policy.
While separating cause and effect is nearly impossible in an unprecedented situation like COVID, Dr. Kerr’s team is proud of the role Covasim may have played in local COVID responses: “One example I'm personally proud of is that for the first phase of the pandemic, the three U.S. states that were using Covasim – Washington, Oregon, and Hawaii – were three of the four states that had the lowest cases and deaths in the entire country.”
Gichuhi said tools like Covasim can not only inform action, but also galvanize change. “The biggest challenge has been misinformation both about the pandemic and the vaccines, leading to low vaccine uptake. Covasim could prove beneficial by simulating different scenarios that would happen if, for example, the vaccination is not wholly taken up. As the famous adage goes ‘a picture is worth a thousand words,’ and these graphical simulations would prompt our policymakers to take action,” he said.
The granularity of its outputs – as well as how quickly it was developed – are not the only things that make Covasim unique. The model was based off another model that had nothing to do with a communicable disease.
“It had a somewhat unusual history. It actually came out of a model of family planning in Senegal. I'm pretty sure it's the world's only COVID model that started as a family planning model, but that was the model we had that was closest to what we needed in terms of being able to account for individual people and their behaviors,” said Dr. Kerr. Today, nearly two years after Covasim’s first version was released, Dr. Klein and his team have ported back learnings to other disease models. “Ironically,” Dr. Kerr added, “one of the biggest technology transfers we had was back to the family planning model.”
Mathematical models that project transmission have existed since the 1760s, when Daniel Bernoulli used predictability to “evaluate the effectiveness of smallpox inoculation.” Since then, disease modeling has been used for everything from HIV to rubella to Ebola, but not to the degree seen during COVID-19. “This is the most significant global pandemic we've had since 1918, when modeling was still in very early stages,” said Dr. Kerr, adding that our current ability to access so much data in real time, powered by super computers, helped every researcher’s efforts. “Bird flu, swine flu … there has been modeling involved. But I think what distinguishes COVID is that it was such an all-hands-on-deck experience,” he added. In a world rushing for answers, data became a tool as essential as masks and vaccines.
Data and modeling, however, are far from fail-proof. Some scientists – like Dr. Fen Chen, a biostatistician from China’s Nanjing Medical University – have argued that the “accuracy of prediction is limited by insufficient, inaccessible, or inaccurate data,” and that for a model to work accurately “greater information sharing and methodological innovation to deal with uncertainty are needed.” Dr. Klein and his team have acknowledged those challenges.
To facilitate collaboration and transparency, Covasim has been fully open-source since day one, with regular updates based on user input and to keep up-to-date with the latest scientific studies. It’s one of the reasons researchers say it’s been so quickly applied in more than a dozen countries in Africa, Asia, Europe, and North America.
For Dr. Kerr, data and modeling are just a small part of what we need to fight the pandemic: “While a tool like Covasim can help a policymaker decide what course of action will be best, without equitable access to vaccines, tests, and treatments, even the best policy won't be enough to ensure good health outcomes for their community. I believe this pandemic has underscored the need for everyone to have access to the tools they need to address these global challenges.”