Epistemix Blog

Evaluating School Reopening Strategies

Written by Epistemix | Jul 9, 2020 10:01:33 PM

“We are living in a very challenging time as we plan for 8,000 people to return to our district (1,000 employees and 7,000 students) in seven short weeks. Data is critical to our success as we strive to balance health and safety with face-to-face interactions.”

— Darcie Mosley, Millcreek School District Superintendent

COVID-19 represents an unprecedented challenge for K-12 educators and administrators who must make crucial reopening decisions under conditions of significant uncertainty. As the pandemic rages across the country, educational leaders have to consider the impact of every decision on students, teachers, parents, and the larger community they serve.

Questions school districts face include: When and how to open safely for students, teachers, and staff? What are the impacts on parents, extended families, the local economy, and the larger community? How to effectively compare and contrast different reopening policies, and which policy is best one for their particular school district? How should policies differ across elementary, middle, and high schools? What is the best way to update policies as new data become available?

These dilemmas are further complicated by the fact that epidemiological risk factors vary enormously based on local conditions, limiting the efficacy of broad, standardized policy responses, and that districts have few avenues by which to learn from each other’s experience. With so many lives and livelihoods at stake, decision makers need better tools to inform their choices.

Epistemix is working to help fill that gap.

Our team has spent decades building epidemiological models to combat infectious diseases in partnership with the National Institutes of Health, Bill and Melinda Gates Foundation, Robert Wood Johnson Foundation, and others. Now, we’re working with schools, health care systems, Fortune 500s, nonprofits, researchers, and federal, state, and local governments to respond to COVID-19.

Using data from the US Census, Department of Education, American Community Survey, and Bureau of Labor Statistics, our synthetic population matches actual local population statistics and demographics, and, unlike many equation-based models that lack the flexibility and granularity to compare the implications of actual policy interventions below the county level, our agent-based model uses realistic social dynamics to simulate epidemiological outcomes from the bottom-up to show the policy impact neighborhood by neighborhood across the country.

That specificity is crucial for districts trying to evaluate the relative merits of alternative school reopening strategies, focusing on health outcomes for students, teachers and the community. Using epidemiological best practice, we can model the specific, local consequences of different reopening policies on students, families, staff and the community—including the number of local COVID-19 cases, hospitalizations, and deaths most likely to result from each strategy. Our simulations account for every individual in a particular region as well as the households, schools, workplaces, disease progression, social-behavioral data, and regulations specific to that geography.

The researchers behind Epistemix studied student-student interactions and contact patterns in Pennsylvania schools when developing the platform by having students wear trackers and record journals of who they were in contact with and where those interactions took place during the day. This is one of the only studies of this kind and ours is the only software that accounts for this level of detail when modeling different policy outcomes. This study enables the Epistemix team to account for differences in grade level interactions throughout the school day and in the classroom.

Every model we build is designed to inform a specific choice the district is facing. We begin by calibrating our COVID-19 disease model to the specific local outbreak pattern, taking into account local closure rules, social distancing behavior, school closure orders, and other local events. In the images below, you can see how we fit our model to the actual disease history of Allegheny County in Pennsylvania—note that in the Total Cases graph on the right, because reported cases are always a fraction of actual cases, our model is likely to be more accurate than the county’s figures.

Once the initial model is calibrated to local conditions, we run scenarios for the particular reopening strategies the district is considering, for example: opening schools as normal, moving everything online, or reducing time spent in school by 50% by breaking students into cohorts. The model then maps out the implications of these strategies under community conditions like the stage of the epidemic and the number of cases in the county, broad social distancing measures like stay at home orders and masks, and macro-dynamics revealed by population density and mobility scores. The model then generates the number of cases, hospitalizations, deaths, and healthcare resources resulting from each strategy under consideration.

In the charts below, you can see the implications if all the Allegheny County school districts reduced physical attendance by 50%—how many students and teachers would be infected in school and how many infections would lead to hospitalized cases under three levels of community conditions.

These outputs arm superintendents with the analytics they need to implement the right policies for their districts. At its best, modeling isn’t about prediction, but insight. It makes assumptions explicit, provides shared context for discussion, and informs high-stakes decisions. It reveals counterintuitive ways in which COVID-19 interacts with the complex adaptive systems that are our schools and communities, and the steps we can take to make things better. For more details, download a complete sample report here.

This data-driven approach to evaluating reopening strategies helps districts safeguard students, teachers, staff, families, and communities. In addition to modeling individual policy interventions, Epistemix is coordinating across districts to establish a network of superintendents and school leaders that can share best-practice in real-time as we continue to learn and adapt to the pandemic—tracking the efficacy of different approaches across similar districts.

So, how can we help your district?

For K-12 schools, Epistemix can model the impacts of different reopening policy options and provide grade-by-grade results for school district teams based on the district’s geography, number of students, population density, and reported cases in the county. For Epistemix to build and run a model for your district and community, district leaders will need to provide basic information including opening plans and contingencies, location of each school, number of students per grade per school, number of teachers and staff, and age ranges of teachers and staff (age is a major COVID-19 risk factor). Epistemix has been working with PA schools, hospital systems, and the state government on modeling various social distancing policies, business reopenings, social gatherings (e.g. high school football to Pittsburgh Pirates baseball games), and interactions between COVID-19 and other diseases like measles, flu, and pertussis.

Consult with us on your school’s reopening strategy.