Introducing our weekly short-term forecast of COVID-19 cases

Faculty, staff and students,

Last Friday, we launched a new COVID-19 dashboard to share metrics such as positive cases, positivity rates, and Student Housing quarantine and isolation data.

Today I am writing to present another new tool: a short-term forecast that projects total positive cases and new positive cases within the KU community two weeks into the future. We will publish an updated forecast each Friday for the rest of the semester.

The goal of this 14-day rolling forecast is to give decisionmakers a sense for where the growth of COVID-19 cases is likely headed based on the most current information available. The forecast is less a tool for exact prediction on any given day and is instead a tool that helps us think about how case growth might unfold if we push trends out 14 days. Keep in mind that cumulative cases and forecasts are not a reflection of active infections.

Collectively, the dashboard and the two-week forecast will be very useful to the Pandemic Medical Advisory Team in its role as an advisory group to KU and county leaders, and to KU decisionmakers as we project campus needs for things like testing capacity and clinic capacity. Beyond that, these tools will also help our partners project community needs for things like hospital space, ICU beds, and ventilators across the region.

The two-week forecast has been developed in partnership with colleagues at The University of Kansas Health System and the Pandemic Medical Advisory Team. The data feeding the forecast and the dashboard is derived from a compilation of lab feeds and collection sites available through weeks of collaboration among Watkins Student Health, Lawrence-Douglas County Public Health, LMH Health, UKHS, and the CRL lab with strict privacy and security protections. Of course, there are testing data we may not capture because of where an individual chooses to test or how lab results are required to be reported to state and local health agencies.

The forecast was created using ensemble modeling, which is a process in which multiple models are used to predict an outcome. From a high level, we built four separate models. Next, we made predictions from each model. Finally, we averaged each model’s predictions to produce a median prediction, a minimum prediction and a maximum prediction, which you can see in the forecast’s graphs as the dashed red lines.

It’s important to recognize the limitations of any forecast of COVID-19 on a university campus. The reality is, there is no precedent for our current situation, meaning there is no pre-existing model or set of assumptions that neatly fit our circumstances. Like universities across the country, we are learning every day about transmission patterns and mitigation strategies, and we will continue to respond to new information accordingly.

For additional discussion of our dashboard and short-term forecast, I encourage you to watch our most recent Chancellor’s Weekly COVID-19 Update, as well as this past Wednesday’s edition of the Health System’s Morning Media Update, beginning at the 31:15 mark.

Thank you for your work

As we’ve said all along, we knew that reopening campus and conducting extensive testing would result in known positive cases. Our testing results so far remain in line with what we’ve expected and are prepared to manage. Our goal has never been to eradicate COVID-19; rather, our goal has been to take science-based steps that allow us to co-exist with it while fulfilling our mission of education, service and research.

Thank you for all you are doing to help us prioritize the health and safety of our community.



Douglas A. Girod