UND Today

University of North Dakota’s Official News Source

Mapping a monster

UND researchers find innovative ways to project and track pandemic’s path

A partial screenshot of Joseph Mbuh’s map-based project, which tracks COVID-19 cases in each of the country’s thousands of counties. This is the map’s website.

If you have followed the global spread of the coronavirus, you are perhaps familiar with John Hopkins University’s map that chronicles the toll of COVID-19 around the world. As useful as that map is, if you live in North Dakota, you might find it lacking in granular local data. And, here is where several efforts at the University of North Dakota to document the virus in our state come in.

Working with Dr. Ryan Adams, director of the School of Electrical Engineering & Computer Science, Sandeep Singhal, assistant professor in pathology and bioinformatics, has built a model to forecast the trajectory of the coronavirus in North Dakota. Senior Associate Dean of the UND School of Medicine & Health Sciences Mark Basson and Pathology Professor Donald Sens have also contributed to the development of the predictive model.

In the Department of Geography, geospatial instructor Joseph Mbuh has put together an interactive, context-rich map that documents coronavirus cases in every county and every state in the U.S. Graduate geography student Anai Caparo, on the other hand, has singled out North Dakota to track the shifts in the virus.

Into the future

While maps allow us to follow the coronavirus in near real time, federal and local officials who manage COVID-19 response actions are also looking into projections about how the virus might behave in the future.

For that endeavor, data is key, said Singhal. Applying his expertise in artificial intelligence and machine learning, Singhal played an instrumental role in creating a model based on publicly available data to prognosticate the pandemic’s life cycle and theoretical ending date in North Dakota.

Continuously updated with the latest data, the forecast is expected to change in real-world scenarios over time. In that regard, the UND model largely differs from conventional predictions that assume future accuracy solely based on present data.

Parsing data from a variety of sources, including the state’s Department of Health, the Centers for Disease Control and John Hopkins, Singhal said the model is currently indicating that the average number of coronavirus cases will significantly drop in the coming weeks. That decline is conditioned on the continued maintenance of social distancing rules, face mask use, hand washing and other preventive measures.

Courtesy of Sandeep Singhal

“We are using this model to predict how many patients we will have in the coming month,” Singhal said. “Together with that, we can infer the hospitalization rate and how many ICU beds we need. The model indicates that North Dakota has a sufficient number of ICU beds and hospital beds.”

The analysis also charts a comparison between North Dakota and the rest of the USA population.

Courtesy of Sandeep Singhal

Adams said that once North Dakota reaches the peak of COVID-19 infections, the number of positive diagnoses will remain high for some time before the virus’ curve begins to smooths out.

Sandeep Singhal

“What the flattening of the curve means is that we have fewer people at one time, but it does also mean that the number of people will remain elevated for a period of time,” he said. “We’re looking at the next week or so, and we’re expecting that the curve will no longer be exponentially growing, but that it’ll flatten out and begin to reduce slowly over the next couple of months.”

To arrive at these projections, the model takes into account factors such as population density, residents’ susceptibility to COVID-19 as well as their age. Some of those factors work to benefit North Dakota, which is a rural state of less than one million citizens, who live in cities that are much more sparsely populated than metros such as Seattle and New York City, which are seeing some of the nation’s largest coronavirus outbreaks.

“When we look at the age and other similar demographics, we’re beginning to see that North Dakota is a little bit more unique,” said Adams. “It’s similar in terms of median age and in population density to neighboring states like South Dakota, Montana and Nebraska, but it’s very, very different than states like New York or Oregon or Washington.”

While such differences underline disparate virus trajectories and response approaches on the federal level, local county disparities reveal diverse rates of infections and hospital treatments in North Dakota.
For instance, while Grand Forks County is a populous region in the state, its hospitalizations might be lower because the median age here is lower compared to other less populated counties, Adams said. The coronavirus usually affects older people and those with underlying conditions more severely than younger people.

Ryan Adams

Adams and Singhal are cooperating with the epidemiology group within the Department of Health, who have met with Doctor Anthony Fauci, who leads the National Institute of Allergy and Infectious Diseases and serves on the administration’s coronavirus taskforce, and Doctor Deborah Birx, response coordinator for the White House. At that meeting, Dr. Birx outlined coronavirus predictions for North Dakota that aligned well with Adams’ model.

Adams said the model can help local decision-makers tackle crucial considerations such as when and how to reopen the local economy.

Meanwhile, with Incoming President Andy Armacost making the connection, a collaboration to share data is currently taking shape between UND and the Massachusetts Institute of Technology, MIT, where scientists are working on their own COVID-19 predicative model.

County by county

“With the initial outbreak of COVID-19, there was a staggering amount of geospatial information that emerged, and several dashboards and near real-time services were built for extraordinarily transparent response efforts,” said Mbuh. “But what we did realize was that most of these dashboards until about a month ago were on a global scale. When the cases started to surface domestically in the United States, I thought it would be a good idea to make a map that shows only the United States, not a global picture.”

Joseph Mbuh

Still, most maps out there relay the total number of COVID-19 cases as well as the tallies of those hospitalized, recovered and deceased (the raw data of the coronavirus spread, so to speak). To Mbuh, that is not enough; it is not a rigorous, exhaustive representation of the real effects of the novel virus.

“You need to present raw data in comparison to the population of each county to help explain the number of cases in that county,” Mbuh said. “You will notice that counties with higher population density would normally lead to a higher number of cases.

“If you are representing only raw data on a map, that is misleading and could very well lead to residents of certain counties to panic even more than they are already.”

With input from the College of Arts & Sciences Dean Bradley Rundquist, Mbuh’s project presents the number of positive COVID-19 cases per every 100,000 residents in every county in the country. (He is also plotting daily changes in reported cases across geographic regions and states.)  In that way, Mbuh “normalizes” (and updates daily) the data on the coronavirus, in order to remove sensationalism and inform official responses to the respiratory disease.

“What this does is it gives us the opportunity to know where the problem is so that we can target solutions quickly and efficiently,” Mbuh said.

To further support actions, Mbuh has developed infographics for each county that succinctly and visually outline metrics such as age of the population, school enrollment, poverty levels and healthcare insurance coverage. Such data is culled from a variety of sources, including the 2014-18 American Community Survey and information applications developed by Esri, a geographic software and solutions company.

“This tells us where to deploy resources but, just as important, it tells us where not to waste resources,” Mbuh said.

For example, Mbuh’s project can guide the placement of testing sites in every state or the distribution of ventilators to regional hospitals.

Day by day

While Mbuh’s project encompasses the daily shifts of the virus in every county in every state in the U.S., Caparo’s map zooms in on North Dakota, tracking total positive COVID-19 cases, recoveries and deaths by day in each of the state’s 53 counties. Unlike Mbuh, she is not standardizing the coronavirus tallies against counties’ populaces. Instead, Caparo is documenting the absolute day-to-day progression of the virus, which she finds to be spreading faster in some counties than others.

Caparo’s map, which can be found on its own website.

For instance, “Cass County now has more cases than Ward County, but Ward registered the first case in North Dakota,” said Caparo, the graduate geography student. “How did this happen?”

By chronicling such peculiarities, using (and manually logging) data from the North Dakota Department of Health, Caparo hopes that, in the future, her project will help inform the study of the coronavirus and the course it is currently charting around the country.

“For me, it is interesting that the coronavirus didn’t first emerge in one of the bigger cities in the state such as Bismarck or Fargo or Grand Forks,” Caparo said.

In a way, Caparo’s map supplements the Department of Health’s public dashboards. For example, the latter keeps track of daily total active and new cases in the state without dividing them by county, as Caparo does. The Department also monitors positive cases, among other things, in each county without separating them by day, as Caparo does.