VICKSBURG, Miss. – When the U.S. Army Engineer Research and Development Center’s Dr. Brandon Lafferty sleeps these days, he dreams about predictive models.
That’s because since mid-March, Lafferty, a researcher from the ERDC Environmental Laboratory, has been helping lead ERDC’s Modelling and Simulation Team develop the ERDC Susceptible Exposed Infected Recovered ⸺ or SEIR ⸺ model for COVID-19, and it’s an intense effort.
“When I go to bed, I think to myself, ‘What can I think about now so I don’t dream about models all night?’” Lafferty said.
The ERDC-SEIR predicts the disease’s spread and provides the foundation for the ERDC approach. The model predictions are provided to the U.S. Army Corps of Engineers Geospatial Taskforce Group, which then summarizes outputs in order for the broader USACE team to advise federal, state and local partners on decisions related to COVID-19 courses of action ⸺ for example, where and when to construct alternate care facilities (ACF).
The SEIR model is just one of several COVID-19 modelling efforts, and Lafferty is handling the day-to-day operations of the team, supporting EL Senior Research Scientist Dr. Todd Bridges, who is at the forefront of the ERDC modelling efforts. The Information Technology Laboratory’s Dr. Larry Lynch is the ERDC COVID-19 Fusion Cell Lead, the overarching coordinator for all ERDC’s COVID-19-related activities.
Bridges described how the team of ERDC scientists and engineers has developed, operationalized and delivered a suite of capability that provides critical information and support to our national response and recovery. “I am amazed by the work this impressive, committed team of 40 has done over the last several weeks,” he said. “It’s a tangible demonstration of the Power of ERDC ⸺ and there’s more to come.”
Another extensive modelling effort, led by the EL’s Dr. Benjamin Trump, has translated the ERDC-SEIR model and population-specific data in Guam, the Commonwealth of the Northern Mariana Islands and American Samoa in support of the USACE Pacific Ocean Division and the U.S. Army Pacific Command, the Army component of the U.S. Indo-Pacific Command. “Our modeling team has been working to support decision making by multiple U.S. states and territories,” said Dr. Igor Linkov, another EL researcher who leads the Risk and Decision Science Team that Trump and Jeff Cegan are part of. “POD, FEMA, INDOPACOM and others are interested in utilizing ERDC’s approach in other territories, regions and countries.”
ERDC-SEIR predictions have also been used by an effort led by Cegan to support the Federal Emergency Management Agency Region 1. Cegan supplements the ERDC-SEIR model with information relevant to hospital resource needs to help FEMA request the resources necessary for response and recovery in the New England states.
The ERDC-SEIR Model
The ERDC-SEIR predicts how the disease will spread across the continental U.S., and the team is getting ready to expand its reach to areas outside the continental U.S., such as Europe.
“It’s the most comprehensive modelling platform we’re working on,” Lafferty said. “It’s being used by USACE as a planning tool for building ACF, and it provides estimates as to how many infected patients states and counties will have.
“The model takes the reported number of active cases in a county’s or state’s geographic area, then uses a parameter optimization process, similar to the way models for weather forecasting are parameterized, and that data is fed into ordinary differential equations to provide predictions for active COVID-19 infections for the future,” Lafferty said.
EL’s Drs. Michael Mayo, Michael Rowland and Todd Swannack; ITL’s Drs. Glover George, Jerry Ballard, Ian Dettwiller, Timothy Dunaway and Brandon Randle; Dr. Gary Brown from the Coastal and Hydraulics Laboratory; and Dr. Matthew Parno from the Cold Regions Research and Engineering Laboratory are all participating in this team effort to develop and run this mathematical model.
EL’s Molly Reif is the liaison officer with the USACE Geospatial Taskforce, providing the ERDC model predictions for the USACE Common Operating Picture, or uCOP. “The USACE Geospatial Taskforce is responding to USACE Fusion Cell requests and incorporating multiple models into the uCOP in an effort to assist decision makers on coronavirus spread and ‘peak day’ projections,” Lafferty said.
Bridges said the goal of the modeling effort is to provide the most accurate and timely projections possible and to ground the modeling solidly on the best available data. “There are over 40 researchers involved in the ERDC-SEIR modeling effort and a system of ongoing checks and balances to ensure that we are providing the best forecasts possible, for what is a very complicated, dynamic, and challenging problem,” he said.
In Alexandria, Virginia, Col. David Hibner, commander of the Army Geospatial Center, and Gary Blohm, director of the Army Geospatial Center and the Geospatial Research Laboratory, are leading the Geospatial Taskforce, which supports the USACE mission, Emergency Support Function-3, Public Works and Engineering. “Under ESF-3, the USACE is called upon by FEMA to coordinate and support certain tasks as part of the federal response to an incident,” Lafferty said. “One of these tasks is to provide modeling and predictive support to help make predictions about what buildings to build where.”
Lafferty described how one major question they’re trying to answer with the modelling is where and when the peaks in the outbreak will occur, so states and their partners can make decisions about ACFs. “For example, in New York, they may have needs now for extra beds, but in one month, the needs will be lower because the peak will be over,” he said. “But if capacity is built in, for example, Alabama, where the crisis is just starting, more may be able to be served for a longer period of time.”
Lafferty said the team is now working on another part of the project. “We just started to model the second wave, and we have to assume a fraction of people will have immunity after the first wave,” he said. “We don’t know when it will hit, but it will be sometime in the future, depending on the timeline for relaxing social distancing and other factors.”
Guam and the Pacific
The researchers doing the modeling at ERDC all agree that the effort has been very challenging. EL’s Dr. Benjamin Trump, who has led the effort to model for Guam, the Commonwealth of the Northern Mariana Islands and American Samoa, describes how he has been working 16-18 hours a day, averaging five to six hours of sleep a night.
“At one time, I used to read one book a day,” he said. “And all the books about how to succeed in a high-pressure environment say the same thing: you have to get that much sleep at a minimum to keep functioning.”
Armed with a doctorate in health management and policy from the University of Michigan, Dr. Trump happens to have a background ideally suited to this project, including coursework in epidemiology.
In 2019, he participated in a Johns Hopkins University program, “Emerging Leaders in Biosecurity Fellowship.” The program participants had three training sessions, including one short-term stay in September at the World Health Organization in Geneva, Switzerland, where they engaged in a highly interactive pandemic simulation. “The way the simulation played out was oddly similar to the current situation ⸺ this exercise was the perfect preparation for COVID-19,” Trump said. “We also learned that decisions in these situations, however small, can have a huge impact on people’s health.”
On Friday, March 27, the ERDC Team received word that the Pacific Ocean Division had an urgent need to model, by Monday morning, what COVID-19 would do to Guam.
Over that weekend, Trump and others from the ERDC Team applied a three-step process for assessing regionally adjusted COVID-19 outcomes. The first step was to develop a number of mathematical models to assess the curve, the number of infections and deaths per day. “This is classic epidemiology; this is a very old field ⸺ there are simple models and more complex ones,” he said.
Once this picture was developed for Guam, the team had to ascertain what the population health dynamics of the area were and other sociological factors that could dampen the spread or make it worse.
The third step of the process was to look at institutional abilities and infrastructural resources available for the local coronavirus outbreak. “Determining how many beds, staff, ventilators and so forth are on Guam isn’t enough to complete the assessment,” Trump said. “That would be a fatal assumption, because not all beds are available for COVID-19.”
Trump and his team asked what the hospital bed utilization rate in Guam was, and found out it was quite high, 50-75 percent. “Given the COVID-19 outbreak that Guam was experiencing, we determined Guam could experience a devastating curve, but the curve can be manipulated by more stringent social distancing and other government actions,” he said. “Once we completed the modelling for Guam, we did the same for CNMI and American Samoa.”
Trump, along with Linkov and Cegan, are also on an interagency detail to FEMA’s Region 1 Data Analytics Team; this region encompasses the area of the U.S. known as New England: Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island and Vermont.
USACE is working and coordinating with FEMA all over the nation, as part of the multi-agency federal, state and local effort to retrofit existing facilities as ACF. The ERDC-SEIR model influences this effort by informing estimates based on need. However, due to the closely integrated relationship between FEMA Region 1 and the USACE New England District, this group’s modeling effort is influencing a range of decision-making in that particular region.
“Cegan is leading efforts to help New England with assessing the number of patients, and then projecting everything from hospital beds to mortuary equipment needs ⸺ everything that goes along with a crisis of this magnitude,” Linkov said. “One thing FEMA Region 1 wanted to know was how many additional ventilators they would need to request from the national stockpile.”
Linkov describes how since models are uncertain, they use a multi-model approach for short term projections, which includes polynomial statistical and exponential distribution models. For long-term projects, the ERDC-SEIR is leveraged to better predict the disease propagation dynamics.
“There are challenges on a daily basis. The main one being that decision makers have very specific questions and needs, and the models are highly uncertain, because we often have fragmented information,” Linkov said. “Our model may result in a life-or-death decision, so that, along with the pressure to respond to emergent needs on a daily basis, makes it very challenging.”
The effort for FEMA Region 1 is still ongoing. “We provide daily updates so they have information as to what they can expect in the next few days,” Linkov said.
ERDC Director Dr. David Pittman recently described how he felt about the teams’ work. “I’m so proud of all of our modelling teams’ efforts,” he said. “The models’ effectiveness demonstrate how the Power of ERDC can help our Nation solve even unprecedented challenges, like the COVID-19 pandemic. Our modelers have brought tremendous resources to bear on this fight for our federal, state and local partners, and their collaborative efforts will go down in history as an outstanding example of interagency cooperation.”
Linkov feels the effort has been very worthwhile. “I feel great about our efforts, because somebody needs to do this work, and I’m excited we’re helping decision-makers make these real-life, important decisions,” he said.
Trump agrees. “This is something I’ve been waiting to do my entire life; at an emotional level, this modelling effort for COVID-19 has been the most meaningful work that I have ever done,” he concluded.