Sgt. Kyle Plumley, an intelligence analyst for Joint Force Headquarters, works three laptop computers at Camp Atterbury, Ind. (Staff Sgt. Chad Menegay)
Army analysts win award for helping commanders mitigate COVID-19
FORT BELVOIR, Va. -- A group of Army analysts were recognized last month for their efforts fighting COVID-19 that included helping top military leaders understand its severity, forecasting infection rates and advising commanders on where to deploy troops.
“[We] helped the Army understand how to keep Soldiers safe, support a national response to COVID-19 and maintain a ready force in the name of national security,” said Maj. Scott Lynch, an operations research analyst for the Center for Army Analysis, or CAA.
At the start of the pandemic, COVID-19 presented a unique challenge for the Army: few of its organizations were able to streamline multiple sources of data into large, actionable information that could be understood by senior leaders.
To overcome this, senior leaders called on a team of analysts, who worked behind the scenes at the CAA on Fort Belvoir, Virginia. They would go on to be called the COVID-19 Analysis Study Team.
Last month, the study team received the Dr. Wilbur B. Payne Award for the best analysis done by the Army in 2020.
The award has been given since 1981 and recognizes excellence in Army analysis. Originally called the Department of the Army Systems Analysis Award, it was renamed after the first deputy undersecretary of the Army for operations research in 1990.
Starting from scratch
The first step in helping commanders was to figure out what they needed to respond to COVID-19. In the beginning, everything was unprecedented, Lynch said.
It boiled down to the most available research being produced by non-military organizations that was not designed for Army operations, which left it lacking the information necessary to ensure readiness, said Robert Ward, a CAA operations research analyst.
The study team examined questions posed by commanders and created tools, estimations, reports and briefings to assist senior leaders in understanding, interacting with and making decisions.
Knowing the mission
In late 2019 the novel coronavirus first emerged and spread quickly across the world. By February 2020, the number of confirmed COVID-19 cases in the United States started to increase rapidly, with state- and national-level concerns about risk mitigation growing just as fast, Lynch said.
To stem the spread of the virus, Army senior leaders were tasked with a dual mission: protect Soldiers while preparing to aid civilian agencies.
It was a balancing act of stemming the spread of the virus, both in and out of the force that conflicted with maintaining Army readiness, then the Army’s top priority, Ward said.
“It was very chaotic, nobody knew what questions to ask or what information they needed,” said Maj. Dusty Turner, a CAA operations research analyst. “They just needed something to help them make sense of this sort of unprecedented thing that was happening.”
Before military leaders could make any hard choices about readiness and COVID-19, they needed a comprehensive analysis of trends to provide them better predictions.
“The team looked into how many cases of the virus would occur and how many hospitals would be impacted by the day,” Lynch said. “By grabbing the raw data from the public domain [and] crunching the numbers together, out the other end came predictions for how the virus would spread in the following weeks.”
Diving into research
One of the study team's core analytical functions was to develop automated techniques for gathering, consolidating, analyzing and visualizing public health data.
After that, it analyzed the data against COVID-19 cases to develop a forecast model to make an agent-based stochastic simulation. The model was based on a well-known tool in public health, known as the SEIR or susceptible, exposed, infected, removed model.”
The model relies on static, deterministic parameters, except for contact rates, and uses that data with a gradient-boosted, machine-learning algorithm to forecast the reproduction rate for each county and country, according to a COVID-19 Data Analysis Project report by CAA officials.
The gradient-boosted model was then trained with 38 variables, such as national mobility, seasonality, geographic attributes, state policies, COVID-19 results and population demographics, among others.
In addition to producing valuable results on its own, the model was also used to inform other analytical tools.
As the study progressed, the needs of senior leaders continually changed with it, forcing the team to regularly improvise unique modeling solutions in order to satisfy the demand to solve problems, Lynch said.
Some changes included analytical tasks, like creating heat maps for the U.S. Army Geospatial Center, where they estimated the timeframe county hospitals had before an influx of COVID-19 hospitalizations would hit, with over 3,000 distinct geographical regions accounted for.
“We realized a huge proportion of the country, especially geographically in mostly rural places, were already out of ICU capacity. There’s a lot of places the counties that had almost zero ICU beds,” Ward said. “A lot of their rural areas do not have large hospitals with intensive care units.”
In addition, the forecast model also predicted cases for 53 countries in the U.S. Indo-Pacific Command’s area of responsibility.
Against the clock
During the time the initial COVID-19 analysis was being conducted, the global crisis was rapidly getting worse. This reduced the time available the study team had to plan its internal organization, knowledge management, tools, and project management.
A typical Army operations research project begins with a clearly defined goal and due date. The plan is then devised to deliver the project on the due date. In contrast, their plan required continuous interaction with senior leaders and the rapid delivery of minimum viable products.
The study team was able to adapt to the constantly changing needs of senior Army leaders, many of whom were unfamiliar with the kind of analytical support they needed or what CAA could do for them, Ward said.
Uses and examples
The SEIR model was initially helpful to U.S. Northern Command and its subordinate units, like the U.S. Army Corps of Engineers, and the Federal Emergency Management Agency, when deciding where to support areas experiencing hospital stress due to COVID-19.
Also, U.S. Army North was responsible for the Army's support of civil authorities' efforts including testing for COVID-19 and building field hospitals in pandemic hot spots.
Overseas, the SEIR projection model attracted the attention of leaders in the Indo-Pacific, who used it to forecast COVID-19 cases in each of the countries within USINDOPACOM's area of responsibility.
Early on, the study team also became involved with the Army COVID-19 Campaign Plan, which organizes periodic briefings for the top leaders in the Army and is headed by the Army’s vice chief of staff and undersecretary of the Army.
During these briefs, the study team regularly presented their findings and provided input into trends, areas of concern and the overall context of how COVID-19 may evolve. Following the briefings with the study team, the vice chief directed the SEIR model to be the forecast model of record for the entire Army.
As a result of their initial COVID-19 work, the study team began answering questions from other commanders, in particular, the leaders of U.S. Army Central and U.S. Special Operations Command.
By altering parameters or initial conditions, the study team has applied the SEIR model to specific resource management questions posed by these organizations.
For example, some commanders have been interested in reducing quarantine times for newly-arrived troops in deployed environments in a safe manner. Depending on the country of origin and the method of arrival, Soldiers may undergo a lengthy quarantine that can account for almost 10% of their deployment.
With the SEIR model, the study team was able to assess the risk of bringing someone to an installation under various quarantine periods by using information about COVID-19 exposure and the likelihood of a positive test.
“This project brought together data, smart people and powerful computers,” Lynch said, which “can be a force for good, help respond to a crisis and help make more well-informed decisions for the Army.”
Among the recipients of the award were:
Lt. Col. Matthew Pacheco
Maj. Chad Chapman
Maj. Maxine Drake
Maj. Trey Harvey
Maj. Sandra Jackson
Maj. Scott Lynch
Maj. Dusty Turner
Thomas Brading, Army News Service