COVID-19 data model quantifies region-specific impact of social distancing orders

// Industrial & Systems Engineering

Photo of students wearing masks

Wearing face masks and maintaining physical distance from others during the global coronavirus (COVID-19) pandemic, undergraduate students Jane Gehringer (left) and Xi (Chelsea) Chen talk outside of Engineering Hall at the University of Wisconsin-Madison on Aug. 7, 2020. The photo was made as part of a UW-Madison Smart Restart marketing campaign and demonstrates health and safety protocols for campus. Photo by Bryce Richter/UW-Madison.

As the COVID-19 pandemic first took hold in regions across the United States in spring 2020, governors, mayors and local leaders hoping to quell the spread of the virus turned to the only actionable defenses available at the time: They closed schools and businesses, banned mass gatherings, issued stay-at-home orders and enforced other social distancing measures.

Photo of Oguzhan Alagoz
Oguzhan Alagoz

Now, a study published in the Annals of Internal Medicine by University of Wisconsin-Madison researchers quantifies the region-specific impact of social distancing measures on the COVID-19 caseload in three distinct areas: New York City, the Milwaukee metropolitan area and Dane County in Wisconsin.

Oguzhan Alagoz, Proctor and Gamble-Bascom Professor of industrial and systems engineering in the College of Engineering, and Nasia Safdar, professor of medicine in the School of Medicine and Public Health and UW Health medical director of infection control, led the research team. The group also included School of Medicine and Public Health researchers Ajay Sethi, professor of population health sciences; Brian Patterson, assistant professor of emergency medicine; and Matthew Churpek, associate professor of medicine.

Using aggregated cell phone mobility data as a way to track how people complied with social distancing policies, the researchers created a computational model to simulate COVID-19 cases based on when social distancing directives were implemented and eased, as well as how diligently people adhered to those orders.

The simulation shows social distancing measures wielded major influence on case numbers, though the impact varied markedly in different areas, even within the same state.

According to their model, the timing of implementing social distancing measures was particularly crucial in New York, where the state restricted mass gatherings March 12 and introduced increasingly stringent measures over the following 10 days.

However, according to the researchers’ modeling, had the state acted one week earlier, the number of cases in New York City would have been 80% less (41,366 instead of 203,261) by the end of May; conversely, a week’s delay would have increased the caseload nearly seven times, to more than 1.4 million.

The impact of the timing wasn’t as dramatic in Dane County, where a one-week delay would have increased its number of cases 36% by the end of July.

“Everybody knows, qualitatively, social distancing measures have made a difference, but I think this is one of the most accurate estimates of how much of a change they really led to,” says Alagoz, first author on the paper. “In places where you have high population density and a lot of movement in and out of the area, the impact of social distancing is significantly greater, compared to other places. Wisconsin, for example, implemented the same social distancing measures statewide, but the impact was different in Dane County, Milwaukee and other areas. Our model actually is able to tell us this quantitative estimate of how much of a difference we are going to see from one region to another.”

The model also takes into account each region’s demographics, infections imported from outside the area, asymptomatic transmission, age-specific adherence to social distancing rules, and limited availability of testing in the early months of the pandemic. The researchers note that a confluence of those factors drives infection rates in different areas, which demonstrates the need for region-specific modeling and policies.

“Context matters for prevention activities, and region-specific data and the expected effect of mitigation is invaluable to get buy-in and engage communities in mitigation efforts,” says Safdar.

Interestingly, the group’s model showed that the timing of lifting social distancing orders wasn’t nearly as consequential—provided the area’s residents largely maintained their vigilance (or, in the model’s terms, only allowed their adherence to social distancing to drop by 5%). If, for example, social distancing measures in New York City had lifted a week earlier—June 1 instead of June 8—the number of cases would have increased by 6,738 by the end of July.

“The fate of this pandemic is in our hands,” says Alagoz. “As a community, it depends on how we respond. It just depends on how people behave.”

Author: Tom Ziemer