'Superspreader' Sites Drive Spread Of Coronavirus, Model Finds

EVANSTON, IL — A computer model developed by a team of researchers from Northwestern and Stanford universities accurately predicted the spread of the coronavirus based on demographics, according to a study published Tuesday in the journal Nature.

Using cell phone location records, demographic data and epidemiological estimates, the model found a large majority of infections could be attributed to a small minority of "superspreader" locations where people come into close contact with each other for sustained periods.

The study also found residents of lower-income neighborhoods face a significantly greater risk of COVID-19 infection — going to the grocery store in a poorer community is twice as dangerous as doing so in a higher-income community, according to the model.

“We built a computer model to analyze how people of different demographic backgrounds, and from different neighborhoods, visit different types of places that are more or less crowded. Based on all of this, we could predict the likelihood of new infections occurring at any given place or time,” Stanford computer scientist and lead researcher Jure Leskovec said in a statement announcing the study.

Researchers used anonymized location information from 98 million cell phone users and socioeconomic data from 57,000 census blocks, and mapped 553,000 public places — gyms, grocery stores, restaurants, religious institutions and more — across 10 of the largest metropolitan areas in the United States, including Chicago.

Using records of new COVID-19 cases between March 8 and May 9, the model was calibrated for each city based on its own mobility and infection factors. In Chicago, it found 10 percent of the locations examined accounted for 85 percent of infections.

Researchers simulated different scenarios involving closing some venues and opening others, or placing occupancy caps on public places. For instance, the model predicted reopening all restaurants at full capacity would lead to the largest increase in new infections — due to the large number of locations, their density and the amount of time people spend in them — followed by fitness centers, cafes, snack bars, hotels and motels.

The model looked at the impact of capacity limits of 20 percent, 50 percent or 100 percent, It found that keeping occupancy below 20 percent would cut down on new infections by more than 80 percent — while reducing the total number of customers by only 40 percent.

---

Researchers did not include congregate settings such as schools, nursing homes and detention facilities among the public places they studied.

Jaline Gerardin, a professor of epidemiology at Northwestern's schools of medicine and engineering, said the model shows higher infection rates among minority groups can be explained by differences in movement patterns alone.

“The really cool part of this work is that it isolates the role of mobility in affecting the trajectory of the SARS-CoV-2 epidemic,” Gerardin said. "By starting with a very simple epi model, imposing the actual pattern of human mobility on it, and pressing 'play,' we were able to see that neighborhoods with lower income and a higher percentage of non-white residents end up with more infections, which is what happened in real life."

The study's authors say reopening plans that cap the number of people inside businesses are likely to benefit lower-income people the most.

For instance, the average grocery store in poorer neighborhoods has 59 percent more shoppers per square foot, and patrons remain inside for an average of 17 percent longer.

“This study demonstrates that the way low-income neighborhoods are constructed, with smaller establishments that serve more customers, is one of the drivers of racial and economic inequality in infections,” said Beth Redbird, a Northwestern sociology professor. “It also shows that, by reducing density in these locations, we might reduce this disparity.”

The peer-reviewed article, "Mobility Network Models of COVID-19 Explain Inequities and Inform Reopening," and its underlying data are available online, along with an interactive simulation of the model.

"In principle," Leskovec said, "anyone can use this model to understand the consequences of different stay-at-home and business closure policy decisions."