A study of how 98 million Americans move each day explains that most infections occur in “high-prevalence” areas, and how mobility helps increase the rate of infection among minorities and low-income populations.
A team of researchers has created a computer model that accurately predicts the spread COVID-19 by analyzing three risk factors for infection in 10 major cities this spring: where people go in a day, how long they stay, and how many people go to the same place at the same time.
“We created a computer model to analyze how people from different demographic backgrounds and different neighborhoods visit different places that are more or less crowded. “Based on all this, we were able to predict the possibility of new infections anywhere or at any time.” Stanford computer scientist Jure Leskovec led the researchers’ efforts. Northwest University.
The study was published in the journal on November 10, 2020 Nature, combines demographic data, epidemiological estimates, and anonymous cell phone location data, and confirms that most COVID-19 transmissions occur on “very common” sites, such as full-service restaurants, fitness centers, and cafes, where people have been nearby for a long time. . The researchers said the specificity of their models could be a tool for authorities to help minimize the spread of COVID-19 by reopening businesses by detecting conflicts between new infections and lost sales. .
The study’s co-author, David Grusky, a professor of sociology at Stanford School of Humanities and Sciences, said the ability to predict is especially valuable as it provides useful new insights into the factors behind disproportionate infection rates in low-income and low-income people. “In the past, these discrepancies were thought to have been established through pre-existing conditions and unequal access to health care, whereas our model of mobility helps manage these non-relative risks,” he said.
Grusky, who also runs the Stanford Center for Poverty and Inequality, points out that the re-opening of the model’s businesses with fewer living conditions tends to benefit disadvantaged groups the most. “Because places that employ minorities and low-income people are often smaller and more crowded, reopening stores can reduce the risks they face,” Grusky said. “We have a responsibility to create reopening plans that eliminate or at least reduce the differences created by existing practices.”
Leskovec said the model’s stay-at-home policy this spring has reduced the number of trips outside the home and slowed the rate of new infections, saying the model “still provides the strongest evidence.”
The study found that 98 million Americans in the country’s 10 largest metropolitan areas went through half a million different businesses, from restaurants and fitness centers to pet stores and new car dealerships.
The group included Stanford doctoral students Serina Chang, Pang Wei Koh and Emma Pierson, who graduated this summer, and Northwestern University researchers Jaline Gerardin and Beth Redbird, who collected job data for 10 metropolitan areas. In terms of population, these cities include: New York, Los Angeles, Chicago, Dallas, Washington, DC, Houston, Atlanta, Miami, Philadelphia and San Francisco.
SafeGraph, a company that collects anonymous location information on mobile apps, provided researchers with data showing which of the 553,000 hardware stores and religious sites they visit each day; how much; What is the square footage of each facility, and most importantly, so that researchers can determine the hourly fill density.
The researchers analyzed March 8 – May 9 in two different stages. In the first stage, the model fed mobility data and designed their systems to calculate a crucial epidemiological variable: the rate of virus transmission under different conditions in 10 metropolitan areas. In real life, it is impossible to know in advance when and where a contagious and susceptible person is in contact to potentially create a new infection. However, researchers have developed and refined a number of equations in their models to calculate the probability of infectious events occurring at different locations and times. The equations were able to solve the unknown variables because the researchers fed the computer to one, an important known fact: how many COVID-19 infections were reported to health workers each day in each city.
The researchers refined the model until they could determine the rate of transmission of the virus in each city. The degree varied from city to city, depending on how often people left home and where they went.
After obtaining transmission rates for 10 metropolitan areas, the researchers tested the model in a second phase, asking to increase the ratio for each city against mobility sample bases to predict new COVID-19 infections. The predictions were closely monitored by actual reports from health workers, giving researchers confidence in the reliability of the model.
Prediction of infections
By combining their models with demographics available in a database of 57,000 census groups – neighborhoods of 600 to 3,000 people – the researchers show how minorities and low-income people leave home more often because of the demand for jobs and shop in smaller, larger stores. Businesses that have higher incomes, can work from home, use home delivery to avoid shopping, and sponsor larger businesses when they go out. For example, research has shown that non-whites are about twice as likely to buy groceries as whites. “By combining mobility, demographic and epidemiological data sets, we were able to use our model to analyze the effectiveness and equivalence of different opening policies,” he said.
The team made their tools and information available to the public so that other researchers could replicate their findings.
“In principle, anyone can use this model to understand the consequences of stay-at-home and business closure policy decisions,” Leskovec said, adding that the group is now working to make the model a user-friendly tool for politicians and public health. officials.
Reference: “COVID-19 mobility network models explain inequalities and announce reopening” Serina Chang, Emma Pierson, Pang Wei Koh, Jaline Gerardin, Beth Redbird, David Grusky and Jure Leskovec, November 10, 2020, Nature.
DOI: 10.1038 / s41586-020-2923-3
Jure Leskovec is an associate professor of computer science at Stanford Engineering and a member of the Stanford Bio-X and Wu Tsai Institute of Neuroscience. David Grusky is a professor at the School of Humanities and Sciences, Edward Ames Edmonds, and a senior fellow at the Stanford Institute for Economic Policy Research (SIEPR).
This research was supported by the National Science Foundation, the Stanford Information Science Initiative, the Wu Tsai Institute of Neuroscience, and Chan Zuckerberg Biohub.