Cookie Settings
close

By clicking "Accept", you agree to the storage of cookies on your device to improve site navigation, analyze site usage and assist with our marketing efforts. See our privacy policy for more information.

Using Data to Create Better Reopening Plans

Replica is a data platform for the built environment.

July 11, 2022
Join a Future Summit


Replica
is a data platform for the built environment. We deliver recent insights into how people live and work by providing a collective representation of the built environment — people, mobility, economic activity, and land use — so you can understand the relationships and trade-offs behind every decision you make.

We use machine learning technology to turn billions of de-identified data points into insights, providing a quick and accessible way to combine traditional data about cities (like census data) with new sources of information (like smartphones and payment data). We combine these various data sources to surface patterns and trends about how groups of people move and interact with the built environment.

Beyond the classic urban planning applications, we can show representative interactions between synthetic users at different locations. This allows us to quantify how many of these interactions happen at work or at a shopping center or while eating out.

To protect public health, Replica can model which businesses or other locations (e.g., schools) could gradually reopen while limiting people’s interactions. Recently, through the Reopen Mapping Project, Replica co-authored a research paper with a team of Harvard, UC Berkeley, Stanford and University of Chicago researchers from a range of fields within economics and transportation science evaluating:

which mix of policies could help a local jurisdiction safely reopen their businesses or schools while balancing public health concerns

Consider the City of Chicago, along with its surrounding geography. We simulated the movements of over 7 million people in one day pre-pandemic. These synthetic people visit almost half a million points of interest, encompassing shops, restaurants, grocery stores, pharmacies, workplaces, schools. We also simulate household composition.

Compared with other studies, our work does not consider interactions solely at public places but also the importance of interactions at work, schools, and at home, and takes into account the demographics of synthetic users. Our analysis relies on assessing potential contact between any two synthetic users if they are simulated to be present at any location in a short time window or live together. We built graphs of interactions between any member of the synthetic population, with over hundreds of millions of links, to better understand which locations and activities are contributing to disease spread.

Table 1. Contact by “Point of Interest” Category in Chicago

Source: Replica

Figure 1: Contact Matrix by Age in Chicago

Source: Replica

Contact risk varies by demographics.

It is well known that a given number of individual contacts and age can factor greatly into the severity of the disease. In Figure 1 above, we depict Chicago’s resident population and estimate the volume of contacts a typical Chicago resident would have given their age in comparison with other age groups. As you can see, a typical Chicago resident in their 70s is likely to meet 43% as many people as the average person in their 20s (114 compared with 248). Yet seniors are more likely to develop severe forms of the disease, risking hospitalization and even death.

In addition to age, public officials should also consider exposure by industry and type of employment. Replica’s synthetic population matches American Communities Survey data to ensure it is representative. Based on a synthetic person’s industry of employment and their age, we evaluated how many contacts a person typically has in a given day. For instance, someone working in the healthcare industry (NAICS 62) tends to interact with many more people than someone working in manufacturing (NAICS 31–33). Therefore, when reopening workplaces, considering contact risk by industry classification could be useful to prioritize which industry to open first in order to support the twin goals of economic recovery and disease containment.

Once the graph of contacts between any two users in a geographical region is understood, we can then simulate epidemic spread and the consequences of various policies on public health, unemployment, and the expected strain on the healthcare system, leveraging a detailed variant of the classical SIR (Susceptible Infected Recovered) disease model calibrated using parameters from the Covid-19 spread characteristics.

While reopening, jurisdictions must acknowledge associated trade-offs amongst a range of different policies. A policy is one that may restrict certain types of movement, thereby limiting the spread of the disease.

Source: Reopen Mapping Project

Replica, as part of the Reopen Mapping Project, simulated six types of policies on our contact graphs:

  • No Policy (NP): All contacts are allowed, pre-pandemic level.
  • Shutdown (EO): Only essential businesses remain open. Schools, workplaces and most businesses are closed.
  • Cautious Reopening (CR): Workplaces and schools are open, but neighborhood interactions are reduced to 10% of pre-pandemic levels (reflecting enforced social distancing measures at most points of interest and recommendations to use caution).
  • Alternate Schedule (AS): Students and workers in all schools and workplaces are split into two groups that do not interact. Neighborhood interactions are reduced to 10% of pre-pandemic levels.
  • Work from home when possible (WFH): Individuals who can work from home do, non-work-from-home types go back to work, schools are open. Neighborhood interactions are reduced to 10% of pre-pandemic levels.
  • Isolate 60+ (60+): Individuals 60+ must limit their contacts to their household. Neighborhood interactions are reduced to 10% of pre-pandemic levels.

We considered each policy in three steps: (1) before the pandemic, there was no policy; (2) then a shutdown was enacted for 75 days; and (3) then one of the policies discussed above is put in place.

For instance, when the EO (shutdown) policy is imposed, only slightly more than 60% of individuals in Chicago are actively employed — either as workers in essential industries or by being able to work from home. This number decreases to 55% as individuals contract the disease and either get sick, die, or become quarantined and unable to work from home.

The outcome of each simulated policy is evaluated along three dimensions:

  • The health outcomes for the population are measured in terms of death and infected rates. A disease spread model from the epidemiology literature, calibrated to the existing data on the Covid-19 pandemic, is simulated on the graph model created from the Replica synthetic model of population movements, including all activities of any synthetic resident in a given day.
  • The economic impact is evaluated with unemployment statistics. By industry and age, we estimate who can work from home. For a particular policy, a different set of residents may be affected.
  • The healthcare burden is estimated in terms of ICU admissions and healthcare costs, by leveraging theCovid-19 Research Database, and the health outcomes obtained from the epidemic model.

Table 2. Comparative Policy Outcomes (-%/+% is reduced/increased volume) in Chicago

Source: Replica

From Table 2 above, “no policy” means a complete reopening after a 75-day shutdown, which leads to the worst-case scenario in terms of loss of life and healthcare burden, which also has costly implications. Continuing the shutdown for a prolonged period of time strongly protects the health of the population and reduces loss of life but causes a severe jump in employment loss. In between these two extremes, a range of options exist. For example, policies promoting people to work from home or policies whereby school and work schedules are alternated (splitting classes and employees in groups) seem to be very effective in limiting contact risk and contagion spread while also containing employment losses.

Key Findings

From Replica’s data-driven insights, we found that:

People’s movements are key to understanding viral spread. To reduce contact risk, policymakers should take into account people’s movements as they relate to different points of interest and consider which ones pose a higher risk of spreading Covid-19.

Public officials need to consider contact risk by industry classification. This can also be useful in prioritizing which industry to open first in order to support both economic recovery and disease containment.

Developing policies in their local context can be more effective, taking into account:

  • Age/industry of synthetic population’s employment.
  • Economic impact of closing each POI varies on the mix of businesses supporting the local economy.
  • Policies promoting residents to work from home, or policies whereby school and work schedules are alternated (splitting classes and employees in groups) seem to be very effective in limiting contact risk and contagion spread while also containing job losses.

For more information on Replica, please contact us.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

Learn More?

Contact us to learn how Replica can help bring insights like these to your organization.
Thank you! Someone from the Replica Team will be in touch soon.
Oops! Something went wrong while submitting the form.

Trusted by: