- A new open-source dataset matches county-level COVID-19 case statistics with county-level data on non-pharmaceutical interventions (NPI; e.g., stay-at-home orders) in 1,320 U.S. counties
- The dataset covers closures of nonessential workplaces, stay-at-home orders, enforcement of stay-at-home orders, limitations on public gatherings, school closures, public transit closures and publicly available testing
- Across the country, there was significant variability in NPI implementation among counties, both within and across states (P<0.001)
- NPIs for any given county were largely uncorrelated, with the exception of a strong positive correlation between workplace closures and stay-at-home orders (r=0.835)
- Illustrating the utility of county-level NPI implementation and analysis in the U.S., an exploratory analysis showed that closures of nonessential workplaces were associated with reduced COVID-19 incidence (P=0.004)
Countries around the world have prevented or contained COVID-19 outbreaks through non-pharmaceutical interventions (NPIs) such as issuing stay-at-home orders. In the U.S. there has been little coordination of national, state and local responses, resulting in wide variation in NPI policies both geographically and temporally.
Jagpreet Chhatwal, PhD, associate director of the Institute for Technology Assessment at Mass General, and colleagues have created an open-source, county-level dataset that permits research into how county-level NPI policies can affect COVID-19 rates. In the Journal of Medical Internet Research, they report their methodology and the results of an exploratory analysis.
In an effort led by Cray V. Noah at Hikma Health, 104 volunteers, mostly graduate students, were recruited and remotely trained to collect NPI data and case rates from 1,320 U.S. counties in a standardized way. The period of interest was March 1 to July 20, 2020. The NPIs included were:
- Nonessential workplace closure
- Shelter-in-place/stay-at-home orders
- Enforcement of shelter-in-place/stay-at-home orders
- Size restrictions on public gatherings
- School closures
- Public transit closures
- Publicly available testing
The effort began with the 500 most populous U.S. counties; the other counties were selected in real-time based on high incidence rates. The researchers included timestamps for when each NPI was put in place and lifted at a county level, coupled with county-level COVID-19 case statistics from The New York Times.
Variation in Policies
Across the country, there was significant variability in NPI implementation among counties, both within and across states (P<0.001).
Correlation Between Policies
NPIs in any given county were largely uncorrelated, with the exception of a strong positive correlation between workplace closures and stay-at-home orders (r=0.835).
Workplace Closures and Case Growth
Workplace closures were associated with reduced COVID-19 incidence (P=0.004). For example, in California, case growth was minimal during the period of workplace closures, and in both urban and rural counties, the end of workplace closures was followed two to six weeks later by an increase in cases.
Toward Better Policymaking
Previous national and international studies have also found positive correlations between implementing workplace closures and lower future rates of COVID-19. This growing body of literature suggests that NPI policies help limit the effects of the COVID-19 pandemic.
Because COVID-19 behaves very differently in different counties, additional local analyses of NPIs are urgently needed. Suggestions for future research include studying relationships between NPIs and COVID-19 cases, deaths, health care utilization and social inequities.
View all COVID-19 updates
Learn more about COVID-19 research at Mass General