Background

With limited clinical data, no known treatment, and a vaccine likely a year away, the only tool that public health officials have right now in the fight against COVID-19 is social distancing. However, to allocate limited resources to those in greatest need and protect the most vulnerable populations, policy makers must be able to understand the impact and effectiveness of social distancing measures in near-real time.

In collaboration with epidemiologists within the COVID-19 Mobility Data Network including the Harvard School of Public Health, Direct Relief, Princeton, and many others, we built this dashboard as an enhanced offering to public health officials that builds further on others’ early work product. This effort provides a more accurate and actionable understanding of the effectiveness of social distancing and other policy interventions aimed at reducing or slowing the spread of COVID-19.

Methodology

In developing this dashboard, we began with several goals. First, we worked with public-health researchers to understand what they needed and provide them with the most-important data and metrics. Second, we worked with experts to ensure that we used privacy-forward practices in developing these metrics. Finally, we aim to iterate based on new information and feedback from the public and researchers to continue to aid the fight against COVID-19.

In achieving these goals, we started by applying our experience at Camber and the experience of the epidemiological teams, and taking into account the differences between aggregated human movements that are predictable and movements that are not, notably socialization. Presenting both radius of gyration and entropy gives a more complete picture of socialization patterns within a county. For example, a high RoG and low entropy could indicate a population that needs to travel far to go to work or the grocery store, but are otherwise staying home; a low RoG and high entropy could indicate a more dense area where people are staying near home but are still moving in their neighborhood. Both are important signals needing different interventions.*

Further, we decided to merge multiple data sources together, so we can reduce errors and be more representative of the larger population, while yet further increasing privacy. Although we have the same limitations other dashboards have, we believe that by refining, cleaning, anonymizing and carefully standardizing this data, public-health researchers are better able to evaluate the similarity and differences between the results across the multiple datasets.

With these broad data sets and tailored metrics, researchers are able to better understand how population mobility data can inform social-distancing interventions in response to COVID-19. Building on the work done by others, this dashboard is an effort to contribute to the invaluable work being done to support public health researchers in some of the hardest questions and decisions of our lives. The dashboard itself, to protect privacy and other civil liberties, are aggregated at the county. For researchers, we provide data at a lower level, but still anonymized and aggregated, so that deeper conclusions can inform public decision makers.

Other mobility data dashboards

There are other mobility data dashboards created by partners in this effort that have informed our work, for example, the Unacast and the New York Times mobility-data dashboards. Taken as a whole, public dashboards can help improve decision-maker's and the public’s understanding of social-distancing patterns and trends over time.

Data limitation and bias statement

The data populating this dashboard is from smartphones - that, by its very nature, limits who is included in the overall data set. Additionally, we have combined data sets to create as representative of a population as possible, but there will always be more work to be done to improve that. We compare our data with census information to normalize as best as possible. By acknowledging these flaws up front, we do our best to put this information into context. Furthermore, because historic and current racial disparities are reflected in where people live and work, it is important to consider movement in that context and not use this information to make decisions that would further exacerbate those disparities.