The Data for Good team at Facebook shares privacy-preserving datasets that researchers and public health officials have been using to combat COVID-19. With three billion people using our services worldwide, aggregated Facebook data offers a unique window into epidemiological factors like population movement, population density, and social connections.
Informing physical distancing policies in the United Kingdom
Ealing, London. Photo by Edward Howell on Unsplash.
Despite increased transmissibility of the virus in the United Kingdom, data has shown that the January 2021 lockdown had one third less impact on movement than the March 2020 shutdown. Some regions were found to be moving at more than 50% of pre-pandemic levels. Population movement is analyzed via the Oxford COVID-19 Impact Monitor, which uses aggregated and anonymized mobile phone location data to help tackle the pandemic. Facebook’s High Resolution Population Density Map is used for validation purposes, providing the United Kingdom population density map base layer for the COVID-19 Impact Monitor. Dr. Matthias Qian comments, “Our mobility measures show signs of lockdown fatigue among Britons. Despite the overflow of hospitals with COVID-19 patients, the behavioural change and mobility patterns responded less than during the March lockdown.” Read more here.
International risk estimates from new COVID-19 UK variant
Empty airport in Germany during the COVID-19 pandemic. Photo by Markus Spiske on Unsplash.
A fast-spreading SARS-CoV-2 variant identified in the United Kingdom in December 2020 has raised international alarm. Researchers collected data on the variant from nineteen countries, and estimated the probability that this variant was introduced by travelers from the UK. Using Facebook mobility data, they examined the extent to which the virus was being locally transmitted, based on the changing proportion of the new variant among infections identified in the UK and population mobility from the UK to each country. These projections suggest that countries with substantial population movement from the UK were likely to harbor cases of the new variant by late October 2020. Read more here.
Improving COVID-19 Forecasts in Germany
Bielefeld, Germany. Photo by bennett tobias on Unsplash.
Researchers at Ludwig Maximilian Universitat and the German Research Center for Environmental Health developed a model to forecast weekly COVID-19 cases at the local level in Germany. To quantify social and mobility patterns, they leveraged three Facebook datasets: data on friendship ties via the Social Connectedness Index, colocation probabilities via Disease Prevention Maps, and the proportion of people staying put via Movement Range Maps. Using COVID-19 case data, they validated the model on the current week of study, and forecast the upcoming week. Results indicated that their forecasting approach had notable improvements over all other existing approaches for forecasting COVID-19. Read more here.
Analyzing COVID-19 Spread in France
Paris, France. Photo by Fran Boloni on Unsplash.
Researchers at Ecole Polytechnique in France and Athens University of Economics and Business in Greece developed a model to study the impact of population movement on the spread of COVID-19. To support their model, they used Facebook’s Disease Prevention Maps to look at human mobility via administrative regions. Focusing on four countries—Italy, Spain, France and England—they were able to demonstrate the superiority of their approach in forecasting the spread of the disease compared to all other models. This research was supported by the French National Research Agency and by the European Union. The model offers useful insights to policymakers on appropriate interventions and resource allocation. Read more here.
For more examples of how Facebook data has been leveraged during the COVID-19 pandemic, click here.
Facebook Data for Good partners with over 450 organizations across nearly 70 countries in the world. Darker shade of blue indicates a higher number of partners.