The Center for Data Innovation spoke with Olivier Telle, an urban health geographer at the French National Centre for Scientific Research. Telle discussed how he is using mobility data from Facebook’s Data for Good program to better understand and mitigate disease diffusion within and between cities.
Hodan Omaar: What is the goal of your work, and how are you using Facebook data to achieve this goal?
Olivier Telle: Our work has two ends. The first is epidemiological: We are trying to understand the way mobility impacts disease diffusion. The second and more important is socio-geographical: We are trying to understand how the socio-spatial dynamics of cities impact epidemic diffusion. Consider dengue, for example, a disease that is transmitted by the Aedes mosquito which is particularly well adapted to urbanized areas. The incidence of dengue has grown dramatically around the world in recent decades, with reported cases increasing more than eight-fold in the last 20 years. But biological processes alone cannot account for these numbers, especially in hard-hit countries such as India. Rather, a range of factors play a role, particularly increasing urbanization and human mobility. This means we need to re-evaluate the geographical models we use to understand public health problems, as currently no one knows how, or even where, to counter these diseases.
The first step in doing this is to better understand the commuting patterns of individuals. Historically, much epidemiological work trying to model disease diffusion has been done at the international level, because that is the only level of data that has been available. For example, in trying to model the diffusion of COVID-19 between Paris and Delhi, researchers could easily use data provided by flight operators. But this does not represent the reality of mobility, which is much more intense at the granular level. To overcome this barrier, I have been working with Facebook’s Data for Good program to use real-time, granular mobility data to model disease diffusion of ongoing pathologies that have been with us for hundreds of centuries, such as dengue, and new diseases that have emerged more recently, such as COVID-19.
The specific Facebook datasets we are using are called origin-destination matrices. These data represent movement at the city and suburb level, aggregated over thousands of individuals. We compare the patterns in these datasets against incidences of disease. One interesting difference between dengue and COVID-19 is that with dengue, we are interested in studying how mobility impacts diffusion, but with COVID-19 we are also interested in the inverse: How disease diffusion is impacting the mobility of individuals. For example, we have an ongoing project aimed at defining if the lockdown has impacted the pandemic curve and the spatial diffusion of COVID-19.
Omaar: As cities grow larger, smarter, and more interconnected, the opportunity for diseases to spread also gets bigger and the responses of different cities become more interdependent. How can policymakers use insights from your project in their city planning efforts?
Telle: Because all parts of cities are now hyper-connected, emerging diseases such as dengue and COVID-19 affect all parts of a region, be it a central, peripheral, or rural zone. Yet, there exists inequality in disease management within large cities and in the regions between them.
Take dengue management in New Delhi. More than 35 hospitals in the city record the number of patients suffering from dengue. According to the data they collect, New Delhi is officially the most affected city in India, with cases seven times higher than in Mumbai or Chennai, which are (officially) hardly affected. Accordingly, disease management efforts are focused in New Delhi rather than in the surrounding regions. But this leaves out the millions of suburban dwellers who travel in and out of the city, including the more than seven million people who travel from Mumbai everyday.
Our research is trying to help policymakers understand that they cannot control disease using a fragmented city-level approach. By improving their understanding of how people—and consequently disease—move at the granular level, policymakers can better coordinate disease management at a higher level, to ensure interventions are efficiently deployed.
Omaar: While the risk to people from different socioeconomic backgrounds contracting diseases like dengue or COVID-19 might be the same, the vulnerability to impact from disease, be it economic impact or impact to health, disproportionately affects the poor. How does this factor into the geographical models you use to understand public health problems, and how does that translate to policymakers when they are deciding how best to allocate resources?
Telle: It is true that the risk of certain diseases cannot be predicted on the basis of socio-economic factors alone. Our research on dengue in India illustrated this, showing that the disease is as prevalent—and sometimes even more prevalent—in richer regions. But while dengue does not discriminate in who it affects, mobility data shows that dengue systemically emerges from the poorest and most deprived areas, creating stark parallels with the 1832 cholera outbreak in Paris. The lessons from that outbreak are instructive.
In the mid-19th century, Paris’ squalid housing and ancient public hygiene system, where people threw sewage into gutters running down the street, allowed cholera, a bacterial disease that spreads through contaminated water, to rip through the city at an alarming rate. More than 30,000 people died; the rich and poor alike. The bourgeoisie understood that to avoid cholera themselves, they needed to provide clean water, electricity, and a good living environment for the poor. The elite were not driven by philanthropy. Rather, they recognized the link between infectious disease and inclusive urban planning, leading the decision-makers of the day to implement the proper drainage systems and wider streets Parisians enjoy today.
Many regions in Delhi still do not have such infrastructure. In the poorest regions where dengue is rife, people do not have proper access to clean water which means they store water in-house, creating a suitable source for mosquitoes to breed. As a result, these regions have a lot more mosquitoes in winter compared to richer parts of Delhi, enabling the virus to survive in the most deprived areas, even when the climatic conditions are not suitable for their survival. If policymakers were to provide basic access to clean water and to invest in proper sewage systems, they might not only protect those in deprived regions from dengue, but they would make the city as a whole more resilient.
Omaar: Facebook’s mobility data does not include users under the age of 18 or users who have their location services turned off. How do you ensure you are not missing a huge part of the population or reinforcing inequalities in disease governance?
Telle: We validate the Facebook data we use against more traditional data sources, such as the census or through surveys, to understand the relationship between numbers of Facebook users and demographics of a particular area. For example, in one area of India, Samuel Benkimoun, a member of our team, noted an inverse relationship between the number of women and the number of Facebook users—the more women, the fewer Facebook users. We found the same inverse relationship in areas with large tribal populations. In this way, we are able to see the impact of the digital divide and understand the limitations of our study.
The reason I use the Facebook data however, is because it is completely anonymized. It only provides collective mobility data on trips that are longer than 6.2 km, and we cannot reverse engineer the dataset to identify the movements of specific individuals. Equally important, I use Facebook data because it is free. We do not have the money to buy access to such data, nor do we have the time or resources to collect such data ourselves.
Moving forward, I would like to see more companies make their mobility data freely available. Not only so that we can use the data directly, but also so that we have more data sources to cross-check against, in order to ensure we are reducing bias as much as possible.
Omaar: Looking forward, what do you see as the main challenge policymakers need to overcome in order to better contain the spread of infectious diseases?
Telle: While the emergence of disease is the result of complex interactions between mobility, accelerating climate change, and the emergence of new viruses like COVID-19, the resilience of territories lie in their ability to maintain a measure of spatial and social equity. Certainly, diverse management efforts inevitably lead to these viruses spreading locally, regionally, and internationally, even in areas that should be less-vulnerable. The main challenge for every country that faces these risks is to make infectious diseases that are currently invisible more visible. We need to encourage the understanding of virus distribution models in territories by better integrating an essential component of the infectious risk: Collective mobility.