The mystery of SARS-CoV-2 transmission
Although significant progress has been made in understanding the specifics of the SARS-CoV-2 virus behind the Covid-19 pandemic – its genetics, protein structure and medical effects on infected individuals – much less is known about how it actually spreads from human to human in different environments. That explains why super-spreading events have attracted so much attention: Why does the same virus in some cases not infect direct family members, while in some cases a single human carrier can infect scores of individuals?
For instance, an infected individual from the city of Yongin in Korea was found to have visited at least five nightclubs in Seoul’s entertainment district of Itaewon on the night of May 1-2. He subsequently headed to other districts in the capital and neighboring provinces before testing positive for Covid-19 four days later. Alarmingly, contact tracing for this super-spreader connected him to a staggering 1,300 individuals. This notorious example clearly demonstrates the amplifying effect that some places might have in spreading disease.
Super-spreading events, however, should not be linked just to particular individuals with specific traits or behaviors, but must also be connected to local characteristics directly related to patterns of human movement, physical proximity and interaction. Contrast, for example, transient contact among many shoppers in a hypermarket with a few nursing home residents and staff in close, sustained interaction. These patterns are remarkably complex, especially in bustling, dynamic cities. Nevertheless, there are many components that can be employed to build this model to a reasonable level of applicability.
The multiple data streams we generate already provide valuable insights into the movement of people, objects and therefore, disease. Tracking human mobility patterns is especially critical because people transit in and through cities, potentially carrying and spreading disease to geographically disparate areas. As early as December 31, 2019, the Canadian health-monitoring platform BlueDot notified its customers about a new coronavirus outbreak. The company’s proprietary disease surveillance program applies natural language processing and machine learning techniques to analyze news reports in 65 languages, together with airline data and reports of animal disease outbreaks. Its access to global airline ticketing data helped forecast travel patterns of infected individuals and its algorithm correctly predicted that the then-unnamed virus would propagate to Bangkok, Seoul, Taipei and Tokyo after its initial appearance in Wuhan.
Besides air travel information, ground-transport data have become significantly richer with the growing prevalence of ride-sharing services that collate detailed records of every journey. Such statistics can complement public-transport trip data already being tracked by bus, subway and train operators. Cities such as Paris, Singapore and Taipei have made some of their human-mobility data publicly available, but more granular information will facilitate constructing human-flow networks with the required spatiotemporal resolution – tracking human flows down to precise locations and to the hour or even minute. Similarly, smartphone apps for scheduling, navigation, payment, health and communication offer detailed and diverse information, presenting a composite picture of people’s daily activities and interactions.
Integration is key to understanding data
Integrating such human-mobility and activity data with known factors of Covid-19 transmission can help model disease spread in a metropolitan area and help us pinpoint super-spreader locations and vulnerable groups. As more granular data is collected, the model should also afford greater precision over time. This endeavor requires a coordinated, large-scale multidisciplinary effort in three largely distinct domains: