illustration of social networks

Abby Rudolph is an infectious disease epidemiologist in the Department of Epidemiology and Biostatistics. Her research examines how social relationships influence risk behaviors and health service use. In July, she was named the 2020 recipient of the Freeman Award for her work in the field. The International Network for Social Network Analysis (INSNA) grants the award “for significant contributions to the scientific study of social structure,” to scholars who are under 40 years old or who have received their PhD within the past 10 years. 

Rudolph’s examinations of social networks have included studies of HIV/HCV risk behaviors and the role of peer networks in opioid overdose prevention and injection cessation. We asked her how social network analysis may play a role in understanding the COVID-19 pandemic and the strategies used to contain the virus.

How are the recommendations to social distance in order to "flatten the curve" with COVID-19 based on social network theory?

When you think about how an infectious disease is transmitted in humans, it is based on the types of interactions that can transmit a particular disease. With HIV, we are concerned mostly with identifying and limiting high-risk sex and injecting relationships. With a respiratory infection like COVID-19, we are most concerned about limiting relationships where we are in close proximity to someone who is infected and we can breathe in infectious respiratory droplets or aerosols, or situations where we may touch an infected surface and then inoculate ourselves by touching our mouth, nose, or eyes.  

Given that SARS-CoV-2, the virus that causes COVID-19, is a new virus, everyone in the population was initially susceptible. The basic reproduction number for an infection, R0, is defined as the average number of secondary cases produced by a single infection, in a population where no one is immune. R0 is influenced by three factors. First is the probability that when a susceptible person interacts with an infected person, the susceptible person will become infected. Frequent hand washing and wearing a mask can reduce the probability of infection per contact between a susceptible and infected person, just like using condoms can reduce HIV transmission. Second is the average rate of contact between susceptible and infected individuals. Physical distancing reduces the rate of contact. The third factor is the average duration of infectiousness. Quarantine and isolation guidelines that are based on duration of infectiousness can also reduce the rate of contact between infectious and susceptible persons.

Contact tracing is another important strategy for containing the spread that’s rooted in social networks. Because people are contagious before they develop symptoms, contact tracers aim to identify people who may have been exposed to the virus and notify them to recommend quarantine. You’re trying to disrupt the transmission chain. 

How does network analysis complement the molecular epidemiology that scientists are performing to analyze particular strains of the virus? 

Over time, the virus mutates and we can look at the similarities in the mutations to construct something like a family tree of the virus. This can be used to determine its origin. In fact, this has confirmed that it was not created in a lab. The initial cases in Washington state were from China, but the virus circulating in NYC was much more similar to the virus circulating in Italy. By looking at air travel patterns, we can see that international flights facilitated the global spread of the virus beyond where it originated. When you overlay different types of networks (i.e., viral sequences similarities, flight routes and contact tracing), you get a more complete picture of how the virus spreads.

What role will social network analysis have once a vaccine for COVID-19 is developed?

The goal of a vaccine is to create "herd immunity," or a situation in which enough of the population is immune to the disease that the rest of the population which is not immune to the disease is protected. For example, if 90 percent of the population is immune to a virus, this means that if an infected person comes into close contact with 10 other people, only one person will become infected. At the start of the epidemic, everyone was susceptible, so all 10 could become infected. By reducing the number of people who can become infected, we can dramatically reduce the spread of an infectious disease so that it can more easily be contained. When an infectious disease is more contagious, we need to vaccinate more people to create herd immunity. We can also use information about how people interact with one another to determine ways to prioritize vaccination—to protect those who are more susceptible to infection due to their inability to physically distance from others or greater exposure to the virus via their occupation (e.g., healthcare professionals or factory workers). Specific models that predict disease transmission rely on assumptions about how members of the population interact with one another. These are often very simple assumptions, but if we borrow ideas from the field of social networks we can create models that more accurately reflect how people interact with others and our predictions will likely be more accurate. These models can be used to prioritize vaccination.

Many people hear "social networks" and think of Facebook or Twitter. Does social network analysis say anything about the role platforms like these may play in the COVID-19 situation?

Network analyses involving social media can examine the spread of information or misinformation (like conspiracy theories) through networks and define social norms regarding mask wearing and social distancing. People may post pictures to demonstrate compliance or lack of compliance, which then promote that norm within their networks. Online social networks often polarize attitudes and viewpoints because people tend to be friends with, follow, like, share or re-tweet people and messages that they agree with. Over time, people tend to see more messages that affirm/validate/confirm their own beliefs and attitudes than those that oppose them. This makes people perceive that their viewpoints are normative (even if they are only normative among a small group of densely connected individuals). For example, two people having two different perspectives may each perceive their views to be mainstream due to the different content that each interacts with.