A social network analysis of the spread of COVID‑19
https://www.nature.com/articles/s41598-021-87837-0
In the early stages of the pandemic, many countries initialized a digital contact tracing system to monitor the spread of the Covid virus, using information technology. Particularly, South Korea’s Center for Disease Control and Prevention employed a digital contact tracing system, pulling data from mobile phone carriers, immigration services, police, credit card companies, and public transit companies to create a comprehensive route/map/network of covid infected cases using location data, immigration records, closed-circuit television footage, credit, and debit card transactions, and public transit records. The location history of infected patients and their routes were made publicly accessible.
This paper uses the data of 3283 confirmed patients in Seoul city from January 20, 2020, to July 19, 2020, to recreate the Covid-19 infection network from real-world data and examine the infection network and its structural characteristics. From January to July, the South Korean government had three different phases in its policy and response to the pandemic: (i) early stage (01/20/2020–03/21/2020); (ii) social distancing stage (03/22/2020–05/05/2020); and (iii) distancing in daily life stage (05/06/2020–7/19/2020). For each stage, the number of nodes and the mean distance of the network is calculated.
The mean distance between each node noticeably decreased during the social distancing period, suggesting that social distancing was effective in decreasing the spread of the virus.
The study also went on to do something more interesting, where they removed nodes with the most out-degrees by each threshold, creating iterations of network models. Jo states that there were 17 values between the maximum and minimum out-degree values, thus performing 17 virtual node eliminations. What this paper found out with the elimination of top nodes is that with a high out-degree threshold, the number of nodes eliminated may be relatively small. However, from the collective network, the number of uninfected nodes from the network was large. As illustrated in Figure 4, the larger out-degree threshold was more efficient in reducing the number of infected nodes.
I found it interesting that analyzing the network models can help identify what influences the structure of the network so that authorities can better respond to a pandemic. The study demonstrated that instantaneous contact tracing and quarantining the infected are very effective in reducing the total number of infected as the effect of elimination of top out-degree nodes on the whole network shows. By comparing three phases of South Korean government policies, it was apparent that Covid measures set by the government have a meaningful impact on the infection network structure. Social distancing showed to reduce the mean distance between nodes, reducing the size of the infected network. Authorities can also learn from the study that with positively skewed distribution in high out-degree nodes (what we call a superspreader) it would be more efficient and feasible to allocate resources to screen and trace individuals with high infection potential and quickly quarantine them rather than trying to moderate the whole population’s infection.