New Paper (Other): Potential roles of social distancing in mitigating the spread of coronavirus disease 2019 (COVID-19) in South Korea

[adamlmaclean]

Title: Potential roles of social distancing in mitigating the spread of coronavirus disease 2019 (COVID-19) in South Korea

General Information

Link: https://github.com/parksw3/Korea-analysis/blob/master/v1/korea.pdf Data & code link: https://github.com/parksw3/Korea-analysis

What is the paper's Manubot-style citation?

Citation: tag:Park2020_distancing

Is this paper primarily relevant to Background or Pathogenesis?

• [x] Background
• [x] Pathogenesis
• [ ] Methods

Please list some keywords (3-10) that help identify the relevance of this paper to COVID-19

• population dynamics
• transmission
• social distancing
• contact tracing
• reproductive number
• time-dependent reproductive number

• [ ] virology
• [x] epidemiology
• [x] biostatistics
• [ ] immunology
• [ ] pharmacology
• [ ] other:

Summary

If you would like to submit a summary of the paper, please copy and paste the following into a comment.

Suggested questions to answer about each paper:

• What did they analyze?

  • Analyzed epidemiological data (daily cases) from January 20 - March 16, 2020, and transport data (Metro travel) in South Korea between 2017 - 2020.
  • Studied the differences in epidemiological dynamics between Daegu (central to the SK outbreak) and Seoul.
  • Inferred the time-dependent reproductive number (R_t) in each location.

• What methods did they use?

  • Inferred time-dependent backward onset-to-confirmation delay distributions from daily case data using negative binomial regression.
  • Used weakly informative priors for fixed effects ~ Normal(0,2)
  • Incubation period modeled by a Gamma dist, generation time by Negative Binomial

• Does this paper study COVID-19, SARS-CoV-2, or a related disease and/or virus?

  • Epidemiology of COVID-19

• What is the main finding (or a few main takeaways)?

  • social distancing and contact tracing in Daegu were effective (at least in part) at reducing the number of new cases
  • Suggests that the COVID-19 epidemic can be suppressed with less extreme measures than those taken by China
  • R_t in Daegu is now well below 1, however in Seoul is still above 1, this could lead to secondary outbreaks
  • Region-specific analysis is important: "recent decrease in the number of reported cases in South Korea is driven by the sharp decrease in Daegu. Our analysis reveals that the epidemic may still persist in other regions, including Seoul and Gyeonggi-do"

• What does this paper tell us about the background and/or diagnostics/therapeutics for COVID-19 / SARS-CoV-2?

  • Tells us what the epidemiological dynamics of the disease are early, and lllater during an outbreak when social distancing measures are in place
  • Highlights the importance of focusing on region specific dynamics

• Do you have any concerns about methodology or the interpretation of these results beyond this analysis?

  • Choice of priors on the incubation period and generation time, is there support for these?

Any comments or notes?

Good supplemental data to #69 on dynamics of infection.

[adamlmaclean]

Hi @agitter how should this paper be cited? preprint on github; no doi afaik. Thanks!

[agitter]

When this gets added to the manuscript, we can use url:https://github.com/parksw3/Korea-analysis/blob/master/v1/korea.pdf as the Manubot citation. Tag me in the pull request so that I can follow up after it is merged. I'll edit manual-references.json so that the citation is correct. I tested Manubot with that PDF URL, and the metadata isn't extracted correctly.

I'll probably create a citation tag as well. That way, once this is posted somewhere formal and receives a DOI we can easily switch over.

[agitter]

@alavendelm I created the manual reference in #120. Once that is merged, you can cite it as [@tag:Park2020_distancing].

[hufengling]

@alavendelm, regarding your last point: "Use of priors on the incubation period and generation time, is there support for these?"

This paper uses a Bayesian framework for their modeling, which allows for the mathematical integration of current data and "prior beliefs" (previous evidence) to generate a posterior distribution of the parameters they were looking to estimate (in this case, reconstructed incidence and R_t.)

In this paper, the authors use a hierarchical Bayesian framework - essentially, this provides mathematical language to say, "I have previous knowledge, but there's actually uncertainty about that previous knowledge too." This framework gives the model structure and flexibility, while allowing the data to speak. In my understanding, this is fairly common practice.

[adamlmaclean]

@hufengling, I was referring to the choice of priors used, not sure these are supported by evidence from the literature. Not the use of priors itself (I edited above to make this clear). I have no concerns about Bayesian modeling framework -- this is standard best practice.