Math May Distinguish Flukes from Flu Epidemic Outbreaks


Home / Math May Distinguish Flukes from Flu Epidemic Outbreaks

One sick person can transmit a disease far and wide.  Image by SCA Svenska Cellulosa Aktiebolaget

Implications for Public Health in Recognizing a Flu Epidemic

This paper has two obvious implications for public health and clinical research organizations.

  • First, when routine sentinel surveillance can reliably determine that an outbreak is highly unlikely to lead to an epidemic, a full outbreak investigation would not be initiated. Instead, resources would be directed to isolating the reservoir and protecting those most likely to become the head of a chain.
  • Second, when an outbreak investigation is required, this method can determine the threat of an influenza epidemic much more quickly and economically by estimating ‘R through determining ‘F’, rather than by exhaustively reviewing each patient’s source of infection./li>

Lead Author Dr. Simon Cauchemez Responds to Questions

Decoded Science had the opportunity to ask Dr. Simon Cauchemez, lead author of the research paper, some questions to clarify this report.

Decoded Science: Under routine sentinel surveillance, R=1-G and is a trustworthy estimate if G is large enough (close to 1). Would a smaller value of G trigger the outbreak investigation, or cast doubt on the accuracy of the estimate? If so, what value of G would be recommended to trigger further investigation?

Dr. Cauchemez: “If the sample size is large and G<=5%, I’d strongly suspect that R may be above 1.” (further investigation would be indicated in this situation.)

Decoded Science: Both ‘G‘ and ‘F‘ must be between zero and one, so the estimate for ‘R‘ cannot be greater than one. But can ‘R‘ actually be greater than one?

Dr. Cauchemez: “The reproduction number R in an outbreak can be larger than 1. Interpretation of statistics G and F works as follows:

  1. If 1-G<1, this gives evidence that R is smaller than 1 and we can estimate R as R=1-G;
  2. If 1-G~1, this means that R is above or larger than 1 but we can’t provide a precise estimate of R.

…Our method is designed to estimate R in the context of subcritical outbreaks, i.e., R<1.

Decoded Science: Is it worth determining a value for R>1 by the old-fashioned and labour intensive process in the outbreak investigation?

Dr. Cauchemez: “Indeed, we’re not claiming that this approach should replace outbreak investigations. It’s just adding another method to the toolbox of epidemiologists, allowing simple interpretation of standard surveillance data. It’s also filling a gap for situations where it’s very hard to track cases. Outbreak investigations remain needed to get a precise estimate of R when R>1 and there are many other reasons why we should keep on doing outbreak investigations.

Decoded Science: What would be a threshold value for F in an outbreak investigation that would trigger the more intense approach?

Dr. Cauchemez: “The main concern for public health is if R becomes larger than 1 so (F close to zero). But we’re worried that a virus adapts to increase its ability to transmit between humans. So, if we could document an increase in transmissibility (e.g. from R=0 to R=0.7) – that would be source of concern, even in R remains <1.


Cauchemez S, Epperson S, Biggerstaff M, Swerdlow D, Finelli L, et al. Using Routine Surveillance Data to Estimate the Epidemic Potential of Emerging Zoonoses: Application to the Emergence of US Swine Origin Influenza A H3N2v Virus. (2013). PLoS Med 10(3): e1001399. doi:10.1371/journal.pmed.1001399. Accessed March 5, 2013.

Leave a Comment