Understanding CDC Data on SEIR Model: Interpreting Flu Outbreak Reports

[tom8]

I am doing a project about the SEIR model, which is an epidemic model consisting of 5 ODEs, that classifies people according to 5 categories:

S: Susceptible
E: Exposed
I: Infected
R: Recovered

I am looking into CDC's website and would like to take Flu as an example. My medical knowledge is limited so I am finding difficulty interpreting the reports. For example, here is a recent report.

If I understand correctly, the report provides data about (some of the) infected people only. There is no data about how many has recovered. Also, I am not sure how can find any data about what fraction of people are susceptible and what are exposed?

 

[Simon Bridge]

Oh good choice - you'll learn a lot about real-world limitations.

If I understand correctly, the report provides data about (some of the) infected people only.

Yes. Well ... mostly...

There is no data about how many has recovered.

No direct data - however, they do talk about the mortality ... i.e. how many people died.
What happens to those infected who do not die?

The equations would normally start out with some infected, nobody dead, nobody recovered ... if you jump into modelling a real outbreak after it has started (the usual situation) then you have to decide if you need to account for the people already recovered before you came in.

Also, I am not sure how can find any data about what fraction of people are susceptible and what are exposed?

You have to start out with the definitions.IRL cases, the susceptible and exposed figures are guesswork. You can improve your guesses by looking at studies in peer-reviewed journals. i.e.
http://www.ncbi.nlm.nih.gov/pubmed/18598231

Since flu is very contagious, and changeable anyway, you can probably guess a high susceptability for people who have not been vaccinated and the statistical rate published for people who have been vaccinated... then use published vaccination rates. Exposure is much harder since it depends on things like if someone stays in bed or goes to work (in flu - people usually treat the symptoms and go to work). Again - you'd probably have to guess. What seems reasonable?

Usually models are used to assess different scenarios rather than to track an actual epidemic.

There are a lot of technical terms in the report you cited - you should look up all the terms: they often have very precise definitions.

 

[tom8]

Thanks for your reply. I agree that I have to read more about this, especially given my lack of knowledge in this health-related field.

 

[Simon Bridge]

Curiously, this sort of study is sometimes done using "Zombie Outbreak" as the disease. This is simpler since everyone has easy access to the information that is needed.
i.e. the disease is spread by bites, so, physical contact. It is not contagious until symptoms show - by the time they show, it is too late to intervene. Stuff like that that be obtained from popular movies and TV shows so they just pick one.
Most of the analyses use an SIR model ... basically the E parameter is implicit.

The hard part in math modelling is figuring out how the real life statistics fit into the model. A good model uses real life collectable data as inputs, but this is usually not sufficient so there is always some guesswork and extrapolation. You main problem is understanding what the reports actually mean - which is hard because you run up against in-field jargon and CDC shorthand.

You could google for "SIER model influenza" and see how other people have handled the same thing.