Comments on: Modeling a Pandemic like Ebola with the Wolfram Language

By: Richard

Richard — Wed, 29 Aug 2018 10:05:25 +0000

Could you please tell me where you received the real data from about the initial number of infected that equals 5% of the population?

By: Jos klaps

Jos klaps — Tue, 09 Jun 2015 20:36:26 +0000

Congratulations, outstanding report !

By: Jacky

Jacky — Tue, 10 Feb 2015 14:52:21 +0000

Mathematical modeling, through use of a powerful computer program, of the spread of an infectious disease can not only save a lot of human lives but can even prevent future epidemics and pandemics.

By: Vitaliy Kaurov

Vitaliy Kaurov — Mon, 09 Feb 2015 20:26:10 +0000

At the end of the blog there is a download link to all files including “airports.dat” and “route.dat”. In the blog text when the data are introduced there is a discussion and link to the data webpage. If you have any technical questions could you please post them on our related forum discussion: http://community.wolfram.com/groups/-/m/t/326240

By: Judiths

Judiths — Thu, 05 Feb 2015 08:50:39 +0000

Excellent! It’s a good model about ebola_virus spread to reference. There’re some problems for me to resolve. One of the most important problems is the data, such as the “airports.dat” and “route.dat”. Could I have a look the data_source about the “airports.dat”?

By: Josh

Josh — Thu, 27 Nov 2014 02:45:12 +0000

You listed North America twice when generating Vitality data.

By: Kari L.

Kari L. — Tue, 18 Nov 2014 23:00:26 +0000

Wow this is very informative to see it laid out like this! It’s important to have some of these graphics in order to see the full scope of the epidemic. I hope we continue to see more resources like this. I have also been following updates on the Ebola Virus from many resources such as: http://www.metrex.com/news/ebola-virus

Hope we can bring more resources and visibility to Ebola!

By: Marco Thiel

Marco Thiel — Mon, 17 Nov 2014 18:07:08 +0000

I agree that the SIR model is quite primitive. It has, however, been quite successfully used to model very homogeneous populations.The main model assumption is that if the people in each population behave in a similar way their social connectivity can be “parametrised” by the parameter for the infection rate. It is quite clear that this is not quite true globally, but I hope that the model suggests one way of taking this into consideration: by including a network of sub-populations which can all behave differently. By using country wide data, we still model at a rather coarse level, but it should be clear how to go forward from here. We could model the cities/regions in each country by individual SIR models and adapt the parameters so that they reflect the local behaviour. We could then go to even smaller sub-populations and so on. There are two main problems with that approach: (i) one is that we could get more and more equations and the required CPU time and memory would very soon require very large computers.(ii) the second is that for each additional layer of sub-populations we would dramatically increase the number of parameters. We could try to determine them from data, such as mobile phone data (http://www.technologyreview.com/news/530296/cell-phone-data-might-help-predict-ebolas-spread/) and other data sources, but more parameters do not necessarily improve the predictions of the model.

I think that the two points you raise at the beginning (that Ebola spreads very inefficiently in human populations and the importance of the health care systems) are not problematic for the model we present. The low infection rate can easily be modelled by choosing the corresponding parameter to be small. The importance of the health care system is actually used in the model. The problem was how to parametrise this… We used the life expectancy as a primitive proxy for the quality of the health care system. This is of course only a crude guess, but as you say it is very relevant for how the disease spreads and your feeling that this is important is reflected by our analysis.

I also totally agree that we do not model interventions (e.g. quarantine), and that certainly is very relevant. There is of course a relatively easy way to model this, which is to decrease the infection rate over time, as the countries take counter-measures. In countries that invest more into the fight of ebola the infection rates could be assumed to decrease faster. It only takes a line of code to take this into consideration, and I have simulations for that. (They were too detailed for this blog, but if you are interested I can post them on the Community.)

Finally, I think that all you say about the social connectivity, the low infection rates, interventions etc. can be modelled – at various degrees of granularity – if you tell me more about which assumptions you want to implement. If you have a better idea of how people move, e.g. you have telephone data, or use “Where is George” (http://www.wheresgeorge.com) we can implement your assumptions into the model, just as we did with the life expectancy and the population density. It is fun to add new processes and see how the model reacts. I hoped to introduce a model that is flexible enough for everyone to include their own ideas and test what impact they might have.

Also note that this is supposed to be a conceptual model. It certainly has limitations but it does show some reasonable features, e.g. France is at a higher risk of infection than many other European countries. This makes sense as there are many links (including flights) to former colonies in Africa. It is also reasonable that Europe gets infected before the United states and Asia… So the model does make some reasonable predictions.

I’d love to discuss that with you in the Community…

Cheers,
Marco

By: David Gurarie

David Gurarie — Thu, 13 Nov 2014 23:55:29 +0000

The use of geospatial data and the transport network is good, but the underlying SIR is way too primitive to take seriously Marko’s predictions.
The spread of Ebola in West Africa is first and foremost the result of inadequate (non-existent) primary healthcare system. Unlike respiratory (air-borne) infections (flu, SARS) ebola is highly inefficient in human spread (BRN close or less than 1).
Any serious attempt at modeling its spread should account for (i) social connectivity and behavior (ii) environment and interventions (e.g. quarantine). Population level SIR is a far cry.
Model of these type are more appropriate for simulating Ebola panic rather that Ebola infection.

By: Vitaliy Kaurov

Vitaliy Kaurov — Thu, 13 Nov 2014 21:09:59 +0000

Thank you! We’d love to hear more of your ideas. I think the best place for a technical discussion is our Wolfram Community – you can also attach files there if you wish to do so for your paper: http://wolfr.am/1BB1_ZVL