Reader Comments

Post a new comment on this article

Complex Systems Group Journal Club

Posted by Complexity_Group on 01 Mar 2007 at 13:10 GMT

Complex Systems Group Journal Club

Date: February 22nd 2007

Participants: Gilles Daniel, Enrico Scalas, Joseph Wakeling

Paper discussed: The Waiting Time for Inter-Country Spread of Pandemic Influenza

Authors: Peter Caley, Niels G. Becker, David J. Philp
From the National Centre for Epidemiology and Population Health, Australian National University, Canberra, Australia

Comments and/or questions to authors:

1. Discrete vs. continuous-time models

To what extent discretisation of variables or their continuous limit can play a role?
See, in particular the following paper:

The Importance of Being Discrete - Life Always Wins on the Surface
Nadav M. Shnerb, Yoram Louzoun, Eldad Bettelheim, Sorin Solomon

later published in Proc. Nat. Acad. Sci. USA

2. Possible modifications of parameters in the present model

General comment: it seems that this paper deals with a "least bad" scenario in terms of potential pandemic insurgence. and the "rural origin" limits the number of potentially infected travellers. Moreover, there are only 2 areas considered. The at-risk zone is not receiving infected people from more than one geographical region. Only air travel is taken into account, etc.. Even though, the main message is that none of the two provisions suggested could give more than a few days of delay.

2.1 It's important to see that the "rural origin" aspect gives you a conservative assumption for number of travellers.
What would change if Hong-Kong is the source?

2.2 What about taking into account asymptomatic screening?

3. Relationships with "networkology"

There are at least two points where there could be an influence of the structure and dynamics of networks:

a. for social networks: on the likelihood of an infected traveler and all their in-flight offspring to fail to or to initiate an epidemic on arrival (random variable D_2);

b. for travel networks: on the chain of events leading to the entrance of undetected infected passengers into the at-risk zone (random variable D_1).

For the latter, a couple of examples of the questions that can be asked. If we think about a stupidly simple network,

source neighbour indirect neighbour

... and there are flight-based quarantine measures in place along both routes, will the indirect neighbour get a cumulative benefit from the time offset of the screening measures? Not necessarily cumulative in the sense of doubling the figures given in the paper, but at least in
the sense of the benefit of screening measures increasing in some proportion to network distance (path length).

More sophisticated networks, where regions have multiple neighbours and there are multiple paths between any two regions, will complicate matters but the question is still worth considering.

In the case of more complex networks, one could consider whether _severe_ quarantine restrictions at a few highly-connected hubs (e.g. quarantining all travellers regardless of symptoms and strongly restricting the amount of travel) might be enough to prevent or strongly
delay spreading to the rest of the network. If so, most air travel could continue unrestricted in exchange for very strong quarantine measures at a few important places, and collective economic and other effort could be put in place to offset the negative effects on these regions and travellers to those regions.

Regarding travel networks there is a recent paper:

Colizza et al., PNAS 103, 2015-2020 (2006)