Posted by
Douglas Roberts-2 on
Mar 30, 2007; 9:45pm
URL: http://friam.383.s1.nabble.com/One-of-my-projects-tp523626p523627.html
A few of you have asked questions about the the EpiSims-Grid project, so
I'll try to answer them here, in roughly inverse order that they were
received:
From: Paul Paryski:
For someone like me who rarely works with such complex models this is a
> very interesting discussion. Out of my ignorance a couple of questions have
> popped into my aging synapses:
> -does the model include mutation and other adaptations by diseases?
>
No, we have only simulated one pathogen at a time, to date, and it does not
mutate.
-are you going to study past massive epidemics to see what patterns are
> applicable (bio mimicry and incidence of natural immunity, cultural
> practices)
>
Yes, we have done this fairly extensively. Lots of data exists from the
1918 pandemic flu outbreak, for example.
-who will make the political choice to use the info/models when the time
> comes?
>
Good question. See my response to Laura Mac's questions bleow.
>From Laura MacNamara:
Being someone who studies people who use models, I'm curious about how you
guys are relating to your user community. Who are the intended analysts
(the ones that you hope know what you're doing)? At what point do you guys
start engaging them? Do they treat your simulation as black box?
Our last study was commissioned by a high-level consortium of
Department-level representatives -- Dept. of State, Dept. of Treasury, Dept.
of Homeland Security, Dept. of HHS, and the office of the White House. The
purpose of the study was to help them identify relative measures of
effectiveness regarding what intervention strategies would provide the most
benefit in the event of a pandemic flu outbreak. Examples of intervention
strategies that were modeled included
1. Self-isolation (staying home when symptomatic)
2. Social distancing (telecommuting, scheduled trips to the store with
minimal contact to other shoppers, in general minimizing physical proximity
to other people) during an outbreak
3. Closing down schools and non-critical workplaces
4. Treating critical infrastructure workers with anti-viral treatments
(remember -- it was a pandemic being simulated, there were no vaccines)
5. etc.
The intent was to help government officials develop a response plan in the
event of an outbreak. I was quite impressed with the expertise with which
the leader of the study, the White House representative, directed the
study. He was one of the most knowledgeable and intelligent of any of the
customers that I have aver worked with. The simulations used in the study
were most definitely not treated as black boxes. Rather, the strengths and
weaknesses of each of the three models were thoroughly explored.
The consortium of users approached the leader of the MIDAS project and
requested our participation on the project last summer, at which point we
immediately engaged with them to develop an experimental design.
>From Robert Holmes:
Fair enough: big simulation answers some questions, small simulation answers
others. So what are the specific questions that a big epidemiological
simulation can answer? It can't be anything too predictive ("ohmigod, New
York has just fallen to small pox. Which city is next?") because that
depends (I'd guess) on something that is unsimulatable ("errr.... dunno.
Kinda depends which flight the guy with small pox got onto"). What are the
questions that can only be answered with a big model?
EpiSims was by far the most detailed of the three models used. It is an
individual-based ABM in which the second-to-second movements of every
individual in the 8.6-million population city were modeled for 60
consecutive 24-hour days. Further, each individual was fairly completely
characterized demographically -- race, inccome, marital status, number of
children, etc. Also, family household structures are created by EpiSims, in
which the same adults and children come back to the same household every
day.
This level of detail allowed us to run experiments on specific demographic
subsets of the population that were not possible with the other models. For
example, we ran a series of experiments for which social distancing was less
effective among lower income people, because they could not afford to stay
home -- they had to work. These runs were compared to runs where all
working members of the population had the same compliance when social
distancing measures were imposed.
Another example of experiments that were conducted with EpiSims that could
not be achieved with the other models: we ran several experiments in which
the imuno-response of lower economic segments of the population was less
effective in resisting the pandemic virus then for those more affluent
members of the population. The reasoning being that poorer people have less
access to health care.
Remember, the intent of these studies was to establish a relative
effectiveness ranking determination of various intervention strategies for
future use establishing a response strategy in the event of a pandemic
outbreak. The intent was *not* to model "ohmigod, New York has just fallen
to small pox. Which city is next?" types of human behavior in response to an
outbreak.
I hope this addresses some of your questions. Thanks for your interest!
--Doug
--
Doug Roberts, RTI International
droberts at rti.org
doug at parrot-farm.net
505-455-7333 - Office
505-670-8195 - Cell
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://redfish.com/pipermail/friam_redfish.com/attachments/20070330/9514c858/attachment.html