Modeling Pandemic Diseases

Janine Burbage, IBM Healthcare & Life Sciences, Business Development


There isn't a day that goes by that I don't get at least 36 Google alerts on "avian flu" or "pandemic." I'll be blogging on this and other important public health issues in future posts. But if your summer vacay left you a little behind on the subject of pandemics, let me give you the low down..

It has been 36 years since our last pandemic (the last one was the Hong Kong Flu in 1968-1969) and many scientists believe that it’s only a matter of time until the next one occurs. Its severity cannot be precisely predicted, but modeling studies show that the effect in the United States could be severe. A highly contagious crisis could affect 66 million Americans, with two million people hospitalized and as many as 500,000 deaths. With a little over 965,000 staffed hospital beds in registered American hospitals, imagine how overwhelmed the U.S. healthcare system would be. As the toll from Hurricane Katrina shows, the economic impact could be as massive as the loss of life, between $71.3 billion and $166.5 billion.

So how likely is it that the Avian Influenza will be the next pandemic? No one knows. The Avian Influenza virus (H5N1) was first identified in 1961 in South Africa, but it mainly affected birds and animals. H5N1 is zoonotic, which means it is capable of jumping from animals to humans. The real danger is if the virus mutates to become contagious from person to person. In 1997, the Avian Flu that emerged in Hong Kong killed 6 out of 18 infected. The outbreak was halted until 2003. As of August 5, 2005, the World Health Organization has reported 112 confirmed cases with 57 deaths. The Avian Flu has made its way to Russia, and a few unconfirmed cases are being reported in Finland.

Let's talk a little bit about computer modeling...tools and techniques that not only predict how and where the disease will spread, but can also quantitatively evaluate different response strategies. IBM doesn’t develop epidemiological models itself, but we have the expertise in building modeling frameworks for the kinds of activities mentioned above. For example, IBM Research has developed Spatial-Temporal Epidemiological Modeler (screenshot, left), a public health forecasting tool that aids policy makers and planners in evaluating appropriate response strategies to natural or biologic threat of emerging infectious diseases.  It is available for researchers to download and use for free in building their own pandemic models.

But how useful are models? Some skeptics of modeling feel that regardless of how much modeling one performs, the results are only as good as the data used to run the model. Point taken. So I ask you, how do we improve modeling and simulation technologies so that they become trusted and valuable tools for the advisors and decision makers in state, local, and national governments and public health departments?

Comments

[Lei]

Another question might also be how do we separate politics from public health so that people are protected, not political interests?

This blog is shaping up to be an interesting read with all the topics being covered and the impressive panel of bloggers.

[Janine Burbage]

True - but I think that we need better communication between policy makers, planners, and decision makers and epidemiologists. I don't think that the problem w/ detecting and responding to incidents is technical - it's related to policy. There are a lot of generic plans, but we need operational blue prints. Each agency and organization may implement the blue print a little differently, but modeling solutions, and frameworks exist that can help those making decisions when a crisis erupts. Did you see that there was a model created for "Hurricane Pam"... a simulation of a cat 3 hurricane that would strike New Orleans? Even after tabletop exercises and action plans that could be implemented, I don't think that anyone took this plan or simulation seriously. Do you think this is because models have made people "cry wolf" in the past?

[Lei]

The one thing I don't think models are good at predicting is human behavior. Despite all the public health messages over the many years of the health risks of certain behaviors like smoking or high fat diets, it's been hard to really get people to change. The situation with Hurricane Katrina goes way beyond predicting disasters via sophisticated models.

[James Kaufman]

Actually some of the models are quite good at capturing human behavior. Perhaps the behavior is not quite what one would always like to see - but it can be predictable. Perhaps with a national health infrastructure we could move beyond modeling and towards better policy making. With real time data it might be possible to discover emerging threats to public health much earlier and use modeling to evaluate alternative public health policies.

[Janine Burbage]

Let's go to this capturing human behavior issue... do you think that models are not good at capturing what humans will do bc of too many external factors that we don't account for in the model? I think that can and should be built in. The problem is that lack of computing power and memory to handle all of the agents.

[Susan R]

Health department was working to ensure supply of quality medical drug for the consumptions of general public but we have no receive any better service in these matter.

[Susan R]

Health department was working to ensure supply of quality medical drug for the consumptions of general public but we have no receive any better service in these matter.