Friday, June 5, 2009

Modeling the Swine Flu Epidemic: Hyped up or Toned down?

-C. Paula de los Angeles

In its report on May 15, the CDC estimated that there are upwards of 100,000 cases of swine flu in the U.S., though there are just over 7,000 cases reported. However, these figures are far from conservative estimates of around 2,000 to 2,500 cases from two different leading research groups at Northwestern University and Indiana University made in the last days of April.

So, what happened? Engineering professor Dirk Brockmann of Northwestern blames the mistake on incorrect initial estimates of cases in Mexico that were magnitudes lower than reality, that were applied to a model that is very sensitive to initial conditions. A redeeming trait about Brockmann's model is that he correctly predicted the geographical spread that highlighted California, Texas, Illinois and Florida as "hot spots".

On the other hand, informatics professor Alessandro Vespignani of Indiana University, saying that he was misquoted or misunderstood, and in actually predicted just over 9,000 cases. However, this estimate is still orders of magnitude off.

While both models need refining, Dirk Brockmann states, “For this disease, we won’t put out another projection,” he said. “Once it’s in the dispersal phase, exponential growth kicks in. You don’t need a sophisticated model anymore.”

Another difficulty with refining the model, is that the CDC has decided to stop confirming all swine flu cases in the laboratory. Without good data, modeling is almost impossible.

When Dr. Bob asked us at the beginning of the swine flu scare if we thought it was all hype, I was of the belief, that over-hype is better than no hype. In this case, lack of hype is disconcerting. The extremely conservative estimates give off a false sense of security to the general public and health officials. Models, in general, are difficult for the public to accept, and ones that are as inaccurate as these, do not add any confidence. These conflicting interpretations demonstrate how sensitive models are to a variety of factors. I am hopeful for more reliable data-keeping, better refinement of models, and new interpretations of the models.

News link here:

No comments:

Post a Comment