[Paleopsych] IDSA: Wall Street Methods Used to Predict Flu Outbreaks

Premise Checker checker at panix.com
Wed Oct 19 01:50:42 UTC 2005

IDSA: Wall Street Methods Used to Predict Flu Outbreaks
By Rabiya Tuma, Ph.D. , MedPage Today Staff Writer

SAN FRANCISCO, Oct. 7 - Using Wall Street traders' techniques, healthcare 
workers predicted flu outbreaks two weeks before the actual events, according 
to a pilot study.

The CDC tracks influenza data but lacks a method of predicting future trends, 
the researchers reported here. In the business world, so-called "prediction 
markets" have been used to accurately forecast the outcome of a variety of 
chance events, including political elections, the number of tickets sold on the 
day a movie opens, and equipment sales.

To test whether such market methods would be useful for prediction of 
infectious diseases, researchers at the University of Iowa recruited 60 
healthcare workers from their state and organized a stock-market scenario. They 
described their methods and results at a meeting here of the Infectious Disease 
Society of America.

All the "traders" were given 100 flu dollars at the start and were asked to buy 
and sell "stock" according to how much influenza activity they thought would 
occur in a given future week. The goal was to try to predict the level of an 
outbreak ultimately recorded by the CDC, which is reported by a five-stage 
color code.

If they saw hints of flu in their patients, for example, they might want to 
sell any shares they owned that predicted a low-level week in the near future 
and buy stocks that predicted many cases.

With 52 traders actively participating on the website, market predictions were 
able to hit the right color for the state of Iowa 80% of the time, two to three 
weeks ahead of the actual cases. That means they were three to four weeks ahead 
of the CDC's announcement of the actual data, which comes out the week 
following the outbreak week in question.

"Ninety-percent of the time, the market was able to get within one color four 
weeks ahead of the event," said Forrest D. Nelson, Ph.D., an economist at the 
university who has developed prediction markets and is a member of the research 

At the end of the study, the flu dollars were converted into U.S. dollars, and 
each participant won a grant of that amount. The most successful trader's grant 
was worth $213.

"There is currently no method for forecasting influenza activity, but the 
information needed is out there," said Philip M. Polgreen, M.D., the infectious 
disease specialist who led the study. "But it's spread out." The prediction 
market aggregates individual experiences into collectively useful data, he 

Two weeks of advance warning about an influenza outbreak would permit 
significant preparation, including immunization of the most vulnerable 
population, making sure pharmacies have adequate antivirals in stock, and 
increasing staffing at hospitals.

The team plans to expand the "market" to include 100 traders in Iowa this year, 
and then to cover the U.S. within a year after that. It is not yet clear how 
many traders will be needed to cover the whole country. "We hope to be posting 
prediction data next year," Dr. Polgreen said.

The researchers also plan to open a "strain market" to predict the prevalence 
of individual flu strains.

More information about the paleopsych mailing list