Health Professionals Dabble in ‘Flu Prediction Market’

By Sandy Kleffman, Contra Costa Times, Walnut Creek, Calif.

Oct. 7–SAN FRANCISCO — Forget corn and soybean futures. The market to watch may soon become flu futures.

Several dozen health professionals have been wheeling and dealing, predicting when the annual influenza season will arrive.

Think the bug is around the corner? They sold off “no activity” shares and nabbed ones indicating the flu is about to spread.

Guess wrong and it costs them. Get it right and their funds expand.

This isn’t a macabre way to earn a fast buck or easy entertainment.

Health leaders say it could be a crystal ball in an area where scientists have few answers now.

Everyone knows the flu season will arrive each year, but when and how severe it will be remains a mystery. Last year, a mild season occurred late, amid widespread angst over a vaccine shortage. The year before, the flu hit abnormally early.

The test run indicated the pooled knowledge of “futures” investors produced an earlier and more accurate forecast.

“Even one or two weeks’ notice would be helpful,” said Dr. Philip Polgreen, an infectious diseases associate at the University of Iowa. “For example, it would provide an opportunity to increase vaccination rates. It would also help us plan for an increase in admissions to the hospitals.”

Pharmacies would know when to order additional antiviral medications.

Nursing home administrators could make sure their residents were protected.

Polgreen outlined the results of last year’s flu prediction market Thursday at the annual meeting of the Infectious Diseases Society of America.

The idea is simple. Conceived by professors at the University of Iowa, it bears similarities to markets that forecast political election results and Academy Award winners, or the Chicago commodities markets that trade in corn and soybean futures.

The 60 “traders,” operating in eastern Iowa included physicians, nurses, pharmacists, clinical microbiologists and epidemiologists. Each day, they bought or sold shares based on what they were seeing in their practices. If a microbiologist suddenly noticed an increase in respiratory cultures that tested positive for the flu, he or she would leap into action, unloading “no activity” shares and buying up those for “widespread activity.”

At the end of the season, researchers compared the market activity with what actually occurred, based on reports from the national Centers for Disease Control and Prevention. They found the market to be remarkably accurate.

“Our group is presenting some promising pilot data suggesting that prediction markets can forecast influenza quickly, accurately and inexpensively,” Polgreen said.

“Using this method, we were able to predict influenza two weeks in advance and we were able to provide important information to health care workers four weeks in advance to help plan ahead.”

While the CDC collects information from health experts throughout the nation, by the time it publishes the data, it is two weeks old. The prediction market enables health experts to pool their observations quickly by use of the Internet, Polgreen said.

He and his colleagues plan to expand the market statewide in Iowa this year, and hope within the next five years to make it nationwide. They also want to use it to predict the strain of influenza that should be included in next year’s vaccine, and may eventually open an avian influenza market to help monitor the emerging health threat.

None of the traders will be buying million-dollar homes or BMWs with their proceeds.

At the beginning of the flu season, they each receive a $100 educational grant to use for their trades. Last year, the most successful trader ended with $213. All of the winnings must be spent for educational purposes, such as buying books or enrolling in a medical conference.

Researchers have not determined how many traders would be needed for a nationwide market, but it might be as few as 50 per state, said Forrest Nelson, a University of Iowa economics professor.

“What you need is good sources of information,” he said. “So if you have the right people and the right distribution, you can get by with very few.”

Businesses have begun to use similar models.

“Google now runs these to predict for strategic information,” Polgreen said.

“Hewlett Packard has run prediction models to forecast printer sales. But to my knowledge, we’re the only group using these models to predict infectious diseases.”

—–

To see more of the Contra Costa Times, or to subscribe to the newspaper, go to http://www.bayarea.com.

Copyright (c) 2005, Contra Costa Times, Walnut Creek, Calif.

Distributed by Knight Ridder/Tribune Business News.

For information on republishing this content, contact us at (800) 661-2511 (U.S.), (213) 237-4914 (worldwide), fax (213) 237-6515, or e-mail reprints@krtinfo.com.

GOOG, HPQ,