HIT National Editor

Surveillance — watching where the flu, or some other pandemic outbreak, breaks out – is a basic methodology, except that on a much-travelled, small-planet environment, the flu may travel faster than our systems to track and keep up with it. As quickly and easily as a transoceanic or cross-country plane flight, it can move from continent to continent, country to country, state to state faster than reports can be turned in, aggregated, analyzed and acted upon.

Thus, delivering the therapeutics to deal with outbreaks may lag up to two weeks or more behind their ability to make real impact.

But here's an alternative technique: predict where a flu will occur, stay ahead of it or at least be able to make very good guesses where it is headed, how virulent it might be, how long it might last, how many people may be effected

And researchers at the University of Iowa (UI; Iowa City) are trying to make doing this almost as easy as saying it, using "prediction" markets, a technique for gathering a variety of expert opinions, putting them together in real-time fashion, and refashioning and making more accurate these predictions as new evidence comes in.

This is one of the applications that has been developed by the prediction market known as the Iowa Electronic health Market (IEhM), a system already shown able to predict such things as how well a movie might do, to who will be the next president, to where a flu might occur next and how long it might last.

Forrest Nelson, professor of economics in the Tippie College of Business at UI and the IEhM's director, told HealthCare InoTech that he and "two other guys" devised the prediction market concept in 1988 and that the technique has proved valuable both for making predictions and for attracting considerable attention to the issues being predicted.

The initial work in this area was established as the Iowa Electronic Market, with the IEhM then spun off to focus on healthcare issues.

In May the IEhM opened several new markets specifically to track and forecast trends in the current H1Ni swine flu outbreak.

The H1N1 swine flu market can be found online with the IEhM's other health markets at Initial offerings in the market serve to aggregate information on five separate questions:

– How long will the current outbreak last in the U.S.?

– How many U.S. states will be affected?

– How many cases will be confirmed in the U.S.?

– What will the mortality rate be among confirmed cases?

– How many other countries will be affected?

The 'futures market' model

Nelson told HIT that a prediction market is really a type of "futures" market used by traditional stock exchanges, such markets essentially allowing traders to make their most informed guesses concerning, say, the future prices of gold or grains, with those participants backing their best guesses with their hard dollars.

Rather than real dollars, IEhM participants are provided "funny money," virtual dollars that are traded as "contracts," essentially the same way as a futures market, the value of those contracts eventually determined by how well their predictions match the emerging reality.

Participants are invited to be market "traders," based on their expertise in the particular market, such as, in the case of the flu, nurses, doctors, diagnosticians, public health officials, or anyone who may have insight or data concerning the disease and its progress.

They trade shares based on the certainty they have about their knowledge, Their payoffs are two-fold, according to Nelson, their inherent interest in the accuracy of their predictions and, if they are accurate, receiving "educational grants" that they can use to buy text books or pay for attendance at a conference.

"The more confident traders are [about their predictions], the higher price they can pay," he said

Nelson in mid-May estimated about 60 active participants "trading" on the flu markets established by the IEhM. He said that the system's predictions have been shown to be highly accurate with as few as 20 trading participants, but he acknowledges some misses, such as in the 2004 U.S. presidential election.

"It's not perfect; just like any tool you can make a mistake," he said "But our predictions, compared to those available from other sources, stack up pretty well."

Filling the predictions 'vacuum'

He added: "More to the point, sometimes there are no other predictions," citing the case of its market for predicting avian flu. "There were no public predictions. There were guesses, speculation, but no hard numbers."

Thus, he says it is a "good thing" to offer "some information" where there might otherwise be an information "vacuum."

He says that the prediction market received the earliest significant notice when participants in the market for the 1988 U.S. presidential election saw a rising interest in the campaign of Jesse Jackson, a trend that was otherwise not seen or simply dismissed by the traditional media.

The Robert Wood Johnson Foundation (Princeton, New Jersey) is backing the system's predictive value with a $1.1 million grant to support the IEhM's testing and expansion of its seasonal influential prediction markets.

Nelson downplays the value of predictions for election campaigns, because having largely entertainment value, and says that the value of healthcare predictions are discernibly and obviously more important.

The marketing application is obvious: for instance, predicting how well a new product might do in commercialization, and in healthcare the technique has been used to forecast outcomes such as which drugs have the best chance of advancing through clinical trials and on to commercialization.

Nelson said that the IEhM is the first research group to use prediction markets for public health issues.

"To date, the IEhM has run markets to predict seasonal influenza activity at a state level and also the number of avian influenza cases," he said. "Both of these markets have provided accurate and timely infectious disease forecasts."

Besides recruiting members of the clinical, public health and animal health communities as participants, the IEhM said it also is "considering" participants from other fields as well.

No policy-making – yet

Nelson says the prediction markets are meant entirely as a method for supplementing other information, where it is available – providing "another piece of the puzzle," Nelson said – and that they have not yet been used to establish public policy-making of any type.

It has, however, caught the eye of the Centers for Disease Control & Prevention (Atlanta), which has seen steady increases in syphilis since 2002. A prediction market for syphilis potentially could be used for deploying resources to head off predicted increases in various region of the country and that CDC officials have shown interest in the technique.

Nelson says that, as a general rule, health officials "tend not to be in the game of predicting anything," perhaps because the whole effort smacks of predicting the weather, which is so highly unpredictable, and health officials may fear the consequences of making predictions that are wrong.

"[Health officials] do a good job of surveillance – yesterday, last year, today – but nothing at all about what's happening tomorrow. But you would think that is most important."

Providing this sort of benefit should be the major purpose of a serious prediction market, Nelson told HIT. "If it's just used for entertainment, it's not worth doing. If it could actually inform decision-makers, that's the ultimate goal."