Skip Navigation

header

Electronic Health Markets Predict Spread of Dengue in U.S.                   

By Crystal Franco, Tara Kirk Sell, Anson Tai Yat Ho, and Phillip Polgreen, November 5, 2010

Knowing the number, severity, and location of cases is essential to responding effectively to an infectious disease outbreak. Many federal, state, and local surveillance and reporting systems operate with the goal of providing outbreak information to aid public health, medical, and political decision making. However, information collected through traditional surveillance methods often lags days or weeks behind the unfolding epidemic, making it difficult to implement epidemic control measures or mobilize the resources needed to support medical and public health response.1,2

In an attempt to eliminate this lag, researchers are exploring the use of “prediction markets,” a methodology proven to be of value in other fields, to supplement traditional disease surveillance. In August 2010, the University of Iowa, along with input from the Center for Biosecurity, opened dengue fever markets for trading in the Iowa Electronic Health Markets.

Dengue Cases Rising

There is considerable evidence that case numbers for this mosquito-borne viral disease are rising and that geographical distribution is widening within the U.S. and around the world. As of November 4, 2010, 368 cases of dengue have been reported among the 50 United States.3

The accuracy and frequency of dengue morbidity and mortality reporting varies geographically. Because traditional methods of disease surveillance may not accurately capture the true impact of this disease in a timely manner, we wanted to see if synthesizing and quantifying the observations and predictions of experts would accurately forecast the course of the growing epidemic in the U.S. Professionals in public health, medicine, and vector control were invited to participate.

What are Prediction Markets?

Traditionally, prediction markets have been used as an economic forecast for such things as commodity prices or movie box office returns, or to predict the outcomes of events such as sports matches and political campaigns.4,5,6,7 More recently, these markets have also been used to forecast developments in infectious disease outbreaks, including influenza activity.

Prediction markets quantify professional opinion, knowledge, and experience by asking participants to trade shares based on their degree of certainty of a particular outcome. The dengue markets allow experts in dengue-related fields to purchase shares (which are called contracts) that are associated with outcomes they expect to occur and to sell shares they believe are associated with unlikely outcomes.

The prices that traders are willing to pay reflect the strength of their belief in particular outcomes. The final value of each share (when trading is completed) is determined by actual events as determined through traditional surveillance. For example, when the market closes, a trader will receive $1.00 for each share s/he owns that accurately predicted the number of dengue cases, and $0 for each share that incorrectly predicted the number of cases.

Dengue Market Methods

Participants in the dengue markets are given $100 of a valueless currency with which to trade and make predictions in 4 categories:

  • Total number of dengue cases reported in the 50 United States in 2010

  • Percentage increase in clinical dengue in the Americas in 2010

  • Number of states in U.S. that will report locally acquired dengue cases in 2010

  • Percentage increase in severe dengue cases in the Americas in 2010

For each of the 4 market categories, participants are provided with choices of outcomes. For example, question 3, which asks participants to predict the number of states that will report locally acquired dengue cases by the end of 2010, has 4 possible predictions:

  • 1 state

  • 2 states

  • 3 to 5 states

  • 6 or more states.

Results Thus Far

As of October 30, 2010, 70 participants were actively trading on the dengue markets. A total of 2,477 transactions occurred with a total of 5,284 contracts traded since the markets opened on August 17. Predictions about the total number of U.S. dengue cases up to October 8, are plotted in Figure 1. The height of each shaded region at any particular date represents the predicted probability, that the corresponding event will occur as of that date.          

This example graphed below shows that, on September 24, market prices indicated an 80% chance of greater than 301 U.S. dengue cases—the most likely outcome. And for this market, traders accurately predicted that the number of cases would rise above 301 as early as September 16—more than 3 weeks before the CDC reported cases above 301 on October 8.9

Figure 1

*Figure provided by University of Iowa Electronic Health Markets group

Conclusions

The consensus opinion of each dengue market has accurately forecasted changes in dengue activity. This outcome provides preliminary evidence that, by aggregating the experience and knowledge of experts, markets are potentially useful complements to traditional disease surveillance.

Interested readers may register and participate in current health markets through the Iowa Electronic Health Markets websitehttp://iehm.uiowa.edu/iehm/index.html. Participation is free.
  

References

  1. Harmon K. Advances in disease surveillance: Putting the “public” into public health. Scientific American. March 13, 2010. http://www.scientificamerican.com/blog/post.cfm?id=advances-in-disease-surveillance-pu-2010-03-13. Accessed October 29, 2010.

  2. Jajosky RA, Groseclose S. Evaluation of reporting timeliness of public health surveillance systems for infectious diseases. BMC Public Health. 2004;4(29). http://www.biomedcentral.com/content/pdf/1471-2458-4-29.pdf. Accessed November 1, 2010.

  3. U.S. Centers for Disease Control and Prevention. Notifiable Diseases and Mortality Table. MMWR. November 5, 2010; 59(43):1416-1429. http://www.cdc.gov/mmwr/preview/mmwrhtml/mm5943md.htm?s_cid=mm5943md_w. Accessed November 4, 2010.

  4. Smith VL. Constructivist and ecological rationality in economics. American Economic Review 2003;93:465–508.

  5. Pennock DM, Lawrence S, Giles CL, Nielsen FA. The real power of artificial markets. Science. 2001;291: 987–988.

  6. Cowgill B, Wolfers J, Zitzewitz E. Using prediction markets to track information flows: Evidence from Google. Dartmouth College. 2008. www.bocowgill.com/GooglePredictionMarketPaper.pdf. Accessed November 3, 2010.

  7. Dvorak P. Best Buy taps 'prediction market’. Wall Street Journal. September 16, 2008. http://online.wsj.com/article/SB122152452811139909.html. Accessed November 3, 2010.

  8. Polgreen PM, Nelson FD, and Neumann GR. Use of prediction markets to forecast infectious diseases activity. Clin Infect Dis. 2007;44(2):272-279.

  9. U.S. Centers for Disease Control and Prevention. Notifiable Diseases and Mortality Table. MMWR. 2010;59(39):1282-1295. http://www.cdc.gov/mmwr/preview/mmwrhtml/mm5939md.htm?s_cid=mm5939md_w. Accessed November 1, 2010.
       



Crystal Franco, MPH, is an Associate at the Center for Biosecurity; Tara Kirk Sell, MA, is an Analyst at the Center for Biosecurity. Anson Tai Yat Ho is a PhD student in the Department of Economics at the University of Iowa. Philip Polgreen, MD, MPH, is an Assistant Professor in the University of Iowa Carver College of Medicine, and the Director of the Infectious Disease Society of America’s Emerging Infections Network.