Tuesday, May 14, 2013

Live from FOCI 2013 - Using Prediction Markets to Crowdsource Medical Diagnoses

-->

Jared Heyman, CROWDMED

Think of this:  Wikipedia is more accurate than Encyclopedia Britannica was.  Because of the accuracy issue – and because of the cost factor (free vs. expensive), the Britannica is now out of business.  Wikipedia is an excellent example of the accuracy, usability, and practicality of crowdsourcing. 

Prediction markets are embedded in contemporary life – from the Iowa elections predictions to Oscar winners.

On a beach in Brazil, Jared wondered: If crowds are so smart and prediction markets are so accurate, could they be applied to medical diagnoses?

The prediction marketing technology (CROWDMED) was already in place at Jared’s company, so it didn’t seem like a long stretch to use the platform for gathering medical condition knowledge.  With Jared’s sister’s illness as the test case, Jared put his theory into practice. 

In three days, the crowd came up with five possible disorders that could be the cause of Jared’s sisters’ symptoms.  The crowd was able to make bets on the correct diagnosis, and the decision was that she had a condition called: Fragile X Associated Primary Ovarian Insufficiency (FXPOI).  An interdisciplinary team later confirmed that this was the correct diagnosis, and she was successfully treated, returning to normal in just three weeks.

An estimation of medical costs for diagnosis services is about $4,078,655.  This doesn’t include the human costs that become entangled with the medical diagnoses processes.  For $200, the CROWDMED website lets people submit symptoms or contribute a diagnosis.  The process is very gamified.  Prediction markets seem to work best when people are betting something of value to them.  In fact, play money that has value to people can be as effective as real money.

The “medical detectives” earn points and select a case that has been submitted.  Interestingly, longer cases have higher rates of participation than do shorter cases.  Participants seem to like the challenge.  The diagnoses that people submit are linked to Wikipedia articles about the diagnosis.  The medical detectives can then bet on a diagnosis.  Interestingly, more points are awarded for betting on a less popular diagnosis.  Finally, the participants determine how much they will bet on a diagnosis, up to 1,000 points.  The amounts are allocated to the diagnoses that have been selected, and the potential points that will be awarded  - based on the diagnoses chosen - are also shown at this time.

The success of the CROWDMED project has driven interest in this company through the roof, even though the company has just gone public.   The platform serves as a demonstration of power of shared knowledge, and demonstrates that convenient, speedy, and focused communication works best, particularly – apparently – outside of traditional systems.

Gigi DeVault writes a market research column for About.com.  http://marketresearch.about.com/

1 comment:

Anonymous said...

Really interesting, thanks ​!

​Given your interest, I think that you (and the other readers here) would be really interested in some recent research that I have come across that theorizes about crowds and such similar phenomena.​ ​

It’s called “The Theory of Crowd Capital” and you can download it here if you’re interested: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2193115

In my view it provides a powerful, yet simple model, getting to the heart of the matter. Enjoy!