Medical detectives deal in challenging diagnoses on CrowdMed’s virtual stock market

The combined intellectual power of crowds has never been used like this before. CrowdMed’s founder and CEO, Jared Heyman spent a good part of his career studying the wisdom of crowds. So when his sister came down with a debilitating illness during her freshman year of college, the idea of crowdsourcing as a diagnostic tool […]

The combined intellectual power of crowds has never been used like this before. CrowdMed’s founder and CEO, Jared Heyman spent a good part of his career studying the wisdom of crowds. So when his sister came down with a debilitating illness during her freshman year of college, the idea of crowdsourcing as a diagnostic tool seemed like a no-brainer.

Already armed with access to a patented market prediction system through his company infosurv, Heyman plugged in his sister’s case and watched as more than 300 participants discussed and voted on various diagnostic options.

“Through crowdsourcing, the participants were able to identify in two weeks what took doctors more than three years to diagnose,” he said.

Infosurv’s prediction model used people from survey panels — huge databases of Americans who agree to take online surveys. Used mostly for market research purposes, the random sample of participants intentionally reflects the demographics of the general U.S. population.

The current CrowdMed platform is similar, allowing anyone to participate as a Medical Detective.

“We don’t think it’s a good idea to screen people based on formal credentials. We believe that the insight of someone without medical board certification could lead to a correct diagnosis or cure,” Heyman said.

It’s the model itself that figures out who is heading toward an accurate diagnosis. In fact, there’s a “cost” to play. Not a financial cost, but a finite number of points allotted to each Detective and you wouldn’t want to waste those points on a case you couldn’t win.

The difference between a simple survey model and a prediction market is that survey participants are incentivized for participation, while a prediction market incentivizes participants for performing well. So, while anyone can offer a prediction, those who put forth special insights that could help crack the case are rewarded and pushed to the top of the list. Better predictions could be the result of a background in medicine or a personal connection with a certain illness.

The actual platform behaves a lot like a virtual stock market, where each diagnosis is a stock. Each share price moves in reaction to how Detectives bet or “invest” in it.

“It works just like the real stock market, so if there’s a lot of demand for a certain diagnostic suggestion, that drives the price up and drives down the point reward if it turns out to be right,” Heyman said.

Inversely, if a diagnosis is not popular, but a Detective is sure they’re right, the potential reward could be much higher. In a market mechanism, this is called truthful revelation, a phenomenon where people reveal how much confidence they truly have on a given position.

The algorithm takes the aggregate betting behavior over time and assigns a probability to each diagnosis being correct. “At this point, we show the patient which diagnoses we think are best based on the Detectives’ betting behavior.”

Medical Detectives are often physicians, nurses, medical students, and people who have personal insights on complex conditions. Even with a cash reward model in place, Heyman insists that Detectives’ main motivator is helping people.

“Plus, it’s a great learning exercise,” he said. “Not only are healthcare professionals exposed to new and varied cases, they can also learn a lot from fellow Detectives.” All responses are visible in real time to everyone on the site, and Detectives can discuss the cases they’re trying to solve. “Crowds are most wise when they have discourse and debate.”

In terms of revenue, the company is currently charging a 10 percent commission on all cash rewards offered by patients for accurate diagnoses. Future models may incorporate third-party payers.

“Insurance companies have a lot to gain from a site like this,” said Heyman. “I think all interests are very much aligned. The payer and the patient want the patient to get well as soon as possible.”

Of the diagnosed patients the CrowdMed team has been able to track, 78 percent have reported that the diagnosis they received via the site turned out to be accurate.

“Our average patient has seen eight doctors and racked up more than $50,000 in medical expenses by the time we get their case,” said Heyman. “This could revolutionize the way that rare diseases are diagnosed.”

CrowdMed was one of the 46 companies in the Y Combinator Winter 2013 class. The company has raised $2.4 million to date, including money from NEA, Andreessen Horowitz, Greylock Partners, Y Combinator, and SV Angel.