Can Crowdsourcing Your Symptoms Reveal What Ails You?by
Amin Azzam gained a measure of crowdsourcing cachet last year by conceiving of a class where medical students would receive course credit for editing Wikipedia pages about various diseases. The idea landed him on a panel at a conference where another participant, Jared Heyman, asked him to help get his students involved in a more ambitious project: diagnosing patients they had never met.
Heyman is the founder of CrowdMed, a company where patients with difficult-to-diagnose ailments post their symptoms online, offering cash rewards to people in exchange for correct assessments. The company was founded in 2012 after Heyman watched his sister struggle with a rare disease known as fragile X-associated primary ovarian insufficiency. CrowdMed’s target audience is people with so-called orphan diseases, those that affect fewer than 200,000 people. There are about 6,800 such diseases affecting about 25 million Americans in all, according to the National Human Genome Research Center. Such conditions are notoriously difficult to diagnose.
“We’re not looking for patients who haven’t seen a doctor yet,” says Heyman. “We’re looking for patients who have seen two or three doctors and don’t have a diagnosis they think satisfies their symptoms.”
To get the project moving, Heyman needed to build a stable of qualified people interested in spending time poring over medical records for pocket money. The system operates like Wikipedia, with entry open to anyone. But there are two logical sources of such medical detectives: retired physicians and medical students.
Azzam had access to medical students, and jumped at the idea to sign them up. He’s working to get CrowdMed onto the curriculum at the med schools at the University of California in San Francisco and Berkeley, where he teaches, and he hopes to have a pilot running by the next academic year. Azzam believes working on real-life cases rather than simulations helps students learn to puzzle through the diagnostic process. “When we’re teaching those skills to early medical students, they’re not good at it, because they’re not doctors yet,” he says. “Platforms like CrowdMed are perfectly positioned to take advantage.”
While most people would endorse the idea of well-trained med students, not everyone wants to turn his body into a textbook. Azzam likens CrowdMed to a teaching hospital, where patients accept care from physicians with less experience partly because that’s the only way for young doctors to learn. “In any kind of educational endeavor in a health-care setting, there has to be a balance,” he says.
To use the system, a patient uploads information about his condition online and sets a reward to be given to anyone who can come up with a diagnosis that can be carried back to a physician. The platform itself is a sort of prediction market, based on algorithms that seek common threads among the opinions voiced. On average the bounties range from $165 to $300, split between everyone who gives the right answer; CrowdMed itself takes 10 percent. About 170 people have gone through the 90-day process so far, with 80 percent getting what they saw as a useful diagnosis.
That places the company’s total revenue in the low thousands of dollars, and the platform clearly needs more mysteries and more detectives before it becomes a viable business. When asked about growth, Heyman gives an outlandish goal typical of a startup entrepreneur: He says that he wants millions of patients using this every year. The real windfall would come if he can convince insurance companies to foot the bill for their customers to be diagnosed by a range of matriculating medical students, retired doctors, and armchair physicians.
It’s a vision that falls in line with one of the current beliefs of the tech industry: that the right answers will inevitably bubble up if you can just collect enough opinions. And while it’s not clear that this actually gives results that on average are better than what you’d get from a few visits to specialists, it could theoretically attract insurance companies by being far cheaper.
Rip Heminway became a believer soon after he had a hip replacement. At 48 he began suffering from cold sweats and swollen lymph nodes, and specialists were baffled by his symptoms. He uploaded his medical history to CrowdMed but continued to see doctors, and was given conflicting advice either to have another surgery or to take antibiotics. As he was weighing his options, Heminway received a report from CrowdMed: The most common diagnosis on the website held that his infection was caused by the prosthetic, meaning no surgery was needed.
“I got to the point of scheduling the surgery, but I backed off when I got this news,” he says. Of course, his doctor had reached the same conclusion, and while Heminway turned over the CrowdMed report, he’s not sure what his doctor did with it. “I think he looked at it with a little sideways glance. What he did with it I have no idea,” he says. But Heminway thinks the peace of mind was worth the $200 he paid, even if the report ended up in his doctor’s trash can: “It was for me, not for him.”