The encyclopedia · R&D & Science · Product decision · 2013–2015
CrowdMed let a crowd bet on the cause of an undiagnosed illness.
CrowdMed pooled many online 'case solvers' and ranked diagnoses by prediction-market points to help hard-to-diagnose patients.
CrowdMed
the move
Diagnostic error is common and some patients with rare or unusual conditions go years and see many doctors without a cause, so the system of second opinions is gated by referrals and cost.
CrowdMed let patients post their anonymous case, let anyone sign up as a solver and discuss it, then scored the suggested diagnoses by a prediction-market algorithm so the patient got a ranked list of possibilities.
An independent evaluation of 397 completed cases found patients had seen a median of five physicians and spent about $10,000, while many who used the platform said it moved them closer to a correct diagnosis and some reported reduced expenses.
why it works
- A prediction-market algorithm aggregates many independent beliefs rather than trusting one solver
- Open discussion lets a tentative lead be confirmed or rejected by the rest of the crowd
- Anyone can contribute, so expertise is not gated by credentials or referral
- A ranked report gives the patient something concrete to take to a clinician
what transfers
When each mind is fallible but the problem is hard, a crowd that bets on answers and reveals its reasoning can surface possibilities a single clinician may miss.
what came after
CrowdMed became a documented example of crowdsourced diagnosis, studied in the peer-reviewed literature as a complement to clinical care. The evaluation stressed that long-term validation was still needed, since a reported helpful suggestion is not the same as a confirmed correct diagnosis.
references
spotted an error? The archive wants to know.