More on the potential of crowdsourcing

Maybe the Office of Rare Diseases at the NIH could do with some crowdsourcing.  A place for patients to go when all else has failed, this office is a repository for unsolved medical mysteries.  Even with a record of more unsolved than solved cases, this clinic is a source of hope for many, because they're in the business of mining all the information they can on rare conditions.

The patients who go there are desperately hoping for answers, but the clinicians who try desperately to find them are facing a major limitation of science.  Science is about causal regularity in Nature, and does rather poorly with unique observations.  Repetition and prediction are gold-standard criteria for understanding.  Yet these cases are essentially unique!

A story on the CNN website tells of two patients with undiagnosed diseases who spent five days at the NIH, the recipients of exhaustive testing for every possible thing that could be wrong.  The problem with rare diseases, of course, is that they can be so rare that only one case has ever been seen.  Or, at least rare enough that there just isn't enough data in the medical literature, or collective experience with the disease, to facilitate a diagnosis.  

Here's where crowdsourcing could be -- and often is -- useful, as we suggested about last week.  The huge numbers of people who populate the web, and perhaps intensely when they have a medical dilemma that concerns them, provides a potentially unique way in which repetition can be ascertained.  People seem to find each other, and relate their stories.   Of course, these are informal and not rigorous by normal clinical standards, but experience shows that sense can be made of the data, at least to some useful extent.

Doctors try as best they can to pigeon-hole their patients into categories for which there are treatments.  Most doctors work with small sets of patients, or in small communities, where they simply can't be expected to see very rare traits more than once or twice, if at all, and then not at the same time.

Taking advantage of huge population samples, assembling rare information to reveal patterns.  Crowdsourcing.