There's been lots of talk of Health 2.0, but this is the first thing I've seen that might actually do something useful with the gobs of data that's collected. Even talk of using national health record data (e.g. having the anonymized charts of every patient nationwide vs. currently only having administrative data from insurance companies) for serious research has been scarce. The one caveat with this is that if the data comes from a biased sample or if they are confounded in the bajillions of ways epidemiologists have discovered things go wrong, it could easily be worse than giving no patients any information, as the conclusions patients come to would be false. Unfortunately, numbers tend to give a veneer of trustworthiness to almost any evidence.