That is, if health data were anonymized, and was done right, and was made unable to be correlated with any other data, it likely would be sufficient. It's when you start allowing it to be correlated with personally identifiable things that it ceases to be anonymous.
That is, sure, let's take a case where you have a super rare genetic disorder. That, combined with the time in 2005 where you broke your leg, is sufficient to distinguish you from every other person in the country. In short, you have a unique health profile.
So what? Unless there is further information, that can't be traced to you. As an example, it's when we start saying "Ah, and the person is receiving treatment at (facility)" that we now know where you live. It's when we start correlating it with usernames that we start getting an internet trail. It's when we start correlating those with forum profiles that we get a real name, and now we know who you are.
The only other way someone could match that profile with you, is to have access to the profile, and to know you personally. Otherwise it links nowhere.
I agree the risk is huge; people don't do it right. But anonymous health profiles are -not- in and of themselves dangerous; it's when details linking them to further information leaks out that it's a problem.
But, pragmatically, while yes it would be incredibly hard...has anyone here read the rights they're signing away when they go to the doctor? Does everyone here trust every system a doctor uses, every system a health insurer uses, and every system used by marketers and researchers that the feds -do- allow to have access to this data? The real risk of Google would be that they could correlate it with so many other things about you; but the health insurers still have your medical history combined with all your PII.