But we don't know if these are false positives. The connection between all those examples doesn't have to be the suggested link. E.g. the patients could be suggested because they have a common friend.
IIRC, connections like those were often explainable by addressbook data. Please assumed Facebook didn't have access to that when it actually did (I'm pretty careful about this kind of stuff and found mine had been slurped at some point, for instance). It's not inconceivable that a psychiatrist and patient may have exchanged emails or phone numbers.