My friend Gary has an essay on the NY Times web site about using stories about his psychotherapy patients in his books. It turns out that even though he disguises their identities every which way, the real problem isn't that a patient's confidentiality might be betrayed, it's that the patient might recognize herself. Using the therapeutic relationship as a means to an end feels exploitive to the patient, even if nobody could possibly recognize her.
I don't exactly have the same problem -- when I interview people or record their medical encounters, they know that I'm doing it for research and some version of their story may appear in print, without sufficient detail to identify them. But I still have to be very careful. There's this, for example, which I have published:
Well, I went to three actually detoxes. I said - 'cause when I got out there, they kept on switching me from one to another. I wasn’t - I was quitting and, well, I got kicked out of one of them 'cause I had sex in a closet with a nun . . . . So they said, "You gotta get out." But I went to three detoxes. I don't want to get off.
(She did eventually quit drinking and she's doing okay now.) No matter what I did to disguise her identity, there must be people who are familiar with this incident, and would know who this likely was if they read the paper. Fortunately, it's very unlikely they will ever read it. But still.
We always exploit our research subjects to some extent. I get the grants and the salary and whatever prestige comes with publishing. They get 20 bucks for an hour of their time. I get to decide what parts of what they tell me matter and what parts don't, and how to present and interpret their experience. I do my best to give them a voice but in the end, it's really mine after all.