How big is the room?
How loud is the background noise?
Are the speakers itinerant or fixed?
This could be challenging to pick-up the patient's words at all times, considering that they are likely to have considerable variation in dynamics and tone.
I assume that that while the recording is for the good and possibly with the knowledge of the patient, that the device itself would still need to be covert, is that correct?
At least 2 microphones with AGC and a directional pickup pattern that is as tight as the room layout allows is to be expected.
Microphones, just like lenses, have angles of 'views', even if the boundaries are not as clearly defined. For example, using a sensitive omni-directional microphone in this case would probably be a disaster. A cardioid or a shotgun pattern, if feasible would be better.