We have also been thinking about using PRIME-MD. I am looking at the SF-36 and PRIME-MD as I write. I know the SF-36 is not necessarily considered a screen, but it is still useful to compare.
The SF-36 quantifies items (e.g. "All the time", "Most of time" etc. to describe how the person felt in the last month. PRIME-MD uses a Yes/No response. Both can be completed as self-report, but I wonder if a person who is forced to chose between Yes or No for "Feeling down or depressed" will chose No and potentially minimize his or her feelings of depression, even if it is present. I work with cardiac patients and in my experience, a great number of them would be likely to deny or minimize anxiety and depression. If the questionnaire is administered by a mental health professional, most clinicians would be able to elicit more response. But it were administered by the physician, not all docs would be equally adept at assessing beyond the yes/no response. However, in fairness to PRIME-MD, if there is a corresponding manual, I haven't read it and it might do a great job of training professionals how to administer the screen.
On the other hand, PRIME-MD gets at some things that the SF-36 doesn't get at, such as substance abuse, eating disorders, sexual problems, panic attacks, constipation/loose bowels/diarrhea etc. These seem like sensitive items where patients will not volunteer the information unless directly asked. And it seems relatively easy to ask the questions. I think PRIME-MD might be my choice for a screen because I know that our screeners will know how to follow-up with "Yes" responses. However, I think I would not use a self-report version because some patients deny symptoms. Hopefully a clinician adminstered questionnaire will catch some of those responses.
Whether the screen is for a research protocol or is used in routine clinical practice would make a difference in deciding what to use. I'm also interested in hearing from others who use this or other measures as screens.