Meh, to each his own. I don't do it unless it's a patient who is there for a psych reason, such as OD, and I'm evaluating for why it happened and appropriate dispo. But if a patient is on a medical ward for, say, COPD, I don't tend to make a diagnosis because I'm getting only a glimpse into the patient's personality while the patient is under a great deal of physical and emotional stress. My view is limited and in that setting, it's tough to differentiate between borderline personality or impaired coping (and the two are not the same). The only caveat to that is if I suspect BPD and it is affecting medical care by putting the patient at risk while in the hospital.
By definition, BPD must be present in "a variety of contexts," which is difficult to observe in a hospital setting. Additionally, a key component of the diagnosis is the instability of interpersonal relationships. Sure, you can get collateral, but who's sitting on the phone on an inpatient medical unit trying to figure out how many failed relationships/friendships the patient had, unless it directly affects care/plan as in the patient with the OD? Most non-mental health professionals are notoriously bad at diagnosing BPD, which is why it's so widely diagnosed (in many cases, erroneously, similar to all the kids walking around with diagnoses of bipolar disorder). Once it's diagnosed by anyone, it is carried over in the chart repeatedly and no one seems to bother questioning it. This sucks for the patient because, as I've seen happen, most subjective complaints are then attributed to BPD, despite the fact that medical illness is likely contributing significantly to the emotional/affective dysregulation you see on a medical ward. The psych ward is obviously different.
If I'm on a med/surg floor and suspect borderline traits AND it's affecting medical care, I will do some education with the medical team, but failing that, I often refrain from saying borderline traits because as I've learned there are a number of providers who don't know what that means and think I'm making the diagnosis of BPD. I simply describe the traits instead: "patient tends to become dysregulated when family leaves or is about to leave" or "patient tends to become angry when receiving disappointing news" or "patient shows signs of impulsivity, so would monitor..." They don't need the label to get a clear picture of the patient's tendencies and how it can affect the medical admission. I then offer recommendations to combat these tendencies and may include a referral or option for psych eval upon medical discharge if I think the patient could benefit from it.