Discussions of care preferences, such as wishes for cardiopulmonary resuscitation or invasive procedures, are of critical importance at the time of hospital admission ‐ a time when patients are sickest and plans for hospital care are being formulated. However, there are few recent data to describe patient factors asssociated with whether or not doctors document a discussion of wishes for care with their patients.


We analyzed data from the Multicenter Hospitalist Study, a prospective trial of patients admitted to general medical services at 6 academic medical centers between July 1 2001 and June 30 2003. Information regarding the presence of a care discussion (CD) in the first 24 hours of admission was collected from chart notations. Notations by physicians or nurses such as ‘DNR/DNI’ or ‘Full Code’ would not have counted as a CD, whereas notations such as ‘We discussed code status and patient continues to wish full code’ would have been considered a CD.

Summary of Results:

Of 33,638 patients in the overall MCH, 17,097 patients (or their proxies) completed an intake interview and had complete chart data; 1776 of these patients (10.0%) had a documented code discussion at admission. In unadjusted analyses, median age of patients having a care discussion was higher (69 years vs. 57, p < 0.001), and these patients were less likely to be African‐American (30.5% vs. 15.8%, p < 0.001). Patients who had cancer (11.4% of pts with no CD vs. 19.8% pts with CD, p < 0.001), were widowed, had a preexisting wish to be DNR/DNI (1.7% vs. 20.8%, p < 0.001) or other wishes (e.g. no ICU care, 0.9% vs. 6.5%, p < 0.001) were more likely to have a CD. After adjusting for patient age, demographics, site of enrollment, patient functional status, chronic illness, and care in an ICU in multivariable models, patients with cancer (Adjusted Odds (AOR) 1.3, 95% CI 1.1‐1.5), those who were unable to make decisions themselves (AOR 1.7, 95%CI 1.4‐2.0) and those who had a designated decision maker (AOR 1.8, 95%CI 1.6‐2.1) were more likely to have a CD, as were patients where the team spoke with the patient's primary care physician (AOR 1.7, 95% CI 1.3‐2.2). Compared to those with Medicare, patients who were uninsured (AOR 0.5, 95% CI 0.3 to 0.8) or who had Medicaid insurance (AOR 0.7, 95CI 0.6‐0.9) were less likely to have a CD. Patient race was not retained in final multivariable models.

Statement of Conclusions:

Even after taking prognostic information into account, the likelihood of a care discussion is modified by factors which increase physicians' needs to understand patients' wishes. However, how and why insurance type plays a role should be the subject of further investigations.

Author Disclosure Block:

A.D. Auerbach, None; S. Pantilat, None; R. Katz, None; J. Schnipper, None; P. Kaboli, None; T. Wetterneck, None; D. Gonzales, None; J. Zhang, None; R. Bernacki, None; D. Meltzer, None.