Table 3.

Associations Between Patients’ Treatment Preferences, Treatment Site, Communication, and End-of-Life Care

	Intensive Carea				Hospice Servicesb				Other EOL Carec
Predictors	OR	95% CI	AORd	95% CI	OR	95% CI	AORd	95% CI	OR	95% CI	AORd	95% CI
Preference for  life-extending care	4.42	1.74-11.24e	2.91	1.09-7.72f	0.38	0.23-0.64g	0.45	0.26-0.78e	1.74	1.02-2.96f	– h	– h
Treatment site
Yale Cancer Center	3.14	1.16-8.47f	– h	– h	0.52	0.29-0.92f	– h	– h	1.59	0.87-2.92	– h	– h
New Hampshire     Oncology-Hematology	– i	– i	– i	– i	1.66	0.85-3.24	– h	– h	0.81	0.41-1.60	– h	– h
Simmons Comprehensive     Cancer Center	2.30	0.58-9.16	– h	– h	0.64	0.29-1.38	– h	– h	1.44	0.64-3.24	– h	– h
Parkland Hospital	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
EOL discussion	0.17	0.04-0.73f	0.21	0.05-0.94f	2.36	1.41-3.93g	– h	– h	0.58	0.34-0.98f	– h	– h
Understands illness  is terminal	0.57	0.20-1.63	– h	– h	2.33	1.38-3.91g	2.08	1.23-3.53e	0.46	0.26-0.79e	0.51	0.28-0.91f
Age	0.97	0.93-1.01	– h	– h	0.69	0.36-1.32	– h	– h	0.99	0.98-1.02	– h	– h
Black race	3.84	1.51-9.77e	3.29	1.23-8.77f	0.72	0.40-1.31	– h	– h	0.82	0.43-1.54	– h	– h

EOL indicates end of life; OR, odds ratio; CI, confidence interval; AOR, adjusted odds ratio; Ref, referent category.

^aDefined as the receipt of cardiopulmonary resuscitation and/or mechanical ventilation followed by death in an intensive care unit.

^bIncludes patients who died either at home while receiving hospice services or in an inpatient hospice (n=40/191).

^cOther EOL care included inpatient hospital care (n=58; 5 patients received care in an intensive care unit and then were transferred to a hospital floor before death), home care without hospice services (n=20), and inpatient nursing home care (n=13).

^dConfounds that were examined included age, sex, race, treatment site, terminal illness acknowledgement, and EOL discussion. Confounds were included in the models if they were associated (P <.20) with both a preference for life-extending care and any EOL outcome and were retained in the multivariate models if significant (P <.05). The model for estimating the association between patients’ preferences and intensive care included Yale Cancer Center, EOL discussion with physician, and black race/ethnicity. The model for estimating the association between patients’ preferences and hospice services included Yale Cancer Center, EOL discussion with physician, and terminal illness acknowledgement. The model for estimating the association between patients’ preferences and other EOL care included EOL discussion and terminal illness acknowledgement.

^eP <.01 (significant).

^fP <.05 (significant).

^gP <.001 (significant).

^hNot significant.

ⁱNo events at this site.