Demographic and Clinical Determinants of Having Do Not Resuscitate Orders in the Intensive Care Unit of a Comprehensive Cancer Center

Abstract

To understand the needs of patients and family members as physicians communicate their expectations about patients admitted to the intensive care unit (ICU), we evaluated the demographic and clinical determinants of having a Do Not Resuscitate (DNR) order for adults with cancer. Patients included were admitted from June 16, 2008–August 16, 2008, to the ICU in a comprehensive cancer center. We conducted a prospective chart review and collected data on patient demographics, length of stay, advance directives, clinical characteristics, and DNR orders. A total of 362 patients met the inclusion criteria; only 15.2% had DNR orders before ICU discharge. In the multivariate analysis, we found that medical admission was an independent predictor of having a DNR order during the ICU stay (odds ratio = 3.65; 95% confidence interval, 1.44–9.28); we also found a significant two-way interaction between race/ethnicity and type of admission (medical vs. surgical) with having a DNR order (p = .04). Although medical admissions were associated with significantly more DNR orders than were surgical admissions, we observed that the subgroup of non-white patients admitted for medical reasons was significantly less likely to have DNR orders. This finding could reflect different preferences for aggressive care by race/ethnicity in patients with cancer, and deserves further investigation.

Introduction

Many deaths occurring in the intensive care unit (ICU) are associated with decisions to withhold or withdraw medical care.¹ Although critical care should be reserved for patients who would benefit from it, most patients rushed to the ICU are in unstable medical conditions, and their chances of survival on admission are unknown.² For those who do not respond well to therapy, including cancer patients in need of critical care, the unintended result is often futile medical care unless decisions to withhold or withdraw care limit the artificial extension of life. A previous study showed that 32% of deaths in a cancer center that occurred over one year were associated with the decision to withdraw medical care and provide comfort care for patients whose disease had progressed toward a terminal stage, despite the fact that therapy had been provided on admission, with curative intent.³

The Do Not Resuscitate (DNR) order means that a patient intends to forgo cardiopulmonary resuscitation (CPR) in the event of cardiopulmonary arrest.⁴ The DNR order is specific to cardiopulmonary arrest; thus, withholding other therapies is not covered by the order. Some cancer patients who have signed DNR orders are subsequently admitted to the ICU, but DNR orders are usually signed during the ICU stay – frequently after therapy has failed, multiple organ failure is evident, and hope for recovery has vanished.

The characteristics of patients in the general population who eventually have a DNR order have been widely described, which improved researchers' understanding of the process of obtaining this document, helped uncover the needs of family members as physicians expressed their expectations about the ICU patient, helped guide decisions about whether to limit or withdraw care after life has been artificially extended, and reduced barriers between family members and physicians.^5–7 Nevertheless, the characteristics of those designated as DNR have not been described well in cancer patients; thus, little is known about the process of obtaining these orders in this group.

Our objectives were to (1) describe a group of critically ill cancer patients admitted to a high-volume ICU in a comprehensive cancer center who had DNR orders during their ICU stays and (2) investigate the medical and demographic factors associated with these orders. We measured and evaluated factors obtained before patients were discharged (dead or alive) from the ICU.

Methods

Data sources and collection

With Institutional Review Board approval, we reviewed data from electronic medical records, ICU and operating room databases, and the Tumor Registry for all adult cancer patients admitted to our 53-bed ICU from June 16, 2008–August 16, 2008. We excluded patients younger than 18 years because the decision-making process for obtaining DNR orders for that age group may differ from the process followed for adult patients. Patients whose race/ethnicity was unknown (i.e., when patients chose to not disclose this information) were excluded. In patients with more than one ICU admission during the same hospital stay, we selected the first admission only.

We obtained patients' demographic characteristics (e.g., age, sex/gender, marital status, and race/ethnicity), advance planning data, and clinical data. Marital status was categorized as married or not married. Race/ethnicity categories were white or non-white; African Americans, Hispanics, Asians, American Indians, and patients from other groups were classified as non-white. We obtained self-reported data on race/ethnicity from admission forms completed by our patients.

Advance planning data included signed documentation such as advance directives (ADs), power of attorney for health care, or out-of-hospital DNR orders, and included the dates the forms were signed.

Clinical data included type of ICU admission, type of cancer, stage of cancer at diagnosis, length of stay (LOS) in the hospital before ICU admission, LOS in the ICU, LOS in the ICU to DNR order, risk of mortality score on admission, reason for referral to the ICU, use of invasive mechanical ventilation, and discharge status from the ICU.

Our hospital ICU has two types of admissions: surgical and medical. Surgical admissions are typically referred to the ICU from the operating room or post-anesthesia recovery room, but occasionally, patients are sent to the floor directly from the post-anesthesia recovery room and later admitted to the ICU. We reviewed the operating room database to assess the date and type of surgical procedure for surgical admissions to the ICU. The type of cancer was categorized as hematologic (leukemia or lymphoma) or nonhematologic. We categorized the stage of the cancer at diagnosis as post-radiation therapy with no evidence of disease, local/regional, distant, or unstaged/unknown. Given the low frequency of DNR orders in patients in the category of local disease, we decided to merge categories of local and regional cancer.

LOS in the hospital before ICU admission was measured starting from the date of hospital admission. LOS in the ICU was measured starting from the day of ICU admission; and LOS in the ICU to DNR order was determined by date of designation as DNR minus date of admission to the ICU. The risk of mortality score on admission was evaluated by using the Sequential Organ Failure Assessment (SOFA) score, described by Peres Bota et al.⁸ Scores ranged from 0 to 24, with higher numbers representing higher risk of mortality. SOFA has been used previously to assess severity of illness in patients with cancer.⁹

The reason for referral to the ICU was either (1) patients with respiratory failure or (2) other reasons for referral (including patients admitted for hypotension, surveillance and monitoring, or renal failure and in need of dialysis). Respiratory failure included: Type I hypoxemic respiratory failure, defined as having a PaO2 <60 mmHg on room air; and Type 2, or hypercapnic respiratory failure, defined as having a pH <7.35 and PaCO₂ >45 mmHg. The use of invasive mechanical ventilation during the ICU stay was recorded as “yes” or “no” and, when this was used, we recorded the procedure's start and termination dates. The discharge status from the ICU was “dead” or “alive.”

Study outcomes

The main study outcome was having a DNR order or not. We compared two scenarios: (1) Patients had DNR orders on arrival to the ICU or the orders were written at any time during the ICU stay; (2) Patients lacked DNR orders during their ICU stay.

Statistical analysis

We used frequencies and measures of central tendency (means and medians) to summarize patient characteristics. We analyzed continuous and categorical variables for associations having a DNR order by using the student's t test and chi-square test or Fisher's exact test (when cells of the contingency tables had fewer than 5 cases). All tests were two-tailed, and p ≤ .05 was considered statistically significant.

We next constructed a multivariate logistic regression model to analyze factors associated with DNR orders. To include all factors potentially associated with study outcome, we entered demographic and clinical factors with a p value of ≤.20 in the univariate analysis into the multivariate model. Age was entered a priori because previous research showed an association between old age and having DNR orders.^2,10,11 Discharge status from the ICU was excluded from the multivariate model because the aim of the study was to evaluate determinants of having a DNR order before discharge from the ICU. We checked our multivariate model for collinearity among the variables and reported the goodness-of-fit of the model and the variance explained by the Nagelkerke R square test. We evaluated whether the effect of a covariate depended on the level of another covariate by determining the presence of three pairs of relevant two-way interactions. These interactions were race/ethnicity and type of admission, race/ethnicity and SOFA score; and age and type of admission. As an example, the equation used to evaluate the interaction between race/ethnicity and type of admission was logit(p) = log(p/(1-p)) = β₀ + β₁*race/ethnicity + β₂*type of admission + β₃*race/ethnicity*type of admission + ɛ. We used SPSS version 15.0 for Windows 2006 (SPSS, Inc., Chicago, IL) for all statistical analyses.

Results

During the study period, 390 patients were admitted to the ICU; after excluding 19 (4.9%) who were younger than 18 years and 9 (2.3%) with unknown race/ethnicity, 362 patients were included in the study. Only 55 patients (15.2%) had DNR orders during the study period: 5 (1.4%) had the orders before ICU admission, 15 (4.1%) during the first 24 hours after admission, and 35 (9.7%) between day 2 and ICU discharge (Table 1). Overall, the mortality rate in the ICU was 12.4%, with the surgical admissions rate (5.4%) lower than that of medical admissions (21.3%) (p < .001). Among all patients with DNR orders during the study, 15 (27.3%) survived and were discharged from the unit, 3 (5.5%) of whom had the orders written before admission.

Table 1.

Univariate Analysis of Factors Associated with Having Dnr Orders in Adult Patients with Cancer Who Were Admitted to the ICU In a Comprehensive Cancer Center

	No. (%) of all patients	No. (%) of patients with DNR orders before ICU discharge	No. (%) of patients with no DNR orders	p value
Total	362 (100)	55 (15.2)	307 (84.8)	NA
Demographic factors
Mean (median) age on admission, years	57.0 (59)	58.2 (58)	56.8 (59)	.51
Sex/Gender				.29
Male	207 (57.2)	35 (63.6)	172 (56.0)
Female	155 (42.8)	20 (36.4)	135 (44.0)
Marital status				.74
Married	244 (67.4)	36 (65.5)	208 (67.8)
Not married	118 (32.6)	19 (34.5)	99 (32.2)
Race/ethnicity				.10
White	275 (76.0)	37 (67.3)	238 (77.5)
Non-white	87 (24.0)	18 (32.7)	69 (22.5)
Advanced planning documentation
Had advance planning data (i.e., advance directives, power of attorney for health care, and/or out-of-hospital DNR orders)				.19
Yes	117 (32.3)	22 (40.0)	95 (30.9)
No	245 (67.7)	33 (60.0)	212 (69.1)
Clinical factors
Type of ICU of admission				≤.001
Medical	160 (44.2)	42 (76.4)	118 (38.4)
Surgical	202 (55.8)	13 (23.6)	189 (61.6)
Type of cancer				.81
Hematologic	49 (13.5)	8 (14.5)	41 (13.4)
Nonhematologic	313 (86.5)	47 (85.5)	266 (86.6)
Stage of cancer at diagnosis				=.03
Post-radiation therapy (no evidence of disease)	66 (20.2)	10 (18.5)	56 (20.5)
Local/regional	98 (30.0)	8 (14.8)	90 (33.0)
Distant	111 (33.9)	23 (42.6)	88 (32.2)
Unstaged/unknown	52 (15.9)	13 (24.1)	39 (14.3)
Mean (median) length of stay before ICU admission, days	4.1 (0)	8.7 (2)	3.3 (0)	.02
Mean (median) length of stay in ICU, days	4.5 (2)	7.4 (3)	4.0 (2)	≤.01
Mean (median) SOFA score on admission	5.8 (6)	7.6 (7)	5.5 (5)	≤.01
Reason for referral to ICU^*				.02
Respiratory failure	163 (45.0)	33 (60.0)	130 (42.3)
Other	199 (55.0)	22 (40.0)	177 (57.7)
Use of invasive mechanical ventilation				.04
Yes	164 (45.3)	32 (58.2)	132 (43.0)
No	198 (54.7)	23 (41.8)	175 (57.0)

DNR = do not resuscitate; ICU = intensive care unit; NA = not applicable.

= May include other reasons for referral to the unit in addition to respiratory failure.

Percentages may not equal 100%, due to rounding.

About one-third (32.3%) of our patients had signed advance-planning documents before admission. Compared with non-white patients, significantly more white patients had signed these documents (10.3% vs. 89.7%, respectively; p < .001). This association persisted after controlling for age, sex/gender, and type of cancer. The mean age of patients on admission was 57 years (range, 18–90 years). Only 11.9% of patients were 75 years or older, and the percentage of DNR orders in this group (18.6%) was higher than in patients younger than 75 years (14.7%), but the difference was not statistically significant.

Although 13.5% patients had hematological cancer, skin cancer was the most frequent (12.7%) type of cancer among those with nonhematological cancer. Among patients with skin cancer, we found only 3 (6%) had local disease and had been referred to the unit after surgical removal of the cancer.

The mean hospital LOS before ICU admission was significantly longer among patients with DNR orders (8.7 days) than among those without DNR orders (3.3 days) (p = .02). This difference persisted after controlling for having or not having AD documents signed before ICU admission. We found no association between LOS before ICU admission and the time when the DNR order was signed, before or after ICU admission.

The mean LOS in the ICU was 4.5 days (median 2, range 0–42 days). Compared with patients without DNR orders, patients with orders had a significantly longer LOS in the ICU, 4.0 vs. 7.4 days, respectively (p < .01). This difference disappeared after controlling for SOFA score on admission and status on discharge (alive or dead). The mean LOS after ICU admission until designation as DNR was 2.7 days for patients without ADs and 1.2 days for those having documents signed before ICU admission, but this difference was not significant. The mean number of days from the time DNR orders were obtained until discharge was 6.2 (median 1, range 0–105 days); however, for the subgroup of patients whose DNR orders were signed after admission to the unit, this period was only 3.7 days (median 1, range 0–38 days).

The use of invasive mechanical ventilation was associated with having DNR orders. We found that compared with patients who did not have invasive mechanical ventilation, patients with ventilatory support were significantly more likely to have DNR orders, 41.8% vs. 58.2% (p = .04); nevertheless, this association disappeared when adjusting for hospital LOS before ICU admission. Furthermore, in the subgroup with ventilation support, the mean (median) duration of ventilation was 5.9 (2) days; among medical admissions, the mean (median) duration of ventilation was 10.4 (6) days and among surgical admissions was only 3.3 (1) days (p < .001).

We found a statistically significant association between type of admission and number of DNR orders. The percentage of medical admissions with DNR orders (76.4%) was higher than the percentage of surgical admissions with DNR orders (23.6%) (p < .001). This association persisted after controlling for demographic factors (e.g., age and race/ethnicity) and medical factors (e.g., stage of disease at diagnosis, survival, and SOFA score).

In the univariate analysis, only clinical factors were significantly associated with having DNR orders. These factors included type of admission, stage of cancer at diagnosis, respiratory failure as the main reason for ICU referral, use of invasive mechanical ventilation, SOFA score on admission, and hospital LOS before ICU admission. Other factors in the univariate analyses potentially associated (p < .20) with having a DNR order were race/ethnicity (p = .10) and ADs (p = .19) (Table 1). All of these factors, as well as age, were entered into a multivariate logistic regression model.

In the logistic regression model, the type of admission was significantly associated with having a DNR order. Medical admissions were 3.65 (95% confidence interval, 1.44–9.28) times more likely to be designated DNR than were surgical admissions (p ≤ .01). Race/ethnicity (p = .05) approached statistical significance. The variance explained by the model measured by the Nagelkerke R square was .26. The goodness-of-fit of the model had chi-square = 7.9, df = 8, and p = .44. (Table 2).

Table 2.

Multivariate Analysis of Factors Associated with DNR Orders in Adult Patients with Cancer in Need of Critical Care

Variable name	Beta	Odds ratio (95% CI)	p value
Type of ICU of admission			≤.01
Medical	1.30	3.65 (1.44–9.28)
Surgical		1
Stage of cancer at diagnosis			.46
Post radiation therapy (no evidence of disease)		1
Local/regional	.01	1.01 (.27, 3.74)	.99
Distant	.45	1.56 (.50, 4.93)	.45
Unstaged/unknown	.86	2.37 (.69, 8.15)	.17
Reason for referral to ICU			.20
Respiratory failure	.58	1.78 (.74–4.28)
Other		1
Use of invasive mechanical ventilation			.56
Yes	–.26	.77 (.30–1.94)
No		1
SOFA score on admission	.07	1.07 (.97–1.18)	.20
Length of stay before ICU admission, days	.04	1.04 (.99–1.08)	.07
Race/ethnicity			.05
White	.80	2.22 (.99–4.99)
Non-white		1
Had advance planning data (i.e., advance directives, power of attorney for health care, and/or out-of-hospital DNR orders)			.08
Yes	.72	2.05 (.91–4.62)
No		1
Age	.02	1.02 (.99–1.05)	.24

CI = confidence interval; DNR = Do Not Resuscitate; ICU = intensive care unit; SOFA = Sequential Organ Failure Assessment.

Hosmer and Lemeshow test Chi-square = 7.93, df = 8, and p = .44. Nagelkerke R square = .27.

There was significant (p = .04) interaction between race/ethnicity and type of admission, suggesting that non-white patients with cancer who were admitted to the medical unit (compared with their white counterparts) were 83% less likely to be designated as DNR (OR = .17, 95% confidence interval, .03–.92). We also evaluated other two-way interactions: race/ethnicity and SOFA score, and age and type of admission, but found no statistically significant associations.

Discussion

In this study, we established demographic and clinical factors associated with DNR directives in critically ill cancer patients admitted to the ICU of a comprehensive cancer center. Our two most significant findings were that (1) admittance to the ICU as a medical patient was an independent predictor of having a DNR order and (2) non-white race/ethnicity was significantly associated with not having been designated as DNR among patients admitted to the same ICU.

Very few patients had DNR orders on ICU admission, despite the presence of critical illness, relatively high proportion of ADs, underlying diagnoses of cancer, and estimated negligible chance of survival associated with cardiac arrest in this population. Previous researchers have assessed percentages of DNR orders in the ICU. Boyd et al.¹⁰ showed that the percentage of patients designated as DNR during the first 24 hours after admission was 5.2% but, by discharge, was 11.4%. More recently, Eachempati et al.¹¹ showed that 12% of patients were designated as DNR by discharge. We showed that 5.5% of patients were designated as DNR before or during the first day of admission, corresponding to the Boyd et al. rates.¹⁰ We also found that the percentage of orders written before discharge (15.2%) was slightly higher than that reported by Boyd et al.¹⁰ and Eachempati et al.,¹¹ possibly because their study populations included patients with diagnoses other than cancer who may have had lower severity of illness on admission or fewer coexistent disorders.

Significant variation in DNR rates has been observed across hospitals. In a recent study that included 68 centers and 6,765 trauma patients, a DNR rate as high as 57% was reported; in about 72% of participating units, however, DNR rates were less than 11%.¹² Only 15% of the units had higher rates than ours.

The Level of Care Study Investigators and the Canadian Critical Care Trials Group found that, in one-third of patients who were mechanically ventilated for longer than 48 hours, DNR directives were established during their ICU stays. In our series, only 19.5% of intubated and mechanically ventilated patients had DNR orders, possibly because of our aggressive use of noninvasive ventilation as first-line ventilatory support in respiratory failure, leading to the designation of DNR before mechanical ventilation is indicated.

We found that medical admissions were more likely to have DNR orders than were surgical admissions, possibly due to systematic bias introduced by selecting patients for nonemergency surgical procedures; also, surgical patients tend to have more localized disease than do patients with cancer who are not candidates for surgery. Moreover, when balancing the benefits and risks of an upcoming surgical procedure, survival is expected, while, in contrast, patients admitted to the medical unit typically have more-advanced disease and could have developed medical complications secondary to chemotherapy or radiotherapy.

Despite others' findings,^10,11,13 we did not find age to be associated with patient designation as DNR, possibly because our patients were, on average, younger than those in the other studies. Our patients' mean age was 57 years and only 11.9% were 75 or older; patients in the Eachempati et al. study¹¹ were, on average, 63 years old, and 53% of patients in the Boyd et al. study¹⁰ were 75 or older.

Our finding that non-white patients admitted to the medical unit were less likely than white patients to have DNR orders corroborates results from other studies. Hofmann et al.⁶ studied critically ill patients, including patients with cancer, and found that, at study enrollment, black patients were 53% more willing than non-black patients to discuss preferences for CPR with physicians but did not do so. Barnato et al.¹⁴ found that blacks, compared with non-Hispanic whites, showed preference for life prolongation, and blacks and Hispanics were significantly more likely to prefer to die in hospitals than were non-Hispanic whites. This finding shows that minorities may perceive palliative care and DNR orders as unacceptable limitations of life-sustaining treatments. Cesta et al.³ showed that white patients were 2.5 times more likely than other ethnicities to have life-support therapy withdrawn, and Loggers et al.¹⁵ found that black patients were 3 times more likely than white patients to receive intensive end-of-life care.

Our study had limitations, including a small study sample from a single institution. However, although the single institution limits generalization of results, our findings were not significantly different from those of other researchers.^14–16 Our sample size limited our power to detect significant differences to only 70%, which could explain why race/ethnicity alone was not significant in the univariate analysis. We could not assess whether end-of-life discussions took place between patients, family members, and physicians before ICU admission because we could not interview all patients on admission, and this information was usually not in the medical records. Fostering such end-of-life discussions has helped reduce aggressive medical care near death in patients with advanced cancer.¹⁷ Indeed, we could infer that those who arrived at the ICU with DNR orders (1.4%) had discussed this topic earlier with their physicians. The details of these discussions, however, were out of the scope of our study.

In summary, the number of DNR orders for ICU patients in our comprehensive cancer center was similar to that found in other critically ill populations. We found the type of ICU admission to be an independent predictor of DNR directive decisions. Non-white patients admitted for medical reasons were significantly less likely than were white patients to have DNR orders, demonstrating a trend toward favoring more aggressive care (e.g., heroic interventions) at the end of life. This finding could reflect different preferences for aggressive care by race/ethnicity in patients with cancer and deserves further investigation of the motivations and beliefs that result in the lower percentages of DNR orders observed.

Footnotes

Acknowledgments

We thank the ICU staff members (especially Rose Jean Erfe, Christina Guajardo, and Mary Lou Warren) for their assistance in this project and Tamara Locke for reviewing this manuscript.

Author Disclosure Statement

No competing financial interests exist.

References

Sprung

, Cohen

, Sjokvist

, Baras

, Bulow

, Hovilehto

, Ledoux

, Lippert

, Maia

, Phelan

, Schobersberger

, Wennberg

, Woodcock

. End-of-life practices in European intensive care units: the Ethicus Study. JAMA, 2003; 290:790–797.

Janssen van Doorn

, Diltoer

, Spapen

. Do-not-resuscitate orders in the critically ill patient—an observational study with special emphasis on withholding of renal replacement therapy. Acta Clin Belg, 2008; 63:221–226.

Cesta

, Cardenas-Turanzas

, Wakefield

, Price

, Nates

. Life-supportive therapy withdrawal and length of stay in a large oncologic intensive care unit at the end of life. J Palliat Med, 2009; 12:713–718.

Cohen

, Lisker

, Eichorn

, Multz

, Silver

. The impact of do-not-resuscitate order on triage decisions to a medical intensive care unit. J Crit Care, 2009; 24:311–315.

Jayes

, Zimmerman

, Wagner

, Knaus

. Variations in the use of do-not-resuscitate orders in ICUS. Findings from a national study. Chest, 1996; 110:1332–1339.

Hofmann

, Wenger

, Davis

, Teno

, Connors

Jr. , Desbiens

, Lynn

, Phillips

. Patient preferences for communication with physicians about end-of-life decisions. SUPPORT Investigators. Study to Understand Prognoses and Preference for Outcomes and Risks of Treatment. Ann Intern Med, 1997; 127:1–12.

Kaldjian

, Erekson

, Haberle

, Curtis

, Shinkunas

, Cannon

, Forman-Hoffman

. Code status discussions and goals of care among hospitalized adults. J Med Ethics, 2009; 35:338–342.

Peres Bota

, Melot

, Lopes Ferreira

, Nguyen Ba

, Vincent

. The Multiple Organ Dysfunction Score (MODS) versus the Sequential Organ Failure Assessment (SOFA) score in outcome prediction. Intensive Care Med, 2002; 28:1619–1624.

Namendys-Silva

, Texcocano-Becerra

, Herrera-Gomez

. Application of the Sequential Organ Failure Assessment (SOFA) score to patients with cancer admitted to the intensive care unit. Am J Hosp Palliat Care, 2009; 26,5:341–346.

10.

Boyd

, Teres

, Rapoport

, Lemeshow

. The relationship between age and the use of DNR orders in critical care patients. Evidence for age discrimination. Arch Intern Med, 1996; 156:1821–1826.

11.

Eachempati

, Hydo

, Shou

, Barie

. Sex differences in creation of do-not-resuscitate orders for critically ill elderly patients following emergency surgery. J Trauma, 2006; 60:193–197discussion 197–198.

12.

Nathens

, Rivara

, Wang

, Mackenzie

, Jurkovich

. Variation in the rates of do not resuscitate orders after major trauma and the impact of intensive care unit environment. J Trauma, 2008; 64:81–88discussion 88–91.

13.

Sinuff

, Cook

, Rocker

, Griffith

, Walter

, Fisher

, Dodek

, Sjokvist

, McDonald

, Marshall

, Kraus

, Levy

, Lazar

, Guyatt

. DNR directives are established early in mechanically ventilated intensive care unit patients. Can J Anesth, 2004; 51:1034–1041.

14.

Barnato

, Anthony

, Skinner

, Gallagher

, Fisher

. Racial and ethnic differences in preferences for end-of-life treatment. J Gen Intern Med, 2009; 24:695–701.

15.

Loggers

, Maciejewski

, Paulk

, DeSanto-Madeya

, Nilsson

, Viswanath

, Wright

, Balboni

, Temel

, Stieglitz

, Block

, Prigerson

. Racial differences in predictors of intensive end-of-life care in advanced cancer patients. J Clin Oncol, 2009; 27:5496–5498.

16.

Blackhall

, Frank

, Murphy

, Michel

, Palmer

, Azen

. Ethnicity, attitudes towards life sustaining technology. Soc Sci Med, 1999; 48:1779–1789.

17.

Wright

, Zhang

, Ray

, Mack

, Trice

, Balboni

, Mitchell

, Jackson

, Block

, Maciejewski

, Prigerson

. Associations between end-of-life discussions, patient mental health, medical care near death, and caregiver bereavement adjustment. JAMA, 2008; 300:1665–1673.