Proxy Perspectives about End-of-Life Care for Person’s with Cancer

Abstract

Background

Each year more than one-half million people die from cancer in the United States. Despite progress in increasing access to palliative oncology services, end-of-life care still needs improvement. Measuring the quality of the end-of-life experience is difficult due to patient debility and reduced consciousness as death approaches. Family proxies have been proposed as valuable informants regarding the quality of end-of-life care. This paper describes family proxy perspectives on care at the end of life in patients who died from advanced cancer

Methods

In the context of a novel outpatient palliative care demonstration project, 125 family proxy respondents completed a structured survey by phone 3–6 months following patients’ death from breast, lung, or gastrointestinal cancer. Four key quality of care indicators were measured: decision-making and physician communication, location of death, hospice involvement, and end-of-life symptoms.

Results

Proxies reported that 80% of patients completed advance directives and about half of them were helpful in guiding care; communication with physicians about end-of-life treatment wishes occurred in 67%, but less than half actually made a plan with the physician to ensure that their wishes were followed; the majority died in their location of choice, most often at home; over half had hospice involvement for an average of 41.8 days prior to death; and during the last week of life the majority of patients experienced troublesome physical and emotional symptoms.

Conclusions

Measurement of proxy perspectives is feasible and these results provide actionable data for areas of improvement in palliative oncology care.

Introduction

Each year more than one-half million people die from cancer in the United States. ¹ Despite progress in increasing access to palliative oncology services, end-of-life care still needs improvement. Specifically many patients still die with significant pain and other symptoms and may continue to receive care that is inconsistent with their wishes. ^2–11

Measuring the quality of end-of-life care is methodologically challenging and can present serious threats to study validity. ^12–16 A measure’s reliability depends on its ability to be completed by the target population. A measure’s validity is determined by the extent to which it captures aspects of quality of life and end-of-life care that are valued by patients and families. ^{17, 18} Quality of life measurement is generally considered most valid through self-report; however reliably obtaining self-report data at the end of life is difficult because patients’ have declining energy and cognition as death approaches.⁷ There is a high likelihood of nonresponse bias, resulting from patients’ diminished capacity with progressive disease. ^18–20

Although there is considerable debate, it has been argued that for some domains of end-of-life quality of life and quality of care, the family can serve as a reliable proxy to capture and evaluate the patients’ experience. ^{7, 20, 21} Based on extensive interviews with patients, families, and clinicians, Teno et al.¹⁷ developed a conceptual framework of patient-centered, family-focused care. This conceptual framework served as the basis for the Toolkit After-death Bereaved Family Member Interview (ADI), a proxy measure of “good dying”. In this paper we describe family proxy perspectives on end-of-life care, as measured by the ADI, in a population of advanced cancer patients.

Methods

We conducted a survey of family proxy perspectives on selected domains of end-of-life care of person’s with advanced cancer who participated in Project ENABLE (Educate, Nurture, Advise, Before Life Ends), an outpatient palliative care early intervention demonstration project implemented at two oncology care settings.²² As a quality improvement initiative, the project was designated as exempt from requiring approval by our institutional review board; however verbal consent of the family proxy was obtained prior to conducting this telephone-based interview.

Setting and Sample

The ENABLE demonstration project has been described elsewhere²² but is briefly summarized here to provide context for this report. The Robert Wood Johnson Foundation’s, “Promoting Excellence in End-of-Life Care” program funded demonstration projects ^{23, 24} of collaborations between hospices and cancer centers to determine the feasibility of integrating palliative care concepts at the time of diagnosis into the care of patients with advanced cancer. ²² The project was initiated at three sites in New Hampshire: a National Cancer Institute (NCI)-designated comprehensive cancer center, a community oncology practice, and a small rural community outreach clinic. Project implementation was incomplete at the rural community outreach practice due to administrative and personnel factors. ²² The 5 ADI surveys from this site were excluded from the analysis.

From 1999–2001, consecutive patients with advanced breast, gastrointestinal, or lung cancer were identified at the time of diagnosis and invited to participate in the project. The ENABLE intervention consisted of an Advanced Practice Palliative Care Nurse Coordinator (PCN) and a series of four psycho-educational seminars, (“Charting Your Course”) derived from focus groups of patients with advanced cancer and their families. The topics of the 4 interactive, discussion sessions included managing symptoms; family and health care team support and communication, community resources; decision making and advance care planning; and spiritual issues, unfinished business, grief and loss. Session materials were available in a self-help format for participants who could not attend the seminars. After enrollment, patients (and their family members) attended the seminars, and PCNs followed them until death including a bereavement phone call to a family member.

Proxy Recruitment and Data Collection Procedures

Three to six months following a patients’ death a letter was sent to the person identified as the “contact person” in the patients’ medical record asking if they would be willing to participate in a telephone interview about the care that the participant received from diagnosis to the final week of life. The letter stated that the goal of the interview was to identify areas for improvement in the care delivery systems. The letter indicated that a member of the project team would contact them by phone within two weeks unless they returned a postage paid envelop indicating they did not wish to be contacted. If they did not have significant contact with the patient at end of life, we requested the name of an alternate contact. The phone interview using the ADI survey was confidential and included topics such as the location of care, communication, and the care and symptoms experienced in the last week of life. The interview was conducted by two trained, non-clinical project staff and lasted between 30 minutes and two hours. Responses were recorded on the survey including any additional comments that the respondent wished to share.

The Toolkit After-Death Bereaved Family Member Interview (ADI)

Teno et al. ¹⁷ developed the ADI as a family proxy measurement tool based on a conceptual model of patient-focused, family-centered end of life medical care. This conceptual model of “good care” at end-of-life was based on a literature review, qualitative interviews with professionals, and six focus groups (held in Arizona (2), Massachusetts (3), New York (1)) of bereaved family members from an outpatient hospice, hospital, nursing home, or non-hospice visiting nurse agency. Survey items were then created to measure the major domains represented in the model. ¹⁷ The ADI Version 4 survey used in this study underwent psychometric testing in 156 bereaved family members representing patient deaths that occurred in outpatient hospice services, a consortium of nursing homes, and a hospital in New England. This version was designed to be telephone-administered. It is comprise of 64-items that retrospectively evaluate eight domains of care: informing and making decisions, advance care planning, closure, coordination, achieving control and respect, family emotional support, self-efficacy, and ratings of care.

The ADI Version 4 was designed primarily to provide data that would be actionable and facilitate quality improvement. It posseses adequate internal consistency and test-retest reliability and acceptable content, construct, and criterion validity. ²⁵ Detailed reliability and validity information are available at: http://www.chcr.brown.edu/pcoc/Satisfac.htm.

Data Analysis

Descriptive statistics and univariate tests were performed on individual survey items. Continuous measures were compared using two-sample t-tests, while proportions and categorical measures were compared using chi-square tests. All analyses were performed using S-plus version 7.0.

Results

Two hundred and sixty-eight of the 380 ENABLE patients died during the three year demonstration project. Of the 268 deaths, 130 proxy respondents completed the ADI interview. Of the 138 decedents for which no responder participated, the majority were due to an inability to locate an appropriate proxy (e.g. no contact listed, bad address/letter returned, proxy not involved in care). In a minority of cases, the proxy declined participation by returning a post card. Five interviews from the rural outreach site were excluded due to incomplete data. The final sample for analysis was 125; 66 % (n=82) respondents from the comprehensive cancer center and 34% (n=43) from the community oncology practice.

Table 1 describes decedent characteristics and Table 2 describes the proxy respondent characteristics. There were no statistically significant differences between respondents and non-respondents relative to mean patient age, gender, or cancer type (data not shown). However, respondents were more likely to be from the cancer center than the community oncology practice. Because there were no significant differences in item responses between the respondents from the cancer center and the community oncology practice the data were collapsed and results are reported for the total sample (N=125).

Table 1.

Characteristics of Decedents (N=125)

	n (%)
Decedent Age [mean (SD)]	64.2 (11.5)
Years of Education [mean (SD)]	14.2 (8.3)
Female	67 (54)
Race, white	121 (98)
Married	90 (73)
Religion, Catholic/Christian	92 (75)
Household Income
Under 11K	7 (6)
11–24,999	25 (20)
25–49,999	28 (22)
Over 50K	34 (29)
Don’t know/declined	29 (23)
Cancer Site
Lung	63 (50)
Gastrointestinal	38 (30)
Breast	24 (20)
Site of enrollment
NCI-Cancer Center	82 (66)
Community Oncology Practice	43 (34)
Site of death
Home	63 (50)
Hospital	35 (28)
ICU	6 (5)
Emergency Department	1 (1)
Nursing Home	10 (8)
Residential Hospice	4 (3)
Other	6 (5)

Table 2.

Characteristics of Proxy Respondents (N=125)

Proxy Age [mean (SD)]	57.1 (15.8)
Education, Years [mean (SD)]	13.9 (3.1)
	n (%)
Relationship to decedent
Spouse	88 (72)
Child	22 (16)
Sibling	6 (5)
Parent	3 (2)
Other	6 (5)
Proxy self-rated health (N=120)*
Excellent	34 (28)
Very Good	42 (35)
Good	29 (24)
Fair	13 (11)
Poor	2 (2)

Note.

^*Variations in N are noted and occurred due to survey skip pattern or if respondent choose not to answer.

Decision-Making, Advance Care Planning, Physician Communication

Table 3 shows the results of selected items related to decision-making and physician communication. Respondents reported that 78% and 81% of decedents had completed a durable power of attorney for health care or living will, respectively. Sixty-five percent indicated that the decedent had discussed their advanced directive with their physician. Of 109 who responded, 46% indicated that these documents helped “a little” or “a great deal” while 36% indicated that advance directives documents had “no effect” on care at end of life.

Table 3.

Items Related to Shared Decision-Making and Physician Communication

	n (%)*
Had DPOA HC	97 (78)
Had Living Will	101 (81)
Discussed AD with physician (N=114)	74 (65)
Role of Advance Directive (N=109)
Helped great deal	38 (35)
Helped a little	12 (11)
No effect	39 (36)
Caused some problems	5 (5)
Don’t Know	15 (13)
Had specific wishes for treatment while dying	64 (53)
Discussed wishes with physician	73 (67)
Made a plan to ensure wishes were followed	61 (57)
Wishes were followed (N=113)
A great deal	66 (58)
Very Much	23 (20)
Moderately	4 (4)
Very little	4 (4)
Not at all	2 (2)
Don’t know	14 (12)
Physician communication about possibility of death	93 (74)
Physician discussed treatment choices	103 (83)

Note.

^*Variations from the total sample N=125 are noted and occurred due to survey skip pattern or if respondent choose not to answer.

Abbreviations-DPOA HC=Durable Power of Attorney for Health Care; AD=Advance Directives

Over half of respondents indicated that the decedent had specific wishes about the treatment they wanted to receive while dying. Respondents reported that 67% of decedents talked to their physician about their wishes for treatment at end-of-life and 57% made a plan with their physician to ensure that their wishes would be followed. During the last week of life, 65% of respondents believed that the patients’ treatment goals were to relieve pain, while 17 % thought they preferred a course of treatment to extend life. Of the patients receiving hospice care, only 14% expressed a goal of “extending life” compared with 22% who were not receiving hospice care. Whereas “relieving pain” was the expressed goal of 71% of patients receiving hospice care compared to 59% who were not receiving hospice care. However, these differences were not statistically significant (p=0.25). Seventy-eight percent of respondents believed that the decedent’s wishes were followed “very much” or “a great deal”.

When asked whether the physician communicated “in an easily understandable way, about the possibility that [the patient] would die from this illness”, 74% indicated that the physician did so. The majority (83%) also believed that during the final stage of illness the physician told them about choices for treatment in an easily understandable way.

Preference for Location of Death

Half of all deaths occurred in the home. Thirty four percent of deaths occurred in the hospital and only 8% occurred in a nursing home (Table 1). Respondents believed that the death occurred in the location that the decedent would have preferred in two-thirds of the deaths (Table 4). In a third of the 27 deaths that did not occur in the preferred location, respondents believed that this was due to an inability to control the decedents’ symptoms.

Table 4.

Location of Death and Location Preference

	n (%)*
Death occurred in preferred location	83 (66)
What prevented dying in preferred location? (N=27)*
Symptom control	9 (33)
Time	4 (15)
Family unable	1 (4)
Other	7 (26)
Don’t Know	6 (22)

Note.

^*Variations from the total sample N=125 are noted and occurred due to survey skip pattern or if respondent choose not to answer.

Hospice Involvement

Table 5 describes proxy perspectives on hospice involvement. Eighty-five percent of respondents believed that the decedents were told about hospice. Fifty-nine percent of decedents had hospice involvement at the time of death, with a mean length of stay in hospice of 41.8 days (range 1–150 days). Of the decedents who experienced home death, 62% had hospice involvement vs. only 33% who did not have hospice involvement (p<0.001). Of the respondents indicating that the patient died in their preferred location, 81% had hospice involvement vs. 50% who did not have hospice involvement (p<0.001).

Table 5.

Hospice Involvement (N=125)

	n (%)
Told about hospice	82 (85)
Hospice involved	74 (59)
Average days enrolled in hospice	41.8 (range 1–150)

Symptoms in the Final Week of Life

Proxies reported that the majority (91%) of decedents were able to communicate during the last week of life. Table 6 describes selected item responses related to emotional and physical symptoms. Fifty-two percent of respondents believed that the decedent had a “very” or “somewhat” difficult time tolerating emotional symptoms and 61% had a “very” or “somewhat” difficult time tolerating physical symptoms during the last week of life. Respondents reported that 68% of the decedents had pain in the final week of life and that for 57% pain was present most or all of the time. Nearly half believed that the physical pain was “extremely” severe. Only 14% of the decedents indicated that they actually spoke to the respondents about their pain directly. Hence, pain presence and severity was based primarily on the proxys’ observations rather than the decedents’ self-report. In response to the question “At any time in the last month of life, do you feel that more should have been done by the doctors and nurses to keep [patient] free from pain?” 75% of the respondents indicated “no”. The two most frequent symptoms after pain that respondents believed caused decedents distress included no appetite (13%) and shortness of breath (11%).

Table 6.

Emotional Symptoms and Pain During Last Week of Life

	n (%)*
Difficulty tolerating emotional symptoms (N=116)
Very difficult	26 (22)
Somewhat difficult	35 (30)
Not very difficult	26 (22)
Not at all difficult	19 (17)
Don’t know	10 (9)
Difficulty tolerating physical symptoms (N=118)
Very much	35 (30)
Somewhat	37 (31)
Not very much	27 (23)
Not at all	11 (9)
Don’t know	8 (7)
Had pain in last week	85 (68)
Pain Frequency (N= 84)
Occasionally	25 (30)
Half the time	11 (13)
Most of the time	29 (34)
All of the time	19 (23)
Pain Severity (N=84)
Not at all	8 (9)
Moderately	36 (43)
Extremely	40 (48)

Note.

^*Variations from the total sample N=125 are noted and occurred due to survey skip pattern or if respondent choose not to answer.

Discussion

Project ENABLE, conducted from 1999–2001, determined the feasibility of early integration of palliative care concepts for patients with advanced cancers at an NCI-designated cancer and community oncology practice. ²² Within this project we assessed selected domains of the quality of end of life care by surveying family proxies after patients’ death using an instrument that was under development at the time, the ADI. Family proxies, who represented half of the decedents who died during this project, were receptive to completing the interview. Non-response was due primarily to an inability to locate an appropriate proxy. Although some respondents became tearful recalling their loved ones’ end-of-life experience, all were able to complete the survey and many shared anecdotes about particularly positive or negative experiences.

High rates of advance directive (AD) completion were reported in our sample; 78% had appointed a durable power of attorney for health care (DPOA-HC) and 81% completed a Living Will. This proportion is markedly higher than the 25% completion rate identified by a chart audit study in the early 1990s. ²⁸ Our completion rates are also higher than those reported in a statewide chart audit study of 782 New Hampshire home, hospital and nursing home deaths from 2002 that found completion rates of 45% and 37% for a durable power of attorney for health care and Living Will, respectively. ²⁹ However, our completion rates are comparable to two more recent reports that used tools and methods that were similar to those used in this study. Teno and colleagues ²⁶ found that proxy-reported AD completion rates were 71% in a national sample of non-traumatic patient deaths (including 25.8% cancer deaths). Using an afterdeath survey of 165 family members of patients dying in 1 of 5 Catholic hospitals in Ohio, Bernal and colleagues found that 70% of proxies indicated that patients had some type of advance directive including DPOA-HC, Living will, or DNR order. ²⁷ It is important to note that this latter study also conducted a chart review in a comparable patient sample and found that 76% of patients had some form of written AD, thus strengthening the accuracy of proxy report. While we cannot rule out the possibility that in our study family could have inaccurately recalled the presence of an advance directive, the consistency of high AD completion rates in multiple more recent studies (that also relied on proxy report) makes this less likely. It may be that public awareness and educational campaigns are having an effect on encouraging the completion of such directives. ^{26, 27}

Perhaps more important than measuring AD completion rates is measuring the communication of end-of-life treatment wishes to the clinician and the overall effect of this communication on end-of-life care.^26–28 From a clinical perspective, advance directives are only effective if the presence of such a document or the wishes contained in them are communicated to the clinicians providing their end-of-life care. In our study it was the respondent’s impression that in nearly half of the cases in which an AD was completed it had some or a great deal of effect on treatment decision-making. Respondents in our study also indicated that just over half of the decedents expressed specific wishes or plans about the types of medical treatment they wanted while dying, but only two-thirds of decedents had discussed these wishes with their physician. In the last week of life, for patient’s who expressed preferences for either “life-extending or “pain relieving” courses of action, proxies believed that the decedents’ wishes were followed in 78% of the cases. Teno et al.²⁶ reported that the presence of an AD or patients’ expressed wishes (even in the absence of a written AD), showed a decrease in resource utilization due to the reduced use of life-sustaining treatments compared with those who had not communicated their wishes in this regard. Taken together these results suggest that there is significant room to improve how we elicit, document, communicate, and carry out patient preferences for treatments and care at the end of life. Improved communication between patients and clinicians about patient’s preferences may result in a better match between the care desired and received. It may also effect the use of life-sustaining treatments and thereby decrease overall resource utilization at end of life.

Respondents also identified that communication between the decedent and their oncologists regarding the possibility of death from the illness occurred in nearly three quarters of the sample. Although patients and families routinely report a desire for open communication with their physicians, ²⁹ open communication about end of life treatment wishes has traditionally been a difficult topic for many oncologists to discuss. ²⁷ Patients’ discussions of end-of-life treatment wishes (and the actualization of these preferences) are more likely to occur if physicians feel comfortable approaching this topic with their patients.

Most patients died in their preferred location which for the majority was in their home. Rates of home death were higher in this project compared to our state and national data. According to the New Hampshire (NH) Department of Vital Records, deaths in 2001 occurred in the following locations: 43% hospital; 30% in nursing homes; 20% in home, 4 % in other facilities and 3% other. Nationally up to 50 % of deaths occur in hospitals ^30–32 and 25% in nursing homes ³³. One possible explanation for the higher rates of home death and lower rate of nursing home deaths in our sample may be the relatively young average age of decedents (62 years) all of whom had cancer. Perhaps younger patients had more family and community resources to accommodate death at home. In NH, the average age at time of death for nursing home residents (all diagnoses) is around 86 yrs ³⁴. Another explanation could be the higher rate and longer duration of hospice involvement in this sample compared to the national average. ³⁵ Others have also found associations between advance care planning, improved communication, hospice involvement and a higher rate of home death. ³⁶

Respondents reported that decedents experienced difficult emotional symptoms and pain during the last week of life. Proxy reports of the presence of pain in 68% of decedents (of which 48% had “extremely severe” pain) are disturbing. Although studies of proxy reports about patients’ quality of life/quality of care show good agreement in some observed behaviors (such as support, care giving, and physical functioning), proxies rarely concur with patients’ reports of subjective symptoms (such as pain, mood, and satisfaction). In cases of poor agreement, without exception, family report a more negative view of patients’ quality of life over what the patient reported ²⁰. This may have been the case in our study. However we were surprised that only 11% of respondents reported distressing dyspnea even though 50% of the decedents died of lung cancer. Rates of 25–50% of unrelieved dyspnea in late stage disease have been reported by others in mixed ²¹ and lung cancer samples. ³⁷ The lower reported dyspnea rate, over what we would have expected suggests that either family had an inaccurately low recall of this symptom or that it was not present (perhaps because it was well-managed).

The results of this survey should be interpreted with caution and within the context of several very important limitations. First, at the time of this study, the ADI was a new instrument. Adequate reliability and validity had been reported, but only after limited psychometric testing. Some questions may not have been clearly understood and the structured response set (e.g. “yes” or “no”, Likert scales) may not have allowed the proxy to express the complexity of the true situation. However, few measures of the target phenomenon (the quality of end of life care) existed at the time of our project. In particular we were interested in obtaining data that could provide actionable results for our quality improvement initiatives. Results of the dying experience in a national sample of adult deaths using a revised version of the ADI have also been reported. ^{21, 26} Subsequent cognitive testing of the ADI has resulted in revisions to some of the domains and items that had been included in the version we used. The ADI continues to undergo development and psychometric testing including test-retest reliability of after death proxy reports and pre-death comparisons between patient and proxy responses related to perceptions of health care team communications and patient preferences for care (Teno, Personal Communication May 1, 2007).

Another limitation was the limited ethnic diversity in our convenience sample of patients (and their proxies) which reflects the racially homogeneous local population of our predominantly rural environment. We had a 50% cooperation rate, however most of the non-responders were due to our inability to locate an appropriate proxy months after the patient died. Although there were no demographic differences between respondents and non-respondents relative to decedent age, gender, or disease type. selection bias in other important unmeasured parameters could exist. Finally, the majority of the decedents were treated in an academic, NCI-designated cancer center. Therefore the end-of-life experience reported by family may not reflect the experiences of the majority of persons with cancer who are treated and die in community settings.

Despite these limitations, the findings of this project provided us with useful benchmarks to consider as we designed patient- and family-centered hospital and cancer center quality improvement projects regarding advance care planning, psychosocial care, and pain management at the end-of-life. ³⁸ This survey also provided us with pilot data to develop a scientifically rigorous randomized clinical trial of a palliative care intervention/care model that is currently underway. ^{39, 40} Our experience allowed us to identify methodological issues and strategies to address them. For example, we realized the need to prospectively engage and maintain longitudinal relationships with family caregivers of persons with newly diagnosed, advanced stage cancers who we hoped to survey after the patients’ death.

In summary, proxy perspectives obtained using the ADI informed our understanding of the quality of care and quality of life of persons with advanced cancer in our region. Some domains of care, such as advance care planning and hospice length of service, reflected higher rates over national averages. However, especially in the realm of pain and emotional symptoms during the last week of life we identified a great need to improve both our measurement of symptoms and the clinical care that is provided at the very end of life. Development of reliable and valid measures of the end-of-life experience of patients using proxy’s estimates of symptoms and quality of care is an important target of research. Ongoing investigation of the reliability and validity of family proxy reports can help us to develop scientifically rigorous measures for use in evaluation of interventions designed to improve the quality of end-of-life care for patients and their family members.