NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Lorenz K, Lynn J, Morton SC, et al. End-of-Life Care and Outcomes. Rockville (MD): Agency for Healthcare Research and Quality (US); 2004 Dec. (Evidence Reports/Technology Assessments, No. 110.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of End-of-Life Care and Outcomes

End-of-Life Care and Outcomes.

Show details

3Results

After a description of the results of the literature search, this chapter first takes up the evidence regarding satisfaction with end-of-life care and the association of satisfaction with other outcomes (Task Order Question 1a). We then address and assess the measures available for the important domains of patient and family experience (Task Order Question 1b). For each of the elements that shape the end-of-life experience and that our work targeted, the ensuing sections take up the topic and address Task Order Questions 2 and 3 around each topic area. Thus, sequential sections of this chapter address symptoms (pain, dyspnea, depression and anxiety, and behavioral symptoms associated with dementia), family caregiver issues, continuity, and advance care planning. Each one generally starts with a summary of the existing systematic reviews, then reviews the interventions that have been studied, and finally reviews the highest-quality observational research. In the sections at the end of Chapter 3, we summarize and synthesize the evidence related to the association of patient, family, and health system factors with those outcomes (2a, 2b, and 2c) and the effectiveness of interventions and population factors related to variation in intervention effectiveness (3a, 3b), so that an overview of the findings related to the questions as asked is readily available.

Results of the Literature Search

The literature search performed by NLM resulted in 16,310 titles. The supplemental library search performed by RAND staff identified an additional 3,748 titles. Library searches performed by NLM focusing on specific clinical conditions of cancer, heart failure, and dementia added 1,187 new titles. In total, the RAND reviewers examined 21,245 titles identified through literature searches, of which 5,563 were considered possibly relevant to our topic areas and continued to abstract review. Out of the 2,493 references used in development of the National Consensus Project clinical practice guideline, our literature searches and title review process had not identified 675. These references were added to the library of abstracts and proceeded on to abstract review. The literature search of the DARE abstracts identified 92 titles, of which 62 were considered potentially relevant to our topic areas and proceeded to abstract review. Another 71 articles were added to the library of abstracts from the NICE guideline, the Health Canada reports, the Toolkit of Measures for End of Life Care, and the files of our content experts. An additional 22 articles were suggested by the TEP and peer reviewers, of which 10 were considered potentially relevant and proceeded to abstract review.

Of the 6,381 titles identified as possibly relevant to our topics, the reviewers screened the abstracts for 5,216 titles; 13 titles were identified as duplicates already abstract screened, and 1,152 titles did not have abstracts to screen. Of the 5,216 abstracts screened, 3040 were excluded for reasons listed as “population, intervention, or outcome exclusion” on the abstract screener: 761 were excluded as not about end-of-life care and outcomes; 26 were excluded as predominately about sudden/violent/ non-chronic death; 148 were excluded as predominately about chemotherapy/surgery/stents/ laser/radiation; 963 were excluded because no outcomes were reported; 620 were excluded because the outcomes were unrelated to patients/families/nonprofessional caregivers; 370 were excluded as primarily useful as background only; 97 were excluded as predominately reporting on prognosis or trajectories; and 55 were excluded because the data were older than 1990. Ten abstracts were excluded because the population discussed was not adults. Thirty-two abstracts were excluded as non-Western in location. Six hundred forty-six studies were excluded due to study design: 239 were qualitative studies; 52 were nonsystematic reviews; 20 were other types of reviews; 138 were observational studies of less than 30 subjects; and 197 had unclear study designs. One hundred ninety-nine abstracts were excluded for topic: 56 as bereavement only; 35 as symptoms other than those included in our scope; 80 as topics other than those included in our scope; and 28 as unclear topic. The remaining 1,289 articles were determined to be potentially relevant to our topic and were ordered.

Of the 1,289 articles ordered, we retrieved 1,274 prior to the cut off date (Sept. 3, 2004). On detailed review of the articles, 363 studies were reclassified as excluded. The remaining articles comprised 134 interventions, 95 systematic reviews, and 682 observational studies of sample larger than 30. These 911 articles were distributed by topic and study design as presented in Table 2. As one article can report on multiple topics the numbers in Table 2 do not add to 911. Figure 1 presents this information pictorially.

Table 2. Study design by Topic Area.

Table

Table 2. Study design by Topic Area.

Figure 1. Article Flow.

Figure

Figure 1. Article Flow. * Abstract screener question relating to exclusion reason is in parentheses

A. Key Question 1a. What individual outcome measures are most strongly associated with overall satisfaction with end-of-life care?

Systematic Reviews and Satisfaction with End-of-Life Care

One particularly salient aspect of evaluation of end-of-life care is whether the patients and families are satisfied with care—in other words, how they subjectively perceive the care provided. We included in this literature the range of articles we identified that subjectively rated either global satisfaction or more specific elements of the care provided to patients or in support of caregivers living with serious and eventually fatal illness.

We evaluated ten systematic reviews that potentially dealt with the subject of patient or caregiver satisfaction. Six addressed the project questions and met implicit quality criteria. One of the reviews focused specifically on a cancer population, and the other five did not limit their reviews to one disease cohort. We went beyond the systematic reviews by including other intervention studies addressing the outcome of patient or caregiver satisfaction published after these systematic reviews or published at any time if not already addressed in a systematic review. In total, we reviewed an additional 12 intervention studies. Finally, we explored the observational literature that used a prospective cohort design and that also presented data separately by race, selected disease cohorts, or selected sites of care. In addition, we identified observational studies that addressed the relationship between satisfaction and other outcomes. In total, we reviewed 17 observational studies.

The remainder of this section summarizes the systematic reviews, meta-analyses, intervention, and observational studies relevant to patient and caregiver satisfaction. We also evaluated the qualitative literature in this area to try to better interpret the strength of the literature and meaning of patient satisfaction with end-of-life care. The relationship of satisfaction to other measures is summarized at the end of this section. Summaries of the association of patient, family, and health system factors to satisfaction and the effectiveness of interventions in improving satisfaction are found at the conclusion of Chapter 3.

Wilkinson et al. conducted an extensive search of the English and non-English literature covering the years 1978-1998 including hand searches of major palliative care journals, reference mining of major citations, consultation with palliative care experts, and a search for gray literature. This review identified 831 documents, of which 688 were retrieved and analyzed. They found 83 papers relevant to patient and caregiver satisfaction with palliative care and were able to retrieve 79 of them. This review described five reports with a randomized controlled design and related to palliative care and satisfaction in the UK and North American literature.

These five reports were from four RCTs and included a study of an inpatient hospice for veterans that found a positive effect on patient and caregiver satisfaction, a study of case management for terminally ill cancer patients in a London health district that had no effect on satisfaction, a study of home-based primary care that included non-terminal and terminally ill veterans (all of whom had advanced illness), and a study of multidisciplinary home care including 24-hour telephone availability for homebound chronically or terminally ill persons. The two RCTs describing home-based services both demonstrated effects on patient and caregiver satisfaction, and these are described below in the Higginson review from 2001, as is the study of inpatient hospice for veterans.

Many of the studies described by Wilkinson et al. related to comparative, often retrospective or cross-sectional assessments of specific inpatient or outpatient supportive services for patients near the end of life. The review described research reports that were heterogeneous in their comparisons and methods, although they generally described hospital care unfavorably compared with alternatives that included a variety of home care and hospice models. The nature of the research designs and heterogeneity of service models did not suggest the superiority of one form of palliative care delivery over another. This review highlighted a number of important methodological issues in end-of-life research in satisfaction including

  • lack of a priori definitions of satisfaction
  • ceiling effects of specific items or measures of satisfaction
  • lack of well-validated measures for assessing satisfaction with end-of-life care
  • the difficulty of assessing association between respondent reports of satisfaction in non-randomized designs because of large observed differences in samples
  • unresolved methodological issues in end-of-life care satisfaction assessment including timing of patient assessment due to frail health states, use of proxies, and questions related to retrospective assessment.

In addition, this review identified a large descriptive study that found differences between cancer and dementia patients' caregivers satisfaction related to differential satisfaction with information and the physical attributes of the hospital environment.

Higginson et al. undertook a systematic review and meta-analysis of the effectiveness of palliative care teams in 2001, which examined satisfaction74 as one of the outcomes. Using a robust search strategy to identify studies of palliative care services and their effects on patients, caregivers, and economic outcomes, the review searched ten databases from 1977 to 1999. The review identified 25 experimental and observational studies with outcomes that could be synthesized. Five studies included satisfaction as a measure and the pooled weighted mean was 0.24 (-0.04–0.52) favoring the intervention. Although not analyzed separately as an outcome, satisfaction was combined with pain, other symptoms, quality of life, referral to other services, and therapeutic interventions. This aggregate variable demonstrated a small effect (weighted mean 0.32 [0.15–0.49]), excluding one outlier) of palliative care services on overall outcomes, although sample sizes of the studies were very small.

Among these studies, one study evaluated the effect of an inpatient hospice on veterans and their caregivers and reported a positive effect of a multidisciplinary team on both patient and caregiver satisfaction, associated with improved ratings of interpersonal care. Several interventional studies described outcomes of home-based services with generally positive effects on satisfaction. One Australian RCT of home-based hospice care compared with regular home care reported greater dissatisfaction among non-hospice patients. The only other difference noted was higher pain duration among non-hospice patients. An RCT of home-based primary care for veterans reported significant improvements in health-related quality of life (HRQOL) among terminal patients but no significant improvement in satisfaction (the effect was positive and moderately large, but not statistically significant). In the larger group of nonterminal but very ill patients who were homebound with CHF and COPD, HRQOL did not improve, but satisfaction showed roughly the same difference between those with and without home-based primary care as was in evidence for “terminal” patients, though the differences were not statistically significant at the p<0.05 level. Caregivers in both groups receiving home-based primary care experienced improvements in HRQOL and satisfaction. A quasi-experimental study of home-based hospice found improvements in pain processes and overall symptoms in hospice but high satisfaction in both hospice and conventional care groups. An RCT of home-based multidisciplinary care for patients with terminal and advanced chronic illness reduced hospitalization, nursing home admission, and outpatient visits and increased home death. In addition, home-based care was associated with greater patient and caregiver satisfaction.

Gysels and Higginson's systematic review of supportive and palliative care for adults with cancer72 identified studies published up to 2003 and was organized around a wide variety of supportive interventions in oncology including coordination of care, patient activation, communication, information provision, psychological support, social support, spiritual support, palliative care services, rehabilitation, complementary therapies, and family/caregiver support. Of 5263 studies reviewed and 443 potentially eligible studies accepted after abstract review, 40 papers describing heterogeneous interventions that measured satisfaction as an outcome were accepted into the review. Of these 40 papers, seventeen RCTs examined effects of an intervention on the satisfaction of either patients or caregivers. This systematic review did not report summary conclusions of the evidence.

Gysels and Higginson's review described randomized controlled trials in the area of care coordination, advance care planning, and information provision to patients. Several of the RCTs identified by this review were described in the context of previous reviews. Among those that were not, one improved the coordination of end-of-life care by using a Patient Care Traveling Record (PCTR). It did not report an effect on satisfaction, but dropout due to patient frailty was quite extensive. Another RCT of a patient held record (PHR) reported no improvement in satisfaction with information—although perception of communication was relatively high in this sample, which included both oncology outpatients and patients who were already enrolled in a home hospice service. An RCT that involved randomizing patients followed by hospital-based specialists to early follow-up that included their primary care physician, and the intervention group reported higher satisfaction. A small CCT (n=24) implemented a “coaching” intervention intended to improve patient interaction in oncology consultations, and it did not result in higher satisfaction, although it did achieve improved patient perception of decision quality and MD-patient agreement. In an intervention that involved a patient-nurse meeting for counseling and education of newly diagnosed cancer patients, both improved information and satisfaction with the consultation were reported by patients and their spouses. A similar RCT that simply involved an informational pamphlet without the personal involvement did not affect either information or satisfaction.

Additional Interventional Studies and Satisfaction with End-of-Life Care

Our review identified a number of additional interventional trials in palliative care that included as an outcome a measure of patient or caregiver satisfaction with care. Several of these addressed comprehensive or coordinated services for chronically ill patients. The following text summarizes these studies, first the studies of comprehensive or coordinated services and then the studies of communication or advance care planning. Within each section, we first discuss RCTs and then articles with other study designs.

Grande et al. conducted a randomized controlled trial of a hospital-at-home service in the UK for the terminally ill.83 This intervention provided in-home nursing support up to 24 hours daily for up to two weeks, predominantly used for terminal care during the final weeks of life. Referrals came from general practitioners and one-third from inpatient discharges. All patients were eligible to receive other care concurrently, including a variety of home services and hospice. Of the 262 referrals, 43 were randomized to control (C) and 186 to intervention (I). The majority of intervention and control patients had cancer. The study incorporated questionnaires that assessed general assessment of care and symptom management. Informal caregivers noted no difference in any supportive services, caregiver support, or symptoms with the exception of pain, which the control group rated as a relatively unmet need (3.00 vs. 2.52). The Jadad score for this study was 3.

Ringdal et al. conducted a cluster randomized trial which involved six Norwegian health districts randomized to an intervention that included an community education and close integration of hospital-based palliative care with local provider activities.84 Within health districts designated for intervention, adult cancer patients with a life expectancy between two and nine months were eligible. Researchers measured caregiver satisfaction using the 20-tem FAMCARE scale, which was developed specifically to measure satisfaction with advanced cancer care. A large proportion of caregivers refused to participate in completing surveys (114 / 426). Of those who completed the study, caregiver satisfaction scores favored the intervention with regard to specific items related to pain management, communication with the family about prognosis, treatments, and involvement of caregivers. Ringdal et al. examined the association of overall satisfaction with caregiver gender, age, education, relationship to the deceased; gender, age, and cancer type of the deceased; and place of death. Satisfaction was higher among spouses than children, higher if the deceased individual was a man, and higher among family of patients who died at home. In a fully adjusted model, in addition to the main intervention effect, only relationship to the deceased was significantly related to overall satisfaction. Spouses scored on average 12.5 points higher on the 0–100 FAMCARE scale than children. The Jadad score for this study was 2.

Hanks et al. conducted a randomized controlled trial of the effectiveness of a UK hospital Palliative Care Team on symptoms, quality of life, and patient, caregiver, and provider satisfaction.85 All non-emergent inpatient referrals of persons who were not immediately likely to die were randomized to either physician-to-physician telephone consultation or in-person interdisciplinary palliative care team consultation. Satisfaction was evaluated with four items from MacAdam's Assessment of Suffering Questionnaire. Caregiver satisfaction was assessed using FAMCARE, Hospital Anxiety and Depression Scale (HADS), and additional items about hospital communication. A more detailed interview was also conducted with caregivers of all discharged patients. The satisfaction of community physicians and nurses of all discharged patients was assessed related to the appropriateness of care and support arrangements and communication with the hospital. Component and overall measures of satisfaction were high in both groups (3.5–3.6 / 4 where 4 is “very satisfied” on all patient measures; 1.9–2.5 / 5 where 1 is “very satisfied” on all caregiver measures) and did not differ at either time point. The Jadad score for this study was 2.

Rabow et al. conducted a controlled trial of an interdisciplinary team that targeted physical, emotional, and spiritual care for 90 patients in two university outpatient general medicine clinics randomly assigned as intervention or control clinics.86 Patients with cancer, advanced COPD, or CHF with a life expectancy of 1–5 years were eligible. The intervention improved dyspnea and sleep quality but not pain. Intervention patients reported higher spiritual well-being overall and in religious activities, and completed more advance directives. However, satisfaction as measured by 25 items (0–100 scale) from the Group Health Association of America Consumer Satisfaction Survey (satisfaction with care, attitude toward care) was high in both groups at baseline (satisfaction 73.7-I, 77.0-C; attitude 13.4-I, 14.0-C) and did not change. The Jadad score for this study was 3.

Brumley et al. conducted a pre-post test at Kaiser Permanente of a palliative care program and compared patients enrolled in that program (n=210) to a group of somewhat comparable patients (n=348) concurrently referred for home care.87 The Reid-Gundlach Satisfaction with Services 13-item instrument (0–48) measures overall ratings, perceptions of providers, and likelihood of recommendation. This analysis reported the change score in patient satisfaction 60 days after baseline in a subset of the original participants who died during the course of the study and completed the interviews (I =161 C = 139). At baseline, both groups reported a mean satisfaction of 40/48 and at follow-up satisfaction improved in both groups. Satisfaction did not differ significantly at either time.

Weisbord et al. conducted a pre-post uncontrolled study of a palliative care consultation in 39 poor prognosis hemodialysis patients.88 Nineteen of them evaluated the program, both before and then again two weeks after their consultation and a follow-up visit. Nine (47%) patients “strongly agreed” and four (21%) “somewhat agreed” that the meetings were useful. A similar proportion of patients also agreed that follow-up by the palliative care team would be useful. This intervention also assessed nephrologist satisfaction for 14 patients, and they also “strongly agreed“ or “agreed” that the consultation was useful for symptoms. Nephrologists agreed that palliative care consultation had provided useful information to 11 of the patients. Nephrologists asked palliative care providers to follow 12 of their patients at the conclusion of the study.

In other studies, Riegel et al. examined satisfaction with care as an outcome of a case management program using a standardized protocol and software support program for CHF.89 In this randomized controlled trial, telephone case management was provided to hospitalized patients with moderate to advanced heart failure (57% of sample were NYHA Class III and 15% Class IV at time of entry/hospitalization). The case manager also coordinated information with the patient's physician. Over the six-month trial, intervention patients received an average of 17 calls. Satisfaction with treatment, convenience, patient education, medication schedule, and MD care was evaluated. Of 358 patients randomized, survey data were obtained on 242, and only 184/242 patients completed a satisfaction survey. The difference demonstrated slightly higher overall satisfaction among the intervention group (22.88-I, 21.66-C), and both groups reported high satisfaction. The Jadad score for this study was 3.

We identified three RCTs that assessed satisfaction as part of a communication or advance care planning intervention. Bruera et al. studied 60 patients with cancer who were randomized to standard care (which included a written summary) or to receive a multidisciplinary outpatient cancer consultation with audiotaped recording to take home.90 Patients returned for follow-up on day 8 and responded then to questions about global satisfaction with the clinic's care, understanding and recall of the original consultation, and ability to discuss their illness with family and friends. Intervention patients compared to controls (31-I, 29-C) reported higher “usefulness” of the clinic (8.7/10 vs. 7.7/10, p = 0.04), but did not describe a significant difference in their perceived understanding and recall of recommendations, or in their perceived ability to discuss their illness with family and friends. The Jadad score for this study was 5.

Schneiderman et al. assessed perceptions using a structured interview as the outcome of a randomized controlled trial of an intensive care unit (ICU) communication intervention by an ethics team.91 The trial enrolled patients in whom treatment conflicts were imminent or already present (considering conflicts within or between the healthcare team and/or family). The study randomized 546 patients (276-I, 270-C) and conducted interviews with 108 intervention surrogates and 272 professional providers involved in 152 patients' care. Both surrogates and providers rated the consultation highly on a number of general attributes (helpful, informative, supportive, fair, respectful of values) and in facilitating specific processes (identifying, analyzing, resolving, educating, and presenting views). Both groups rated the consultation as moderately stressful. The Jadad score for this study was 3.

Molloy et al. conducted a trial of advance care planning in nursing homes using an educational program for staff, residents, and families combined with a validated advance care planning tool (Let Me Decide) that offered choices for life-threatening illness, cardiac arrest, and nutrition.92 Three pairs of randomly selected nursing homes were matched for hospitalization and case-mix, and site of death. This multifaceted intervention succeeded in increasing advance directive completion rates from 57% in control homes to 70% in the intervention homes, where most care plans used the more flexible Let Me Decide directive. Satisfaction was measured using two previously validated 23- and 29-item measures that assess satisfaction with involvement in care.93 Pre-post satisfaction (1–7) was 4.77 and 5.07 in the intervention and 5.09 and 5.10 in controls, and adjusted mean difference (-0.16, 95% CI, -0,41-0.10) was not significant. The Jadad score for this study was 1.

Bookbinder et al. conducted an uncontrolled pre-post study of a continuous quality improvement (CQI) intervention to reduce pain.94 The intervention consisted of intensive staff education and problem-solving targeted specifically at improving pain documentation, nursing pain knowledge, and patient satisfaction. Six hundred ninety-six patients who experienced pain during hospitalization were interviewed (398 pre-intervention and 298 post-intervention) about their overall satisfaction as well as satisfaction with their nurse and physician care. Patients reported a high level of satisfaction in both periods: 71% after intervention, contrasted with 61% before intervention, reported satisfaction with nursing care; 67% after vs. 63% before reported satisfaction with MD care; and 62% after vs. 54% before reported overall satisfaction. Satisfaction correlated with longest time to wait for medication (r=0.335), extent of pain relief (r=-0.304), and time to change medication (r=0.457).

Pietersma et al., in a study in which patients served as their own controls, evaluated patient satisfaction with a food cart on a palliative care service compared with standard food service (e.g., food trays). During a ten-day cart trial, 27 patients consented and participated, and patients were generally more satisfied with the food cart, which allowed them to choose their own items and portion sizes.95

Observational Studies Evaluating Satisfaction in Palliative Care

Of studies that examined racial/ethnic differences, several looked at white/nonwhite differences96–98 and two studies included African-American and Hispanics as separate categories.27, 99 The majority of this literature did not examine racial differences at all—many probably because they were small studies or performed in settings in which there were insufficient numbers of minorities. Of studies that examined racial/ethnic differences, some did not describe any differences,96, 97 although in several studies race/ethnicity was considered an exploratory variable98 or control that was not presented in available published comparisons.27 One study that did report racial differences noted that African-Americans (OR 3.3) and other non-whites (OR 2.5) compared with Whites were more likely to agree with the importance of using all available treatments no matter what the chance of recovery.99

We identified a number of observational studies that addressed end-of-life care within particular settings. A number of studies have addressed end-of-life care for hospitalized adults,27, 96, 98, 100 or more specifically end-of-life care in the ICU.97, 101 We also identified studies describing satisfaction in home care102, 103 or hospice/palliative care services.104–106 Other studies have assessed end-of-life care in general and in doing so, compared satisfaction with care across settings typically including home/hospice, hospital, and nursing homes.27, 107–112 These comparative studies highlight important differences with hospice users or caregivers of patients who died at home generally reporting higher satisfaction with many attributes of care 27, 107, 108, 110 than those who died or were cared for in other settings at the end of life.

With regard to disease, we found little evidence that satisfaction differs by disease.96, 98, 99, 104 At the same time, few studies have examined specific diseases or employed measures that are disease-specific.100, 113 To the extent that a particular disease is well represented by the literature, the experience of cancer patients and their caregivers is best characterized because many of the studies have either focused on cancer or have been conducted in palliative care settings where cancer predominates.

Several studies have evaluated satisfaction with aspects of end-of-life care in the context of large or particularly notable cohort studies. Teno et al.27 evaluated the U.S. dying experience through interviews with surviving family members representing 1,578 decedents from the mortality follow-back survey regarding patient and family centered end-of-life care. Sixty-seven percent of decedents died in an institutional setting while 33% died at home. Of those dying at home, 38% did not receive nursing services, 13% used home nursing services, and 49% had home hospice services. About 25% of all patients with pain or dyspnea at the end of life did not receive adequate treatment and one-quarter reported concerns with physician communication. More than one-third of respondents cared for by a home health agency, nursing home, or hospital reported insufficient emotional support for the patient and/or one or more concerns with family emotional support, compared with about one-fifth of those receiving home hospice services. Nursing home residents were less likely than those cared for in a hospital or by home hospice services always to have been treated with respect at the end of life (68% vs. 77% and 96% respectively). Family members of patients receiving hospice services were more satisfied with overall quality of care: 71% rated care as “excellent” compared with less than 50% of those dying in an institutional setting or with home health services. These data suggest that those dying in institutions have unmet needs for symptom management, physician communication, emotional support, and being treated with respect. Family members of decedents who died with home hospice services were more likely to report a favorable dying experience.

Tilden et al. (2004)114 examined the end-of-life experiences of elderly decedents dying out of the hospital in Oregon through a telephone survey of 1,189 family caregivers of decedents aged 65 and older who died of natural causes in community settings between 2000 and 2002. Outcome variables included advance directives, hospice enrollment, use of life-sustaining treatments, perceived decedent symptom distress, financial hardship, out-of-pocket costs, and family caregiver strain. Results showed that most decedents had an advance directive (78.3%) and were enrolled in hospice (62.4%). Although perceived decedent symptom distress was low overall, certain symptoms (e.g., pain, dyspnea, constipation) were distressing for approximately half of decedents experiencing them. Financial hardship, out-of-pocket expenses, and caregiver strain were frequently reported. American Indian race and younger age were associated with decedent symptom distress. Greater perceived decedent symptom distress, hospice enrollment, more caregiver involvement, and more financial burden were associated with greater caregiver strain. Thus, despite high rates of advance directives and hospice enrollment, perceived symptom distress was high for a subset of decedents, and caregiver strain was common.

Steinhauser et al.99 conducted a cross-sectional, stratified random national survey of 340 seriously ill patients, 332 recently bereaved family members, 361 physicians, and 429 other healthcare providers (nurses, social workers, chaplains, and hospice volunteers) to determine the factors considered important at the end of life. Twenty-six items consistently were rated as being important by greater than 70% of respondents, including pain and symptom management, preparation for death, achieving a sense of completion, decisions about treatment preferences, and being treated as a “whole person.” Results also highlighted differences among the respondent groups. Eight items received strong endorsement from patients but less from physicians (p<.001), including being mentally aware, having funeral arrangements planned, not being a burden, helping others, and coming to peace with God. Ten items had broad variation within as well as among the four groups, including decisions about life-sustaining treatments, dying at home, and talking about the meaning of death. Participants ranked freedom from pain most important and dying at home least important among nine major attributes. The findings from this study suggest that quality end-of-life care is a dynamic process that is negotiated and renegotiated among patients, family and healthcare professionals, a process moderated by individual values, knowledge, and preferences for care.

Fisher et al. conducted a retrospective cohort study using Medicare data including the Medicare Current Beneficiary Study (MCBS) to measure satisfaction.115 They constructed retrospective cohorts of patients hospitalized with hip fracture, colorectal cancer, and acute myocardial infarction. As the main regressor of interest, they considered the End of Life Expenditure Index (EOL-EI) to evaluate whether higher resource utilization at the end of life was associated with beneficial patient outcomes. This study found no association between higher expenditures for end-of-life care in these chronically ill Medicare beneficiaries and satisfaction as determined by 20 items from the MCBS.

Qualitative Studies Evaluating Satisfaction with Palliative Care

We identified 32 qualitative studies that specifically reported satisfaction related to care of the patient at the end of life.23, 24, 116–145 All of these studies reported the importance of health care in relationship to aspects of quality of life, quality of the dying experience, or satisfaction with care. The majority (20/31) examined the experience of patients, but 11/31 examined the experience of caregivers, and 6/31 examined perceptions of end-of-life care from the providers' perspective. Even among studies that incorporated multiple viewpoints, few explicitly compared patient, caregiver, and professional providers' perspectives.133 Most qualitative analyses employed either focus groups or unstructured interviews. With regard to settings, the most frequently studied settings was at home, whether in formal home care or not. We noted relatively few studies that incorporated participants or examined the end-of-life experience in nursing homes.24, 119 or that were relevant to satisfaction with end-of-life care in ICUs,123 although this may be related to our initial exclusion criteria (e.g., excluding cases of sudden, traumatic death). Most studies did not focus on specific diseases, and the majority of studies with a disease-specific focus examined aspects of cancer care rather than patients with other conditions.24, 125, 133, 137, 139, 146

One study that compared CHF with cancer125 noted important differences in the experience of medical care between these conditions. This study suggested the particular importance of information provision in CHF because patients are not ordinarily “expected to die.” Thus, prognosis is not discussed, and providers have little stimulus to acknowledge that advanced CHF will be fatal. CHF patients' care arose almost entirely from a medical model focused on treatment. Patients with cancer receiving treatment experience a rapidly changing clinical condition emphasizing a high need for coordination, and the value of being closely connected to supportive resources. CHF patients experience relatively stable but prolonged functional disability generating a need for support, but such services were infrequently available, at least compared to their availability for patients with cancer. Patients described the relative importance of various symptoms (a feeling of “drowning” in CHF vs. pain in cancer). Vig et al. also examined the quality of life and death for heart disease and cancer patients in an ambulatory setting.137 This study did not find differences in themes, but these interviews were less about the experience of health care than about overall aspects of living and dying.

In the aggregate, this group of qualitative studies shares a strong and striking sense of common themes related to important aspects of health care for people living with serious, eventually fatal conditions. These themes were repetitive across all the studies that examined the experience of patients and caregivers broadly and emphasized

  • professional competence in symptom management
  • continuity and coordination of multiple providers and across settings
  • responsive, flexible care that is available and adaptable to changing clinical needs
  • adequate provision of information about disease course, prognosis, and treatments
  • care from all providers that is empathic and that respects the individual as a person
  • spiritually supportive care and environments
  • adequate practical support for patients and caregivers in the home environment and informational support for practical planning in hospital and institutional settings.

Summary of the Relationship of Satisfaction to Other Measures of Process and Outcome

Several studies described in the context of our systematic reviews noted the association between satisfaction and interventions that improved communication or addressed other interpersonal aspects of care.90, 82, 84, 91 Other important processes or attributes of care that were highlighted by the interventional literature include the relationship of pain management, practical support, enhanced caregiving, and provider accessibility to satisfaction.83, 84, 147 The observational literature was similarly supportive of the importance of these indicators and their relationship to satisfaction. The observational literature adds to our understanding of these relationships by illustrating how these specific processes or attributes of care are helpful in distinguishing healthcare performance in different settings.72, 74

The qualitative literature suggests some important insights related to patient perception of care at the end of life. To the extent that satisfaction measurement reflects subjective perception of care, these qualitative data endorse the fact that patients and caregivers positively regard many of the attributes typified by palliative care (e.g., underscoring the importance of pain and symptom management, continuity, responsiveness, adequate information, respectful empathic, spiritually supportive care, and practical support). To the extent that interventions successfully target them and satisfaction measures embody these domains, they are likely to detect positive effects. In fact, this seems often, but not uniformly, to be the case in the interventional literature. In addition to measurement, as our findings suggest, the qualitative literature also supports the idea that this relationship between interventions and satisfaction could be confounded by other factors including differences in patient, caregiver, or healthcare settings.

B. Key Question 1b. What is the reliability and validity of specific instruments for measuring quality of life or quality of care at end of life?

Measurement of Patient and Family Outcomes

Our literature search identified one comprehensive systematic review of measures relevant to end-of-life care that Teno has published on the World Wide Web.30 The Toolkit of Instruments to Measure End of Life Care (TIME) project, last updated with a literature review current through 2000, created a web-based resource of patient-focused, family-centered instruments that address the needs and concerns of patients and their families at the end of life (see Methods). The Toolkit is a comprehensive list of the highest quality measurement tools for evaluating end-of-life care from the perspective of patient-focused, family-centered evaluation. The Toolkit organizes measures into 11 domains:

  • Pain and other symptoms
  • Emotional and cognitive symptoms
  • Functional status
  • Survival time and aggressiveness of care
  • Advance care planning
  • Continuity of care
  • Spirituality
  • Grief and bereavement
  • Patient-centered reports and rankings (i.e., satisfaction) with the quality of care
  • Caregiver well-being
  • Quality of life.

The Toolkit website, www.chcr.brown.edu/pcoc/bibliographies.htm, gives an extensive summary of the 35 recommended instruments,30 including reports of reliability and validity. Measures that the review process labeled as being only potentially relevant are listed on the web site with a one-sentence summary and corresponding reference. The Toolkit has a number of limitations. Its search terms were limited and inclusion criteria focused on measures that were accessible and easy to use. These criteria suggest that the Toolkit could have missed some important measurement tools for research. The Toolkit omitted clinician/provider focused issues and evaluation of quality end-of-life care from perspectives other than patient and family, even though those perspectives might also inform evaluations of the quality of end-of-life care. Nevertheless, the Toolkit is a remarkable and widely used working document, and wide use is likely to have led to reasonably broad coverage of measurement instruments.

Literature Review of Measures

Given the availability and quality of the Toolkit, our review focused on the literature after 2000 or on reports that were not identified in the Toolkit search. We searched especially for the development of new measures and for reports that describe reliability and validity data on specific instruments. We identified 48 new measures that supplement the Toolkit. Appendix H2 provides detailed validity and reliability data for measures we identified that supplement the 35 recommended Toolkit measures (the extensive data on reliability and validity testing summarized in the Toolkit website was not reproduced in this report). Our discussion below is organized in a similar fashion to the Toolkit. We highlight measures that fit best within the discreet domains as used in the Toolkit, but we grouped together multidimensional measures of quality of life, quality of care, and satisfaction with care. We report on measures to evaluate both overall quality of life and quality of care as well as specific domains relevant to both.

In the course of identifying all citations relevant to measurement, we also identified a number of citations that are important to understanding the application of measurement tools. This literature is not strictly within the scope of the RFTO, which focused on the reliability and validity of measurement tools themselves, but reports of the use of the measurement tools are important to understanding the application of the measures we identified and to assessing the implications for research and research priorities in the field. For that reason, we have included an accounting of these citations as a separate Appendix H1 A summary of the literature describing the properties and psychometric evaluation of measures is provided at the end of this section.

Multidimensional Measures: Quality of Life, Quality of Care and Satisfaction

Measures of Quality of Life

The Toolkit30 reviewed 41 measures of quality of life and recommended four that have detailed data on validity and reliability: McGill QOL Questionnaire (MQOL),148 Missoula-VITAS QOL Index (MVQOLI),149 European Organization for Research and Treatment Core Quality of Life Questionnaire version 3.0 (EORTC QLQ C-30), and the Functional Assessment of Cancer Therapy (FACT)/Functional Assessment of Chronic Illness Therapy (FACIT Fact-G).

The EORTC QLQ-C30, extensively described in the Toolkit, was evaluated in a palliative care population. Validity testing included generally moderate, statistically significant interscale correlations; discrimination by functional status; responsiveness to changes in health status over time and to palliative treatment. Factor analysis showed six factors, and Cronbach's alpha ranged from 0.56 to 0.79.150 A second article reported psychometric data in lung cancer for the EORTC QLQ-C30 demonstrating Cronbach's alpha overall = 0.93, subscales = 0.69 to 0.89 (7 of 12 subscales > 0.80).151 This same longitudinal study reported supplementary data on the Duke-UNC Social Support Scale in this population; Cronbach's alpha overall = 0.94, subscales = 0.88 to 0.92.151

The Brief Hospice Inventory, developed for use in hospice patients, showed two factors in factor analysis; Cronbach's alpha ranged from 0.84 to 0.94.152

The Hebrew Rehabilitation Center for the Aged index (HRCA-QL) index is a version of the Spitzer Quality of Life Index adapted for patients with advanced cancer. For criterion validity, it showed correlations with the Karnofky Performance Scale and an Instrumental Activities of Daily Living index. Cronbach's alpha was 0.7, and test-retest and inter-rater reliability were good. Scores declined as patients became closer to death or health status changed.153

The McMaster Quality of Life Scale was designed for use by proxies or patients. Concurrent (correlation with Spitzer Quality of Life and construct (those able to rate it themselves scored higher than those who could not) validity were tested. Intra-observer and inter-rater reliability were high, and the measure was responsive to perceptions of change in clinical status.154

The Palliative Care Quality of Life Instrument includes 28 items in six scales. Validity testing included face, construct (correlation with AQEL), criterion (ability to predict independent criterion variables, convergent and discriminative. Patients with better and worse Eastern Cooperative Oncology Group (ECOG) status showed significant differences, as did patients before and after treatment. Internal consistency and test-retest reliability were also high.155

Giorgi et al. describe comparisons between a linear analogue scale (LAS) for measuring quality of life in cancer patients and results with categorical unvalidated assessment that was not included in the Toolkit.156 Correlation between the LAS and a performance status measure is 0.46 and the questionnaire and performance status correlation is 0.38. Internal consistency testing for LAS reveals a poor Cronbach alpha for the LAS.

Green et al. proposed a chronic heart disease specific tool to measure physical limitation, symptoms, QOL, social interference, and self-efficacy.157 The Kansas City Cardiomyopathy Questionnaire is a self-administered, 23-item tool that was compared to the SF-36 and Minnesota Living with Heart Failure Questionnaire (LiHFe). Convergent validity was 0.46–0.74 across seven domains. Physical limitation subscale was correlated to the six-minute walk (r=0.48), SF-36 (r=0.84), and LiHFe (0.65). Reliability testing demonstrated Cronbach's alpha of 0.62–0.95 across seven domains; test-retest at three months for patients without exacerbations changed only 0.8 to 4 points on the 1–100-point scale.

Higginson et al. is cited in the Toolkit reporting validity testing of the Support Team Assessment Schedule (STAS) for seven of STAS's 17 items.158 The measures have items scaled 0–4 and use ten items to rate patient and family status and seven items to rate services delivered. Agreement on the seven items for patient and staff (n=62–78) ranged from kappa 0.12 to 0.78, total score Spearman rho 0.66; kappa for family and staff (n=58–67) ranged from -0.06–0.51, total score Spearman rho 0.44. Carson et al. report validity and reliability testing of the STAS in Canada in an acute care oncology unit and a palliative care unit.159 Validity data by comparison to patient ratings resulted in an overall r=-0.09 for the palliative care team and r=0.28 for the oncology team; comparison to family ratings resulted in overall r=0.38 and r=0.37, respectively (all p>0.05). Inter-observer correlations ranged from 0.27 to 1.0 and intra-observer correlations from -0.33 to 0.88. Test-retest correlations were 0.50 for palliative care team and 0.71 for oncology team.

Steinhauser et al. describe the Quality of Life at End of Life (QUAL-E) instrument that consists of 24 items.160 Factor analysis revealed five domains: life completion, relationships with the healthcare system, preparation/anticipatory concerns, symptom impact, connectedness and affective social support; Cronbach's alpha ranged from 0.6 to 0.84 for the subscales.

The Life Evaluation Questionnaire (LEQ) was described in 1996 but was not reviewed in the Toolkit.161 The LEQ is a self-administered, 121-item measure across five subscales (freedom, appreciation of life, contentment, resentment, social integration) that was developed in incurable cancer patients in both outpatient and inpatient care settings. Salmon et al. report convergent validity to the RSCL that ranged from 0.01 to 0.62 (sufficient only for freedom, resentment, and social integration); convergent validity to MacAdam and Smith Support scale that ranged from 0.02 to 0.62; Cronbach's alpha for freedom = 0.70, appreciation of life = 0.76, contentment = 0.76, resentment = 0.85, social integration = 0.78); test-retest in 40 individuals at two to three days were freedom r=0.80, appreciation of life r=0.91, contentment r=0.77, resentment r=0.92, social integration r=0.84).161

Measures of Quality of Care and Satisfaction

The Toolkit30 reviewed 20 measures and recommended the Medical Outcome Study Satisfaction Survey, Toolkit of Instruments to Measure End of Life Care Bereaved Family Member Interview, Picker-Commonwealth Survey, and FAMCARE. Our literature search identified six additional measures in the domain of satisfaction or quality of care that also had available psychometric information. We identified one additional validation study for the FAMCARE scale109 that added data to the Toolkit citation.30, 162 Kristjanson et al. report an inter-item correlation criterion (minimum 50% with r = 0.3 to 0.7) for 18 of 20 items, item correlation to total score of 0.4 to 0.76 for 15 of 20 items, and a Cronbach's alpha = 0.90.109 The authors also reported on two additional measures evaluated concomitantly, the Family Assessment Device (FAD) and the F-Care Expectations & Perceptions Scales. The FAD is a 12-item scale assessing family functioning; inter-item correlations met criterion (minimum 50% with r = 0.3 to 0.7) for 12 of 12 items; item correlation to total score was 0.4 to 0.75 for 12 of 12 items; Cronbach's alpha = 0.93.109 The F-Care Expectations Scale assesses family members' care expectations and was reported to have inter-item correlations at criterion for 13 of 16 items; item correlation to total score of 0.4 to 0.72 for 12 of 16 items, and Cronbach's alpha = 0.88.109 The F-Care Perceptions Scale assesses family members' care perceptions; inter-item correlations met criterion for 18 of 21 items; item correlation to total score was 0.4 to 0.72 for 13 of 21 items; Cronbach's alpha = 0.86.109

The Toolkit After-Death Bereaved Family Member Interview30, 31 is a telephone survey for family members and has versions for hospice, nursing homes, and hospital deaths; U.S. norms are available. There are eight domains. Scales were moderately correlated with overall satisfaction and with the corresponding individual rating question for the construct. Cronbach's alpha was greater than 0.7 for scales with more than three items, and test-retest reliability was high. Families of those who died in hospice reported better care than families of those who did not.31

The Quality of Dying and Death (QODD) instrument is a 31-item family after-death interview across six domains;163 it includes an assessment of frequency and a linked quality ratings; construct validity r=-0.52 against the Memorial Symptom Assessment Scale (MSAS), r=-0.47 MSAS psychological subscore, r=-0.42 MSAS physical subscore; discriminative study with independent symptom questionnaire significant at p<0.01, preferences at p<0.01, and communication p<0.001; correlation to global rating of last seven days of life r=0.55, moment of death r=0.51 (two factors explaining 38% of QODD variance); overall 31-item QODD Cronbach alpha = 0.89.164 A separate report demonstrated Cronbach alpha = 0.96 for a 14-item nurse version of the QODD.165 A study in the of the after-death QODD adapted for the intensive care unit demonstrated interobserver reliability 0.44 for the overall ICU-QODD score (23 item ICU version); components ranged from an intra-class correlation (ICC) of 0.15 to 1.0 for frequency components (mean 0.54), and ICC 0.16 to 0.59 for quality rating component (mean 0.32).166

The QUEST includes four scales for evaluating quality of end-of-life care and satisfaction with treatment: MD care, MD satisfaction, RN care, and RN satisfaction. Face (expert review), construct (moderate correlation with Patient Satisfaction Index), and correlation between subscales and with unrelated constructs were all tested. Test-retest kappas were 0.43–0.86, and Cronbach's alpha was 0.83–0.95. Scores were negatively correlated with symptoms and lower for those with “do not resuscitate” (DNR) orders.167

A four-item measure of patients' assessment of the quality of communication about end-of-life care was highly correlated with overall satisfaction with care. Those with higher-rated communication had clinicians more likely to know if the patient had a durable power of attorney, and Cronbach's alpha was 0.81.168

The WALT measures Willingness to Accept Life-sustaining Treatment. It was reviewed for face validity by patients and experts, and showed correlation with a simple measure of preference. Inter-rater and test-retest reliability were good, and scores were associated with age, ethnicity, and functional impairment in a moderately ill population.169

A relatives' patient management questionnaire was developed to assess families' attitudes, perceptions, and patterns of choice in the management of terminal cancer patients. It includes 21 items and five scales. Construct and discriminant validity were demonstrated through interscale and interitem correlations, and Cronbach's alphas were 0.5–0.69.170

Volicer et al. report the evaluation of three scales for dementia patients including a caregiver satisfaction scale, the Satisfaction With Care at the End of Life in Dementia (SWC-EOLD).113 The ten-item scale was shown to have one factor; item-total correlations range 0.33 to 0.79; Cronbach's alpha = 0.90.

A postal questionnaire to examine caregiver satisfaction with palliative care was described by Jacoby et al.112 This 89-question after-death postal survey of caregivers demonstrated discriminant validity tested with 36 attitudinal questions when health problems identified—only four were significant by Chi square; convergent testing was reported in tabular form in the reference; Cronbach's alpha = 0.68 to 0.84 across seven subsets.

We also identified several needs assessment tools, a domain that measures an element of patient-centered care but was not addressed in the Toolkit. The Cancer Patient Needs Survey has 51 items in five categories, including coping, help, information, work, and cancer shock. Different scores were found for hospice and clinic patients, and Cronbach's alpha was 0.91; this questionnaire was developed for the general cancer population.171

The Concept of a Good Death measure includes 17 descriptive statements of relevant concepts in three subscales: closure, personal control, and clinical criteria. Factor analysis showed three subscales, there was small-to-moderate association with other measures, and test-retest reliability was high. Scores were related to age, gender, and ethnicity.172

Emanuel et al. report the rigorous development of a 13-question clinical screening instrument for terminal care needs, the Needs at the End-of-Life Screening Tool (NEST).173 This multidimensional screening tool was developed from factor analysis of a 135-item survey administered to 988 dying patients. The measure requires further validation and reliability testing.

Finally, we identified two tools for evaluating the quality of palliative care, one for use by both patients and staff and one for use by staff only. Hearn et al. reported development and testing of the Palliative Care Outcome scale (POS).174 The measure was developed by systematic literature review and underwent refinement by a multidisciplinary advisory group over several iterations of pilot testing. The measure was specifically developed as an outcome measure for the quality of end-of-life and palliative care for use in hospice patients. The measure includes 12 items, most using a 0–4 scale and consists of two parts, one patient self-administered questionnaire and one palliative care staff responses. Validity testing was performed across eight sites in England and Scotland with 148 patients completing evaluation. On average, the measure was completed in less than ten minutes for each type of respondent. Reliability testing included test-retest, internal consistency (Cronbach alpha for patient = 0.65, staff = 0.70), and a comparison of staff to patient responses. Validity testing included assessments of face validity and change over time; construct validity achieved a Spearman's rho 0.43–0.80 against ETORTC QLC-C30 and Support Team Assessment Schedule (STAS). The Resident Assessment Instrument for Palliative Care was designed for clinician assessment in nursing homes. It builds on the standard RAI, and includes nine domains. Intra-observer kappas were 0.77–0.9.175

Measures Related to Other Specific Domains

Measures of Pain and Other Symptoms

Sixty-four measures were reviewed in the Toolkit30 and five measures were recommended for assessing either pain or overall symptoms: McGill Pain Questionnaire (MPQ),176 Wisconsin Brief Pain Questionnaire, Memorial Pain Assessment Card, Edmonton Symptom Assessment System (ESAS,177 and Memorial Symptom Assessment Scale (MSAS).178

With regard to the MSAS, we identified a validation trial for the MSAS in non cancer patients where convergent validity to the Piper Fatigue Scale ranged from r=0.15 to 0.56 for cancer patients and 0.29 to 0.61 for non-cancer patients (best for behavioral and sensory subscales of the PFS); factor analysis yielded one psychological factor and one physical symptom with three subgroups; Cronbach's alpha = 0.85 in cancer patients (n=66) and 0.77 in non-cancer end-stage group (n=69).179 Also, Chang et al. report univariate correlations for the MSAS to RAND Mental Health Inventory (MHI) well-being scale -0.60 (-0.53 to 0.66 for three subscales), MHI distress 0.65 (0.48 to 0.80), Functional Living Index-Cancer (FLIC) -0.78 (-0.61 to -0.78, subscales of FLIC range -0.45 to -0.73), SDS 0.79 (0.57 to 0.81), and Karnofsky -0.58 (-0.31 to-0.65); the physical and global distress index subscales performed better than the psychological symptom subscale.180

More recent studies of the ESAS have shown that telephone administration was possible in 62% of palliative care patients.181 correlation to MSAS Global Distress r=0.73; concurrent validity ESAS summary distress score to MSAS demonstrated: TMSAS scale (0.72), Global Distress Index (GDI) (0.73), physical symptom subscale (0.74), and psychological symptom subscale (0.56); ESAS summary distress score to FACT demonstrated: physical well-being subscale (-0.75), sum QOL (-0.69), functional well-being (-0.63), emotional well-being (-0.52) and social/family well-being (-0.25); all item correlations reported as significant; calibration studies showed overlap for median values within scales for all items; Cronbach alpha 0.79; test-retest Spearman correlation 0.86 at two days and 0.45 at one week; all items significantly correlated at two days (r = 0.43 to 0.86) but at one week only pain (0.75), activity (0.65), depression (0.54), shortness of breath (0.53) and distress (0.45) were significantly correlated.182 We identified seven additional measures with descriptions of psychometric properties in the current effort.

The Cambridge Palliative Assessment Schedule (CAMPAS-R) was developed for palliative care in primary care. Patients rate physical and psychological symptoms and their caregiver's psychological symptoms on a visual analog scale. Face and content validity was tested with patients, physicians, and nurses; criterion validity showed correlation with the EORTC and HADS for some items but not for others; and discriminant validity was shown through significant differences between patients who did and who did not survive. Cronbach's alpha for correlation between symptoms was 0.77–0.8.183

The Symptom Monitor is a ten-item diary for physical symptoms, developed for feasibility in patients with advanced illness. Inter-rater intra-cluster correlations were >0.75.184

Two validation reports were identified for the Lung Cancer Symptom Scale (LCCS).185, 186 The measure uses nine patient-scored visual analogue scales and six observer-scored four-point scaled items to measure symptoms prevalent in lung cancer. Construct validity against Karnofsky was 0.15–0.63 across items (symptomatic distress 0.49, effect on activities 0.63, QOL 0.43).185 Criterion validity was reported (patient scale and observer scale, respectively) relative to the Karnofsky (r=0.63, NA), Sickness Impact Profile (SIP) (0.40, 0.56), Profile of Mood States (POMS) (0.67,0.54), American Thoracic Society Questionnaire (ATS 29) cough (0.56, 0.65) and dyspnea (0.46, 0.64), and McGill Pain Questionnaire-short form (SF-MPQ) (items range 0.51–0.67). Internal consistency was done to Brief Symptom Inventory (BSI) (r=0.93), SIP (r=0.94), POMS (r=0.94), SF-MPQ (r=0.91, r=0.64–0.74 for three components). Hollen, et al. describe normative data and trends for QOL in stage III and IV lung cancer using the LCCS.187

Sarna et al. applied the Symptom Distress Scale (SDS) to female lung cancer patients.188 The 13-item, self-report scale was developed and modified in the 1970s to 1980s. In this study, factor analysis with principal components and varimax rotation resulted in a five-factor model explaining 65% variance. The study provides only limited validity data beyond factor analysis but notes negative correlations of certain items to parts of Karnofsky Performance Status (r= -0.27 to -0.48) and an overall r=-0.58.

Warden et al. reports psychometric testing for a novel, disease-specific measure, Pain Assessment in Advanced Dementia (PAINAD).189 The five-item, observer assessment demonstrated convergent validity to Discomfort Scale—Dementia Alzheimer's Type (DS-DAT) and Discomfort Scale—Visual Analogue Scale (DS-VAS) (r=0.76, n=19) and PAIN-VAS (r=0.75, n=18). Factor analysis was noted and also done in different conditions (r>=0.82 for pain with activity). Cronbach's alpha ranged from 0.57 to 0.83 in multiple phases of the study.

Volicer et al. reports the evaluation of two symptom scales for dementia patients: the Symptom Management at the End of Life in Dementia (SM-EOLD) and the Comfort Assessment in Dying With Dementia (CAD-EOLD).113 The SM-EOLD is a nine-item scale shown to comprise two factors; item-total correlations range 0.18 to 0.66; correlation for symptom items on CAD-EOLD r = 0.475 to 0.559; Cronbach's alpha = 0.78. The CAD-EOLD is a 14-item scale with four subscales (physical distress, dying symptoms, emotional distress, well-being); item-total correlations range 0.39 to 0.79; correlation for symptom items on SM-EOLD r = 0.475 to 0.559; Cronbach's alpha = 0.85 overall; subscales (physical distress r=0.74, dying symptoms r=0.70, emotional distress r=0.82, well-being r=0.80).

Measures of Emotional and Cognitive Symptoms

The Toolkit30 reviewed 41 measures and recommended five: Profile of Mood States, Memorial Symptom Assessment Scale, Center for Epidemiolgic Studies Depression Scale (CES-D), and RAND Mental Health Inventory (MHI-5). Our literature search identified six reports describing measures in the domain of emotional symptoms.) A single-item screening for depression, “Are you depressed?” correctly identified depression in all 24 terminally ill patients evaluated.190

The communication capacity scale is a five-item clinician rating scale developed for palliative care populations. Principal components analysis demonstrated only one component, and the scale was highly associated with cognitive items on the MDAS and DRS (delirium rating scale) and not with irrelevant items. Cronbach's alpha was 0.96 and inter-rater kappa was excellent.191

The agitation distress scale is a six-item clinician rating scale. Principal components analysis demonstrated only one component, and the scale was highly associated with agitation items on the Memorial Delirium Assessment Scale (MDAS) and DRS and not with irrelevant items. Cronbach's alpha was 0.91 and inter-rater kappa was excellent.191

Kurlowicz et al. evaluated the 19-item clinician interview Cornell Scale for Depression in Dementia (CSDD) in a study of 642 nursing home patients.192 Oblique rotation four-factor matrix and inter-factor correlation analysis resulted in a 16-item, four-domain measure. Criterion validity was performed and reported. Internal consistency revealed a Cronbach alpha of 0.76.

Hopwood et al. applied two previously developed measures to a sample of 204 patients with breast cancer.193 Only weak validity metrics are reported for the Hospital Anxiety and Depression Scale (HADS) and the Rotterdam Symptom Checklist (RSCL).

Measures of Functional Status

The Toolkit30 reviewed 15 measures and recommended six within this domain: Index of Independence in ADLs, Barthel Index, Physical Self-Maintenance Scale, Rapid Disability Rating Scale, Stanford Health Assessment Questionnaire, and FIM™ Instrument. Our literature search identified four reports describing measures in the domain of functional status with specific psychometric descriptions of measures. Two reports were refinements to the Edmonton Functional Assessment Tool (EFAT); the original measure was evaluated in the Toolkit and not recommended; however, the revision, EFAT-2, was not available at the time of the last Toolkit update.194, 195 EFAT-2 is a ten-item rating assigned by a professional grading symptoms and functions and assigns a summary functional assessment. The Cronbach's alpha was 0.86 and inter-rater correlation was 0.97 for self trained clinicians (n=2) and 0.95 for formal trained (n=2).194, 195 The measure was not correlated with pain but demonstrated discriminant validity in different groups based on discharge location.

Gerety et al. report on a 54-item measure to evaluate frail elderly individuals that requires calibrated specialized performance measuring equipment, the Physical Disability Index (PDI).196 They report discriminate validity against Folstein Mini-Mental State Exam (r=0.11) and convergent validity to the Physical Self-Maintenance Scale (r=-0.71) and Sickness Impact Profile (r=-0.59). Test-retest correlation in 36 patients at two to five days was r=0.97 overall, four subscales 0.92–0.96; inter-rater reliability coefficients ranged from r=0.81 to 0.99 except for the mobility scale which was r= -0.02 to 0.70.

Gloth et al. report on the Frail Elderly Functional Assessment Questionnaire (FEFA), which is a 19-item, interviewer administered tool for the elderly at very low activity levels.197 They report correlation to direct observation (r=0.90), Katz's ADL index (r=0.86), Barthel index (r=0.91), and Lawton's IADL index (r=0.67). Test-retest in 29 patients at a two-week interval revealed a kappa 0.82 overall; all items had kappas greater than 0.40 (0.45–0.91).

Measures of Survival Time and Aggressiveness of Care

The Toolkit30 reviewed four chart-based instruments and three prognostic tools. Several individual questions are recommended, but validity/reliability information on tools is not available. As described in the methods section, a review of prognostication and prognostic indices relevant to the definition of the end of life is included in Appendix A.

Measures of Advance Care Planning

The Toolkit30 reviewed and recommended the Toolkit of Instruments to Measure End of Life Care Bereaved Family Member Interview. Our literature search identified one additional report describing measures in the domain of advance care planning with specific descriptions of the psychometric properties of measures.

Koedoot et al. (2001) describe a measure not captured in the Toolkit that has applicability to advance care planning.198 The decisional conflict scale (DCS) is tested in a Dutch translation version for psychometric properties in a cancer patient group. The measure consists of 16 items, each scored on a five-point Likert scale, across three subscales (uncertainty, factors contributing, and effective decision-making). Construct validity among subscales was measured at r=0.58 to 0.76. Criterion validity on the uncertainty subscale was described as significant between certain versus uncertain group. Prior reliability testing was noted in the report demonstrating internal consistency (Cronbach alpha = 0.78–0.89) and test-retest reliability (r>0.80).

Measures of Continuity of Care

The Toolkit30 reviewed four measures and recommended the Picker-Commonwealth Single Item, Smith-Falvo Patient-Doctor Interaction Scale, McCusker Scale, and Chao Patient Perception measures. Our literature search identified no additional reports describing measures in the domain of continuity of care with descriptions of specific psychometric properties of measures.

Measures of Spirituality

The Toolkit30 reviewed 25 measures and recommended the Meaning in Life Scale, Spiritual Well-Being Scale, Spiritual Perspective Scale, Death Transcendence Scale, Death Attitude Profile, and Herth Hope Index. Our literature search identified one additional report describing measures in the domain of spirituality with descriptions of specific psychometric properties of measures.

The Santa Clara Strength of Religious Faith Questionnaire (SCSORF) 199 is a 10-item scale with good internal consistency (Cronbach's alpha = 0.95) and test-retest reliability (0.82) in a population with mainly early-stage breast cancer. Convergent validity was demonstrated through a strong correlation with intrinsic religiosity and moderate correlations with religious practice, perception of self as spiritual, and comfort derived from religion.

The 45-item Life Closure Scale was developed to measure psychological adaptation in the dying and tested in hospice patients. The content validity index, as assessed by experts, was 0.83, and Cronbach's alpha was 0.80.200

Measures of Grief and Bereavement

The Toolkit30 reviewed 24 measures and recommended the Grief Resolution Index and Anticipatory Grief Scale. Our literature search identified four additional reports describing measures in the domain of grief and bereavement that provided specific psychometric properties of measures.

The CBI (Core Bereavement Items) includes 17 items in three subscales. The measure was developed from the bereavement phenomenology questionnaire. Testing included face validity, factor analysis, and discriminant validity for time and group effects; Cronbach's alpha was 0.91.201

The Hogan Grief Reaction Checklist (HGRC)202 is a 61-item measure across six constructs (despair, panic behavior, blame and anger, disorganization, detachment, and personal growth) that was developed in grieving adults from mixed sources. Hotgan et al. reported convergent validity to earlier measures in general grief that ranged from r=0.20 to 0.78 with significant correlations across subscales; discriminant validity in subset of mothers who experienced death of a child by different mechanisms and by timing of death; Cronbach's alpha overall was 0.90.

An eight-item adaptation of the Bereavement Risk Index showed significant differences in the Brief Symptom Inventory between low- and high-risk group, which were maintained for 25 months after death.203

Feldstein et al (1995) used the Grief Experience Inventory (GEI) measure in a study of oncology nurse grief and summarized the original validation data reported in 1985.204 The measure uses 102 yes/no statements in a self-administered inventory that is further scored into nine composite scales. Data reported includes discriminant validity between bereaved versus nonbereaved individuals at the significance level 0.001 on all subscales, test-retest coefficients 0.53–0.87, and internal consistency Cronbach's alpha = 0.52–0.84 on bereavement scales.

Measures of Caregiver Well-being

The Toolkit30 reviewed 53 measures and recommended the Caregiver Strain Index and Caregiver Reaction Assessment. Our literature search identified two additional reports describing measures in the domain of caregiver well-being with specific psychometric descriptions of measures.

Travis et al. describe the development of the Family Caregiver Medication Administration Hassles Scale designed to capture problems caregivers experience with assisting elderly with medications.205 The 24-item paper survey is designed to capture four subscales (information, safety issues, scheduling, and polypharmacy). Principal components and factor analysis was done (66.5% cumulative variance). Construct validity against the Medication Complexity Index (r=0.19) and modified Caregiver Strain Index (r=0.44) were reported. Test-retest at two weeks (n=53) correlated at r=0.84. Internal consistency was reported at 0.95 (0.800.92 across subscales).

The Cost and Reciprocity Index (CRI) (modified) includes 25 items in four subscales and was modified for use with hospice caregivers. Concepts include social support and conflict. Extensive testing was done with the original instrument in healthy populations; in this study, relations between subscales were consistent with the theoretical framework and Cronbach's alpha was 0.68–0.83.206

Other Measures

A number of measures were identified in our endeavor that did not specifically fit into any domains established by the Toolkit but may have applicability to end-of-life care research. Our literature search identified three reports describing measures in outside of the Toolkit domains with descriptions of specific psychometric properties of measures.

Kristjansson, et al. report on an index of social support developed from data gathered in the Canadian Study of Health and Aging (CSHA).207 The six-item measure was developed from factor analysis (item correlations 0.26 to 0.83) and item response theory (IRT) analysis for half the study population. External (construct and predictive validity on second half of study population), and IRT (r=0.53 to network size)/classical (r=0.61) comparison was done. Cronbach alpha = 0.76; IRT marginal reliability was 0.85.

We identified a number of clinical scoring tools. The Hospice Pressure Ulcer Risk Assessment Scale (HoRT) measures physical activity, age, and mobility. PPV was 50%, NPV 100%.208 The Clinical Dementia Rating Scale looked only at demographic, clinical, and prognosis to death correlation to CDR scores CDR correlates to death during follow up r=0.36.209

Fowell et al. report a novel application of an integrated care pathway (ICP) to gain quality of end-of-life care data.210 The investigators developed and employed the ICP across the healthcare system in Wales and captured data about the end-of-life care experience from variance sheets that were required when the care provided deviated from the expected course of care delineated in the ICP guideline. Although not a validated measure, this quality improvement method provided significant evaluative data about the care of the dying across the healthcare system in Wales.

Summary of Measures

Many new instruments have been developed or have undergone further evaluation in end-of-life settings since the last Toolkit update in 2000, particularly in the domains of quality of life, quality of care and satisfaction, and pain and physical symptoms. However, many articles did not report a theoretical framework or a careful development process, and reliability and validity testing was often limited in scope. Since patients at the end of life often receive care in multiple settings, instruments that are useful longitudinally and in hospitals, intensive care, outpatient settings, nursing homes, and at home are essential for comprehensive evaluations, but most instrument evaluations were limited to a single setting. End-of-life issues and symptoms often also vary substantially with cultural backgrounds. However, development, reliability, and validity studies addressing different populations were also very uncommon. Finally, although end-of-life care varies substantially among different regions of the United States, most studies were conducted in a single center, often in tertiary care settings.

Many commonly used instruments have not been evaluated in end-of-life populations, where psychometrics, burden, or applicability may be very different. Few instruments were developed for or tested specifically in non-cancer populations. In certain areas, particularly continuity, advance care planning, and aggressiveness of care, we found few instruments tested in the end-of-life population. In other areas, such as quality of life or satisfaction, lack of theoretical frameworks, limited evaluations, and lack of consensus often make it difficult for researchers to choose appropriate instruments. Finally, few instruments have been developed or evaluated for the purpose of clinical practice, evaluation studies, or quality assessment or improvement interventions. Improving the quality of the intervention literature requires further evaluation of carefully developed instruments and development or testing of continuity, advanced care planning, and aggressiveness of care specifically for the purpose of evaluating interventions.

C. Key Questions 2 and 3

2. What patient, family, and healthcare system factors are associated with better or worse outcomes at end of life?

3. What processes and interventions are associated with improved or worsened outcomes?

Elements associated with patient experience: symptoms of pain, dyspnea, depression and anxiety, and behavioral issues in dementia

We reviewed 27 systematic reviews or meta-analyses because they addressed selected symptoms of a palliative care population. Of those considered, we identified 12 that addressed the project questions and met implicit quality criteria. Two of the reviews included here focused specifically on a cancer population, one on patients with COPD, three on patients with dementia, and another six did not limit their reviews to only one disease cohort. In our review, we went beyond the systematic reviews by including intervention studies addressing our chosen symptom topics if those studies were not included in the systematic reviews. In total, we identified an additional 18 intervention studies. Finally, we explored the observational literature that addressed selected topics. Specifically, we identified prospective, observational cohort studies addressing any of our selected symptom topics and that also presented data separately by race, selected disease cohorts, or selected sites of care. In total, we reviewed 14 observational studies.

The remainder of this section summarizes the systematic reviews, meta-analyses, and intervention studies for each of the symptom groups separately: pain, dyspnea, depression and anxiety, and behavioral issues for dementia patients. A discussion of all the observational studies is presented at the end of the whole section. Summaries of the association of patient, family, and health system factors to symptoms and the effectiveness of interventions in improving symptoms are found at the conclusion of Chapter 3.

Pain

Systematic Reviews and Pain

Six systematic review publications reflecting five separate reviews were identified that addressed the topic of pain.34, 72, 74, 211, 214, 216 The systematic reviews by Higginson et al.74, 214 include the original report and a peer-reviewed publication from that report and are treated as one review. Two of the systematic reviews addressed pain specifically in cancer populations,34, 72 one included a meta-analysis of the literature on the effects of palliative care teams on pain,74, 214 and two reviews examined the literature on complementary and alternative medicine (CAM) or otherwise deemed “non-pharmacologic” interventions to address pain and other symptoms.211, 216

One of the more recent reviews, conducted by Gysels and Higginson,72 examined the literature on improving support for and palliative care of cancer patients. This review considered studies published before March 2003, including randomized or quasi-randomized controlled studies, non-randomized controlled studies, observational studies and systematic reviews. This review was not organized specifically around pain or other symptoms; however, many symptom-related studies were reviewed in the context of other topic areas, including “coordination of care,” “user involvement in planning, delivering, and evaluating services,” “psychological support services,” “general palliative care services,” “specialist palliative care services,” “rehabilitation services,” and “complementary therapy services.” In total, 44 symptom-related studies were identified, 27 of which addressed pain. Among the studies reviewed was the systematic review by Pan et al.216 on complementary and alternative medicine (CAM), a study separately identified during our search of the literature and which will be described below. Of the 27 studies identified on pain, nine were randomized or quasi-randomized controlled trials and 12 were observational studies. The remainder were qualitative studies (2), a systematic review, and three studies with unclear study designs. Sample sizes in these studies ranged from 9 to 695. Interventions to address pain symptoms included clinical pathways and special clinical teams, education, hospice (either inpatient or outpatient), palliative care teams, specialized home care teams, and massage. Overall, the studies identified in this review reported beneficial positive results in which pain symptoms experienced by cancer patients were alleviated by the interventions. Of the 19 studies reporting beneficial results, 11 were observational studies. One of the qualitative studies identified substantial unrelieved pain in the sample included in its study. There were six studies that reported no significant difference in pain symptoms between the intervention and control groups or between baseline and follow-up. Five of these studies were randomized or quasi-randomized controlled studies and one was a prospective observational study.

A systematic review of the management of cancer symptoms, including pain, was conducted by the New England Medical Center Evidence-Based Practice Center for the Agency for Healthcare Research and Quality.34 This review considered the literature published in or before September 2001 that addressed the prevalence, assessment, or treatment of the selected symptoms. The report considered the full trajectory of disease rather than focusing on end-of-life care specifically. Given the focus of our review, we will only report on the findings from the review of studies related to treatment of pain. Only randomized controlled trials were accepted for this portion of the review. The authors of this report summarized the literature on the treatment of cancer pain published following the publication of a systematic review on cancer pain by Goudas et al.220 A total of 24 studies were identified; one addressed the relative efficacy of particular nonsteroidal anti-inflammatory drugs (NSAIDs) in comparison to other NSAIDs or placebo; six were identified that evaluated adjuvant analgesics in cancer pain management; six compared one opioid with another; five considered bisphosphonates in treating metastatic bone pain (comparing different doses or comparing to placebo) and six studies considered CAM treatments for managing cancer pain. Of these 24 studies, 14 interventions reported beneficial results. The six studies comparing different types of opioids, different dosages of the same opioid, or different means of opioid delivery did not report statistically significant results.

In an extensive review conducted by Higginson et al.,74, 214 the authors explored the role that palliative care teams play in affecting a number of symptoms in end-of-life care populations. The authors searched ten databases, the gray literature, journals, and the references of included studies. The most recent study was published in 2000. A total of 54 studies were identified after excluding case reports. The palliative care interventions identified in these studies included a number of settings: home care, hospital-based, combined home/hospital-based, inpatient unit, and integrated teams. A meta-analysis was conducted with a subset of 19 studies. The study designs included in this review were primarily prospective or retrospective/observational/cross-sectional. A meta-analysis of palliative care versus conventional care based on 13 studies reported an overall beneficial effect of palliative care teams on pain outcomes (OR: 0.38, 95% CI: 0.23, 0.64; odds ratio less than 1 means less pain). When the studies were stratified by study design, a significant effect on pain was only seen among the studies with non-randomized and observational/retrospective designs; there were three RCTs in this review (OR: 0.82, 95% CI: 0.52, 1.28), three non-randomized controlled trials (OR: 0.41, 95% CI: 0.30, 0.57) and seven observational/retrospective studies (OR: 0.30, 95% CI: 0.12, 0.74).

One systematic review produced by Health Canada addressed the symptoms of populations nearing the end of life.211 In this review, the authors focused on managing end-of-life pain and other symptoms through non-pharmacological means. The search incorporated the literature published through mid-2003 in nine databases, the gray literature, monographs, and policy statements. A total of 21 research articles were identified (6 individual studies and 15 reviews). Non-pharmacological treatments of pain were the subject of 17 out of 21 of the research studies reviewed. Topics included acupuncture, hypnosis, music therapy, relaxation, massage, imagery, therapeutic touch, magnets, transcutaneous electrical nerve stimulation (TENS), microcurrent electrical neuromuscular stimulator (MENS), radiation therapy, and pediatric palliative care; however, only a subset of these were the subject of intervention. Where an intervention was conducted, results were generally beneficial, but most studies were observational in design.

One study by Pan et al216 explored the role that complementary and alternative medicine (CAM) interventions might have in reducing or eliminating pain among palliative care populations. The authors searched six databases for CAM interventions, focusing on the following interventions: acupuncture, TENS, massage therapy, behavioral/relaxation therapy, music therapy, and psychological therapy. A total of 21 studies were identified in this review; eleven of these studies were RCTs, two were non-randomized trials, and eight were case studies. The most recent of the studies identified was published in 1998. A total of 14 studies addressed pain as the primary outcome of interest. Although the search criteria did not limit the review to cancer populations only, of the 14 studies, 12 included patients with cancer diagnoses only. One other study examined CAM interventions for pain in an HIV-positive population and another focused on a patient population that had received bone marrow transplantation. We summarize the findings from this review by type of intervention below, and further details regarding these citations can be found in the systematic review by Pan et al.216

TENS: In a double blind RCT of 15 hospice cancer patients, the authors of one study did not have enough power to detect differences on pain measurement between the intervention and control groups, however, overall quality of life improved among intervention patients. A prospective pre/post intervention study of 60 patients with cancer pain included a 2-week intervention with TENS and resulted in 28% of patients reporting an excellent response that decreased to 15% after three months. A case study of nine patients with advanced cancer identified improvement in pain for 66% of them and partial relief in 22% of patients. Another case study including 29 frail cancer patients evaluated the joint intervention of TENS with acupuncture and found that 62% of patients had pain relief and 27% had pain reduction.

Acupuncture: In one study of 92 cancer patients, an intervention of acupuncture for one to two weeks achieved pain relief for one month in all patients with mild to moderate pain and 72% with severe pain. Among 183 cancer patients in another study, 48% had pain relief for three days or more or experienced an increase in mobility after a treatment of acupuncture one to four times weekly. In the only RCT involving acupuncture, 239 HIV-positive patients were randomized to real or sham acupuncture twice weekly for six weeks followed by once weekly for another eight weeks. The study found no statistically significant differences in pain reporting between groups.

Massage: There was one RCT and two case studies that explored the role massage might play in reducing pain symptoms. In an unblinded RCT of 28 cancer patients, men had immediate pain relief lasting for one hour while women experienced no significant improvement in pain symptoms. In a case series of nine cancer patients, patients reported a reduction in pain symptoms following two consecutive 30-minute massages. In another case series of 103 cancer patients, massage plus aromatherapy promoted pain relief in 33% of participants.

Behavioral and Relaxation Therapy: In a case series of 58 hospice cancer patients, participants were referred to relaxation therapy. Approximately 38% of study participants reported reduced pain symptoms following the intervention. An RCT of 94 bone marrow transplant patients with oral mucositis reported pain improvement by relaxation and imagery.

Music Therapy: In an RCT of nine terminally ill cancer patients, no significant difference was reported among groups in pain relief following an intervention of music therapy, although the review authors suggest that there was not sufficient power to detect small differences in the outcome.

Psychological Therapies: An RCT of therapy with or without self-hypnosis was conducted among 58 women with advanced breast cancer. The authors of this study reported that therapy reduced pain sensation and suffering and self-hypnosis provided further relief.

Additional Interventional Studies of Pain

We identified an additional ten randomized clinical trial or controlled clinical trial intervention studies addressing pain in end-of-life or palliative care populations. Six of the studies were focused specifically on cancer pain. Due to the recent publication of systematic reviews addressing cancer pain specifically, we selected those studies for review here that were published between 2002 and 2004,221–226 with the exception of one study published in 1998 but not otherwise addressed by the reviews we identified.223 Of these six studies, one examined the role of hospital-based palliative care teams in improving symptoms of cancer patients, one included an aromatherapy massage intervention, two examined pain relief through medication (one with NSAIDs, one comparing opioids, and one comparing delivery method), and two examined the role of structured assessment on pain and other symptoms. An additional four studies focused on the treatment of pain in palliative care for non-cancer or mixed diagnosis populations. Three of these studies were published between 2002 and 2004.86, 227, 228 Another study, published in 1998, was not previously reported on in any of the systematic reviews we considered and is described here.229 One intervention compared different doses of the same opioid on pain, one looked at the role exercise plays in reducing pain among nursing home residents, one examined the influence of a more comprehensive and coordinated medical record on pain and other symptoms, and one explored the role of an outpatient palliative medicine consultation on various symptoms including pain. These ten studies have been organized into four categories loosely based on the intervention types, rather than by disease cohort. The categories are pharmaceutical interventions, system/institutional interventions, CAM, and exercise.

Pharmaceutical Interventions: In one study, Smith and colleagues225 conducted a randomized controlled trial of an implantable drug delivery system (IDDS) and comprehensive medical management versus medical management alone (control) in 200 outpatients with cancer (101 in intervention group; 99 in control group). While the IDDS and control groups had the same results in terms of pain reduction (≥ 20% reduction in pain as measured by a 100-point VAS) and six-month survival, this finding is limited by a baseline pain assessment for both groups, which ensured some therapy for the control group. Also, the findings are confounded by the longer survival of the intervention group. There was a 50% reduction in toxicity scores for the intervention group as compared to 17% reduction in the control group (p=0.004). The Jadad score for this study was 3.

Buprenorphine is an opioid analgesic that has been mostly available in sublingual and parenteral formulations. In the study by Sittl and colleagues,228 the authors examine the efficacy and tolerability of transdermal buprenorphine. A randomized, double blind controlled trial of 157 patients with cancer- and non-cancer-related pain compared the efficacy and tolerability of transdermal buprenorphine in three doses (35.0, 52.5, and 70.0 μg/h) plus placebo. Patients received a new patch every 72 hours for up to 15 days and were allowed to use sublingual buprenorphine tablets for rescue analgesia. The lower doses of transdermal buprenorphine produced higher response rates (measured as needing ≤ 1 rescue analgesia pill per day) than placebo at 35.0 and 52.5 μg/h (p=0.032 and p=0.003, respectively). There were no significant differences between the largest dose and the placebo. The Jadad score for this study was 3.

In a third drug study, Mercadante and colleagues226 examined the effect of ketorolac, a non-steroidal anti-inflammatory drug (NSAID) on morphine escalation in a randomized controlled trial. Patients with cancer-related pain (n=47) were randomized into two groups: the intervention group received ketorolac (60 mg/daily p.o.) in three doses with opioid escalation as needed and the control group were treated with opioid escalation only. Those in the intervention group used less morphine than in the control (p=0.003) and had less opioid escalation (p<0.0005). The mean weekly pain intensity was significantly less after three weeks in the intervention group than in the control (p=0.005). The Jadad score for this study was 3.

Systems/Institutional Interventions: In a study published in 1998, Latimer and colleagues229 investigated the effectiveness and efficiency of a patient care traveling record in palliative care. The authors randomized 61 patients cared for by a palliative care service to receive or not receive the patient care traveling record, a record of the patient's care from all sources that the patient could take with him/her to all appointments with providers including names of providers, next of kin, prior hospitalizations, medications, advanced directives, etc. Of the original sample, only 21 remained at the end of the follow-up period. Patients who used the traveling record had a larger reduction in reported pain at follow-up as compared to the control group; however, the difference was marginally significant (p=0.05). The Jadad score for this study was 2.

Building on the literature from Higginson et al.74, 214 are two recently published studies examining the relationship between palliative care team interventions and patient outcomes. In the study by Jack et al. published in 2003,221 the authors conducted a controlled clinical trial of hospital based palliative care teams with 100 cancer patients (50 in intervention, 50 in usual care control group) to improve pain and other symptoms. The intervention group had significantly better pain ratings than the control group at the second and third assessments (p=0.029 and p<0.001, respectively). The most recent study, published in 2004 by Rabow et al.,86 reports on a randomized controlled trial to understand the influence of an outpatient palliative medicine consultation team on symptoms in 90 patients (50 intervention, 40 control) with chronic heart failure, COPD, or cancer. There were no significant differences in patients reporting any pain or in their average pain score based on the Brief Pain Inventory. The Jadad score for this study was 3.

Complementary and Alternative Medicine: Soden and colleagues224 conducted a randomized controlled trial of aromatherapy massage versus massage only or no treatment (control) with 42 cancer patients. There were no significant changes in pain assessments between baseline and follow-up for any group, nor were there significant between-group differences in pain assessment. The Jadad score for this study was 5.

Exercise: Simmons and colleagues227 reported on a study that explored the effects of an exercise and toileting program on pain among 51 incontinent nursing home residents in a randomized controlled trial. The intervention included toileting prompts every two hours, five days a week, between the hours of 8:00 AM and 4:30 PM by nursing staff. During this same time period, either before or after toileting, the staff would provide assistance for the resident to walk, wheel, or at least perform sit-to-stand movements. This intervention did not result in significant differences in pain reports between the intervention and control groups. The Jadad score for this study was 1.

Two studies examined the role of structured assessment in improving care processes for cancer patients through better information collection and patient-provider communication. Sarna223 examined the efficacy of a structured symptom assessment on symptom distress in a randomized controlled trial. The study included 48 subjects with advanced lung cancer. Patients were randomized to structured assessment or usual care and assessed several times over a six-month period. A total of 21 patients remained in the study at six months. Pain symptoms (frequency and severity) did not significantly differ between the intervention and control groups across time. The Jadad score for this study was 2.

In a study published in 2002, Detmar and colleagues222 evaluated the efficacy of standardized health-related quality of life assessments in improving patient-provider communication and increasing provider awareness of patient needs. Patients undergoing palliative chemotherapy (n=214) were randomized to the intervention or to usual care. Intervention patients were assessed at three successive outpatient visits. The study reported no statistically significant differences on measures of pain at the final visit for intervention patients as compared to controls. The Jadad score for this study was 2.

Dyspnea

Systematic Reviews of Dyspnea

We identified five systematic reviews addressing the topic of dyspnea in the context of end-of-life care. One of the reviews focused specifically on dyspnea in cancer patients, one on patients with COPD, and three on mixed disease. The review described previously by Gysels and Higginson72 also included studies addressing dyspnea. In this review, the authors summarized six studies regarding dyspnea, including the systematic review by Pan et al.216 Two additional studies were randomized controlled studies, two were qualitative, and one was observational. The sample sizes ranged from 34 to 207 patients and two of the studies focused specifically on dyspnea in a patient population with lung cancer. Interventions included a nurse (RN) clinical intervention, a nurse (NP) CAM intervention, home care with a focus on dyspnea treatment, palliative care services, and one with an unclear intervention. Four of the five interventions described in these research studies (excluding the systematic review) demonstrated beneficial results by reducing the symptoms of dyspnea and/or the anxiety associated with dyspnea.

In a separate review study, Salman and colleagues213 searched three databases as well as the reference lists of selected studies and unpublished studies from meeting abstracts to identify RCTs that included interventions to relieve dyspnea through rehabilitation (either upper-extremity, lower-extremity, and/or respiratory muscle exercises) for patients with COPD. The study authors applied strict criteria for identifying studies with the intended population; however, it is not clear how much of the patient populations included were at the end of life. The authors selected studies in which the clinical status of the patients was reported and in which patients had a diagnosis of COPD and had an forced expiratory volume (FEV1) < 70% of predicted value or an forced expiratory volume / forced vital capacity (FEV1/FVC) < 70% of predicted value. A total of 12 RCTs including in total 723 patients were identified that assessed dyspnea. Intervention studies that included at least lower-extremity training (11 of 12 trials) reported significant improvements in dyspnea. Interventions lasting six months or longer had better outcomes for those with severe COPD while both short and long-term interventions improved dyspnea for patients with mild to moderate COPD. A meta-analysis of the selected studies yielded a beneficial overall effect (OR: 0.62, 95% CI: 0.26, 0.91). Among those with mild to moderate COPD, the total effect based on nine studies was not statistically significant (OR: 0.69, 95% CI: 0.24, 1.14). Among those with severe COPD, the total effect based on three studies was significant (OR: 0.42, 95% CI: 0.02, 0.84).

The study by Pan et al.216 described previously also examined the literature on CAM in treating dyspnea. The interventions included in this review of the literature on dyspnea were acupuncture, acupressure, and behavioral/psychological therapies. A total of six intervention studies were identified; four of these studies explored dyspnea relief for patient populations with COPD and two addressed cancer-related dyspnea. We summarize the findings from this review below by type of intervention, and further details regarding these citations can be found in the systematic review by Pan et al.216

Acupuncture: In a single-blind RCT, 24 COPD patients were randomized to receive 13 sessions of acupuncture over three weeks or sham acupuncture over the same time frame. At the end of the study, the intervention group had less subjective breathlessness and could walk further in a six minute walking test. In a prospective study of 20 patients with cancer-related dyspnea, 70% of patients reported symptomatic improvement lasting up to six hours after acupuncture treatment.

Acupressure: One study was identified that examined acupressure in a single-blind RCT (with crossover) as a treatment for dyspnea. In this study, 31 patients with COPD were randomized to a six-week course of self-administered acupressure alternating with six weeks of sham acupressure. Those in the intervention group experienced a significant reduction in dyspnea symptoms at the end of the study.

Behavioral and Psychological Therapies: Three studies were identified that addressed behavioral or psychological therapies with respect to dyspnea. In one RCT of 20 COPD patients, the authors randomized patients to an intervention of progressive muscle relaxation or usual care and advice to try to relax for 45 minutes a day. The study authors found that, in the intervention group, dyspnea symptoms improved with each session. However, there was no overall improvement over the course of the study. Another double-blind RCT of 65 patients with COPD reported less dyspnea when measured by the Fletcher scale for those who received nurse therapy (which consisted of reassurance without psychotherapeutic training) compared to those who received supportive therapy with psychoanalysis, analytic therapy, and a control group. There was however, no difference in the experience of dyspnea as measured by a visual analogue scale. A third RCT of 20 patients with small cell and non-small cell lung cancer incorporated a one-hour session with a nurse practitioner for three to six weeks in which the patient learned exercises, received counseling, relaxation techniques, and coping/adaptation strategies versus usual care to address dyspnea. After three months, the intervention group reported 35% improvement in dyspnea, 53% improvement in distress, and 17% improvement in functional capacity.

The fourth review study, published by Jennings and colleagues,215 considered the evidence regarding the use of opioids in the management of dyspnea. The authors searched eight difference electronic databases as well as hand searched reference lists of selected articles and textbooks on the subject. Only double-blind, randomized, placebo-controlled trials were included in this review. A total of 18 studies met the criteria for review. All studies had a crossover design. Nine of the studies examined the use of oral or parenteral opioids and nine examined the use of nebulized opioids. Meta-analysis of the studies (the subset of 13 with the necessary level of detail in the data) demonstrated an overall beneficial effect of opioids on the management of dyspnea (standardized mean difference (SMD): -0.31; 95% CI: -0.50, -0.13). Analyses were split by mode of opioid delivery (nebulized or non-nebulized) and the authors found a similar significant effect for the non-nebulized forms of opioid delivery (SMD: -0.40; 95% CI: -0.63, -0.17) but not for the nebulized forms (SMD: -0.11; 95% CI: -0.32, 0.10). An additional meta-analysis was conducted to explore the pooled effect of opioids on exercise tolerance, which failed to demonstrate a statistically significant beneficial effect (SMD: -0.20; 95% CI: -0.42, 0.03).

The most recently published review specifically addressing the research on the treatment of dyspnea was published by Booth and colleagues in 2004.212 The search strategy for this review included a search of three databases, the references of selected papers, and a hand search of key journals in the field. The authors identified 34 randomized controlled trials that examined the use of oxygen in the management of dyspnea for patients with COPD, advanced cancer, and heart failure. Studies were organized and evaluated both around the type of intervention (short- or long-term oxygen therapy) and by patient cohort (COPD at rest, COPD before, during, and after exercise, advanced cancer, and chronic heart failure). All studies, with one exception, included a crossover design. Short-term oxygen therapy for COPD patients at rest led to significant improvement in dyspnea in two out of five studies. Among the studies that included oxygen along with an exercise program, 18 out of 22 had a positive result; in most cases, the intervention led to a slower increase in dyspnea and/or increased endurance rather than simply reduced dyspnea on oxygen. Long-term oxygen therapy had little if any effect on COPD patients with dyspnea. Two of three studies focusing on the management of dyspnea in advanced cancer produced significant improvements in dyspnea with an oxygen intervention. Only one of three studies employing oxygen in the management of dyspnea for patients with chronic heart failure reported a positive finding.

Additional Interventional Studies of Dyspnea

An additional ten randomized controlled studies were identified that explored the role of different interventions on reducing dyspnea in palliative care populations. 230, 231, 222, 232–236 These studies were all published between 1993 and 2003 and were not included in the systematic reviews described above either because they did not meet the disease cohort criteria or the intervention criteria. Eight of the intervention studies focused on cancer patients, one focused on patients with chronic heart failure, and one focused on patients with chronic obstructive pulmonary disease (COPD).

Three of the randomized controlled studies reported on interventions incorporating oxygen to relieve dyspnea in cancer patients.231, 236, 237 Booth and colleagues231 administered oxygen or air to 38 hospice patients with advanced cancer and dyspnea at rest in a single-blind randomized controlled trial with crossover (20 initially receive oxygen, 18 received air). Patients received either oxygen or air for 15 minutes and then were switched to receive the other. This treatment was repeated; however, the authors do not report the number of times the crossover took place. There was significant relief from dyspnea reported for all patients after receiving air (p<0.001) or oxygen (p<0.001) as compared to baseline. However, there was no significant difference in mean dyspnea scores between air and oxygen administration. Analyses of patients stratified by coexisting drug therapy indicate that those on morphine only, benzodiazepine only, or morphine and benzodiazepine had significantly reduced dyspnea with oxygen while those with neither drug therapy had non-significant differences in reported dyspnea with oxygen or air. The Jadad score for this study was 2.

In a double-blind randomized controlled trial with crossover, Bruera and colleagues237 assessed the effects of oxygen on the reported intensity of dyspnea in 14 patients with terminal cancer. Patients were randomized to receive oxygen or air for five minutes at which time patients were switched to the other. This process occurred twice each for oxygen and air. Reports for oxygen saturation, respiratory effort, respiratory rate, and the 100-point visual analogue scale (VAS) for dyspnea were all significantly better with oxygen than with air. The Jadad score for this study was 2.

In another more recent study by Bruera et al.,236 the authors explored the effectiveness of oxygen over air in decreasing dyspnea and fatigue and increasing distance walked during a six-minute walk test in a randomized, double-blind crossover trial. Of the 33 evaluable patients in this study, 31 had lung cancer and all had advanced cancer. Patients were randomized to receive oxygen or air during the first treatment and then switched to air or oxygen for the second treatment. In each treatment phase, the patients performed a six-minute walk test. Contrary to earlier findings by the same author, there were no significant differences between treatment groups in dyspnea, fatigue, or distance walked. The Jadad score for this study was 5.

One randomized controlled230 trial evaluated the effect of specific inspiratory muscle training (SIMT) on dyspnea in 20 patients with moderate heart failure. Ten patients received training in SIMT and the other ten received sham training. Both groups trained 30 minutes a day, six times a week over a three-month period. Inspiratory muscle strength measured by Pimax increased in the intervention group from 46.5 ± 4.7 to 63.6 ± 4.0 cm H2O (p<0.005). Endurance increased significantly in the intervention group (p<0.05) but remained unchanged in the control group. Intervention group members were also able to walk further in a 12-minute walk test than control group members after completion of training (p<0.01). Based on the dyspnea index (0–4 scale), intervention group members significantly improved (p<0.005) while control group members remained unchanged. The Jadad score for this study was 2.

Three studies identified in our review examined the efficacy of morphine in relieving dyspnea.232, 234, 235 All three studies employed a randomized controlled trial design with crossover. Two of the studies employed patient samples with terminal cancer232, 234 and the third employed a sample with COPD.235 Mazzocato and colleagues232 randomized nine patients with lung cancer to receive morphine subcutaneously or a placebo on day 1. The intervention crossed over to the control group on day 2. Morphine doses ranged from 5 mg to 11.25 mg q4h. Mean changes in dyspnea based on a 100-point VAS were -25 ± 10 mm and 0.6 ± 7.7 mm in the intervention and control groups, respectively (p<0.01). Significant improvements were observed in the intervention group relative to the control based on the Borg scale as well (p=0.03). There were no significant changes in somnolence, pain, anxiety, respiratory effort, respiratory rate, and oxygen saturation. The Jadad score for this study was 2.

In the study by Bruera and colleagues,234 ten consecutive patients with terminal cancer were randomized to receive subcutaneous injections of morphine or placebo. Patients were crossed over on the subsequent day. Morphine provided substantial relief from dyspnea at 30-minute (p<0.02), 45-minute (p<0.01), and 60-minute (p<0.01) follow-up assessments. There were no significant differences in O2 saturation or respiratory rate between the intervention and placebo groups. The Jadad score for this study was 0.

Abernathy and colleagues235 evaluated the efficacy of orally administered morphine in 48 patients with predominantly COPD and dyspnea. This study was the only one of the three that was explicitly described as a double-blind trial. Patients were randomized to receive 20mg of morphine sulphate with sustained release or placebo. After four days, patients were crossed over. Thirty-eight patients completed the trial. Based on a 100mm visual analog scale, patients receiving morphine had a mean improvement in dyspnea scores of 6.6 (sd=15) in the morning (p=0.0.11) and 9.5 (sd=19) in the evening (p=0.006). The Jadad score for this study was 5.

In the study by Detmar et al.222 described above in the section on pain, the authors also evaluated the efficacy of standardized health-related quality of life assessments in improving patient-provider communication and increasing provider awareness of patient needs related to dyspnea. No statistically significant differences were found for measures of dyspnea at the final visit for intervention patients as compared to controls. The Jadad score for this study was 2.

The study by Rabow et al. investigated the influence that an outpatient palliative medicine consultation had on symptom relief.86 In this study (described above in the section on pain), the authors reported a significant reduction in patient reports of the degree to which dyspnea interferes with daily activities (p=0.01) but no difference in the frequency that dyspnea limits activities (p=0.07). The Jadad score for this study was 3.

In another more recently published study, Jordhoy and colleagues233 examined how palliative care provided in cooperation between a hospital palliative medicine unit and community-based care improved on patient symptoms relative to usual care. Randomization in this study occurred at the community healthcare district. Cancer patients within these districts received the intervention or usual care (n=235 intervention; n=199 control) and followed for four months. No significant differences in patient ratings of dyspnea were found. The Jadad score for this study was 3.

Depression and Anxiety

Systematic Review of Depression and Anxiety

Our search identified two research evidence reports that covered the topics of depression and anxiety. We chose to address depression and anxiety together because many reports that address one also address the other, although that is not uniformly the case. One report addressed depression as part of a systematic review of the literature on the management of cancer symptoms.34 This study was produced by the New England Medical Center Evidence-Based Practice Center and was described in detail earlier in the section on pain symptoms. The other report is an unpublished review of studies to improve supportive and palliative care for adults with cancer.72 This report, produced by Gysels and Higginson, was also described in detail above in the section on pain symptoms.

The methods applied to develop the evidence report published by the New England Medical Center Evidence-Based Practice Center have been described previously. Only meta-analyses and randomized controlled trials in the topic of depression were included in this reported. Eleven controlled studies were identified that explored the effects of medications on depressive symptoms. Nine were primarily treatment studies on depressive symptoms, and one was a study that explored both pain and depressive symptoms. One study was a depression prevention study. Four studies explored the efficacy of selective serotonin reuptake inhibitors (SSRIs) for depression in cancer patients. Other intervention medications included thioridazine, imipramine, methylprednisolone, mianserin, mazindol, alprazolam, trazadone, and amitriptyline. With the exception of two studies with mazindol and amitriptyline, all medications classified as antidepressants reported benefit for cancer patients. Three meta-analyses were identified that explored the efficacy of psychosocial interventions in treating depressive symptoms in cancer patients. Two of the meta-analyses focused on psychoeducational interventions for general cancer symptoms. One meta-analysis focused specifically on anxiety and depression. The interventions identified in this analysis included individual therapy, relaxation, group therapy, group therapy excluding psychoeducation, and group psychoeducation. A small to medium effect size was reported, but the low quality of the studies ultimately decreased the effect size.

The methods applied to develop the Gysels and Higginson review have been described previously. Twenty-four articles out of a total of 302 studies explored the topic of depression and anxiety. Six of the identified studies addressed depression alone or with other unrelated symptoms (i.e., pain, dyspnea), nine addressed anxiety alone or with other unrelated symptoms, and nine addressed both depression and anxiety. Of the 24 studies, 14 were randomized controlled trials, four were observational studies, one was qualitative, four were systematic reviews and one had an unclear design. Interventions in these studies included behavioral interventions (e.g., group/individual cognitive-behavioral therapy), systems/institutional interventions (e.g., hospital at home, hospice, palliative care teams), education, and complementary and alternative medicine (CAM). The behavioral interventions reported generally beneficial outcomes for anxiety and depression among cancer patients. Systems/institutional interventions produced mixed results; one study on comprehensive hospice care and one on palliative care teams did not report significant improvements in anxiety and/or depression. The other four studies with similar interventions reported beneficial results, however. An educational intervention to address cancer pain significantly reduced anxiety associated with pain. CAM interventions including homeopathy, relaxation, acupuncture, and massage demonstrated reductions in anxiety for cancer patients, but a nurse practitioner-run intervention of CAM did not yield improvements in depression among patients.

Additional Interventional Studies of Anxiety and Depression

Five studies were identified that included interventions to improve depression, and two were identified that addressed anxiety. Schofield et al.238 performed a pilot study using a randomized controlled trial design to investigate the use of the Snoezelen multisensory environment in a palliative day care setting for patients with anxiety. Twenty-six patients were recruited as subjects. The intervention consisted of access to the Snoezelen for one hour on two separate occasions. Control group subjects were given access to a quiet room. Assessments of anxiety were made immediately following access to these two environments. A brief semi-structured interview was conducted with experimental group patients only at the completion of the trial session. A significant reduction in anxiety was seen with the experimental group, but the investigators reported no changes in quality of life. Semi-structured interviews revealed that experimental group patients experienced higher levels of relaxation. The Jadad score for this study was 3.

Soden et al.224 conducted a study to compare the effects of four-week courses of aromatherapy massage and massage alone on psychological symptoms (depression and anxiety) in patients with advanced cancer (this study was previously described above). There were no significant long-term benefits of aromatherapy massage in the improvement of anxiety. There were significant improvements in patients with depression and data suggested that aromatherapy massage may have a beneficial effect on sleep quality for advanced cancer patients.224 The Jadad score for this study was 5.

In a similar randomized study, Wilkinson et al.239 performed a study to assess the effects of massage and aromatherapy massage on cancer patients in a palliative care unit. A total of 103 patients were accrued. Subjects received either massage using an inert carrier oil (control) or a carrier oil plus Roman chamomile essential oil (intervention). Of the 103 subjects, 46 were randomized to the aromatherapy group and 57 to the control group. Unlike the Soden study, this study reported a statistically significant reduction in anxiety across all allocated groups. The aromatherapy group reported a significant decrease in psychological distress with improvement in QOL. The massage group reported improvements as well, but this was not statistically significant. The Jadad score for this study was 3.

In a third study employing CAM, Stephenson and colleagues240 tested the effects of foot reflexology on anxiety and pain in breast and lung cancer patients. In this study, 23 inpatients were allocated to the intervention or control in a quasi-experimental, pre/post crossover design. Anxiety, measured using a 100mm visual analog scale was significantly reduced in patients receiving reflexology relative to the control (p<0.0001). The Jadad score for this study was 1.

Gottlieb et al.241 explored the effect of an exercise program on patients with moderate to severe heart failure on performance and quality of life including depression. Thirty-three patients were randomized to usual care or an exercise program consisting of aerobic training three times a week for six months. Depressive symptoms, measured by the CES-D, did not differ significantly between the intervention and control groups. The Jadad score for this study was 2.

In a randomized controlled trial, Rabow et al.86 explored the efficacy of an interdisciplinary palliative care team on psychological outcomes. This study was described above in the section on pain interventions. There were no significant changes in anxiety or depression levels in the intervention group. The Jadad score for this study was 3.

Addington-Hall et al.242 conducted a randomized controlled trial to explore the efficacy of coordinating care for terminally ill cancer patients on the presence and severity of psychological morbidity. A total of 554 patients were accrued and randomized. Of this total, 318 were randomized to receive coordination care, and 236 were allocated to the control group. The intervention included access to a coordination care team (made up of community-based nurses) who assessed the need for services and offered advice on how to obtain these services. Overall, there were no significant differences between the presence and severity of psychological morbidity across both groups. The Jadad score for this study was 2.

Behavioral Issues in Dementia

Systematic Reviews of Behavioral Issues in Dementia

A total of three systematic reviews were identified that addressed behavioral symptoms in patients with Alzheimer's disease or some other form of dementia.217–219 We summarize the findings from these systematic reviews below. An additional four intervention studies were identified that were either published after the systematic reviews were completed or addressed the same symptoms in another way.

Three systematic reviews were identified that addressed the topic of dementia.217–219 All three reviews addressed dementia in the context of Alzheimer's disease. Two of the systematic reviews focused on interventions for behavioral symptoms in dementia.217, 218 One study by Forbes219 described the use of different strategies to manage behavioral symptoms in Alzheimer's disease. The author searched published and unpublished literature specifically for interventions addressing the following symptoms/activities: aggressive/agitated/disruptive behaviors, social interaction, self-care ability, day/night disturbances, and wandering. Forty-five studies published between 1985 and 1997 were identified. Only one was rated methodologically sound, with the majority (38) being weak or poor. The interventions addressed in the studies to affect the symptoms/activities described included music therapy (most common intervention type), skills training, visual barriers, exercise, bright light therapy, pet therapy, sensory integration, reality orientation, presence, therapeutic touch, life review, and white-noise therapy.

The author reported that exercise in the form of a planned walking program, bright-light therapy, music therapy, written cues, and simulated presence therapy all produced improvements in behavior problems including agitation, aggression, and repetitive vocalizations. Therapeutic touch was the only intervention that did not report beneficial results on behavioral outcomes. Exercise was also successful in increasing communicative function in demented patients as were pet therapy, life-review therapy, and reality-orientation therapy. Music therapy and small-group activities reported non-significant trends toward improvements in communicative function. Music therapy and skills-training interventions were both successful in increasing self-care ability whereas a sensory-integration program did not have a significant effect on this outcome. Bright light therapy and music therapy were used in interventions to normalize sleep patterns and produced beneficial, clinically significant results, although it is unclear that those results were statistically significant. Visual barriers were somewhat effective in reducing episodes of wandering among patients with dementia, particularly Alzheimer's and Parkinson's disease patients.

Opie et al.218 explored the use of psychosocial approaches to behavioral disorders in people with dementia. The authors conducted searches using four databases for materials published between 1989 and 1998. Forty-three papers were included in the review; one had a strong methodological rating, 15 were rated as moderate, and 27 as weak. The following interventions were identified for the review: changes to the physical environment; activity programs; exposure to music, voice, and language; behavior therapy; massage and aromatherapy; light therapy; multidisciplinary teams; and caregiver education.

Changes to the physical environment were made to reduce wandering in dementia patients. The results were mixed; studies that created grids on the floor to disrupt walking patterns were not effective, but covering the doorknobs with fabric or painting them the same color as the door did reduce the number of exits among patients. Placing a mirror in front of the exit door also reduced the number of exits. Sensory stimulation through music, videos, conversation, and exercise was shown to reduce verbal outbursts and repeated requests for attention. Exercise was also successful in reducing wandering, aggressive incidents, and episodes of agitation. Music therapy was the topic of ten interventions in this review; all of these studies reported beneficial results with respect to distress and agitation in demented patients. Behavior therapy was not a common intervention identified in the literature; only two studies were cited and one was a case study. However, in both studies behavioral interventions were successful in training patients with dementia to change negative behaviors (verbal outbursts, entering other patients' rooms and taking personal items). Light therapy was reported as successful in three out of four interventions to reduce agitation and nocturnal disorientation. Massage and aromatherapy have been used with mixed results to reduce agitation among demented patients; two of three studies reported beneficial effects.

Finnema and colleagues217 explored the efficacy of emotion-oriented approaches in the care for individuals suffering from dementia. The definition of emotion-oriented care provided by the authors focused on care aimed at improving emotional and social functioning. The authors searched six databases for studies published between 1990 and 1999 and focused on the following interventions: validation, sensory integration/stimulation, simulated presence therapy, and reminiscence. Six studies focusing on validation were identified; only one was a randomized controlled trial and the remaining five were observational studies of various designs. The RCT was not completed at the time the systematic review was published. The five observational studies reported improvements in behavior and mood using validation. However, the study designs had methodological limitations due to small sample sizes and lack of control groups. Six studies examined the role of sensory stimulation/integration on demented patients' behavior, mood and cognition. Most of these studies reported beneficial outcomes. However, again methodologic flaws in study designs limit our ability to draw conclusions regarding this intervention. Simulated presence therapy is a relatively new form of therapy to reduce aggressive behavior, agitation, wandering, and repetitive vocalizations in patients with dementia. Four studies were reviewed by the authors and reported some beneficial results. Five studies were identified that used reminiscence to reduce negative behavioral symptoms. The work in this area reported mostly beneficial results with regard to decreasing aggressive and attention-seeking behavior as well as disorientation and increasing social interaction.

Additional Interventional Studies of Behavioral Problems in Dementia

We identified four intervention studies published between 1996 and 2003 that addressed behavioral outcomes for dementia patients that were not otherwise summarized in the systematic reviews described above. Out of the five studies that explored dementia, two explored the efficacy of specific pharmacologic therapies on behavioral symptoms related to dementia. Manfredi et al.243 conducted an intervention study to determine the effect of opioids on agitation in demented nursing home residents who were unable to report pain. There was no comparison group for this study, and subjects were not randomized to intervention versus placebo. Subjects were administered placebo for four weeks, and the intervention consisted of a four-week regimen of long-acting opioids. Subjects and nursing home staff were blinded to the medication administered. Twenty-five subjects completed the regimen. Of the 25 evaluable subjects who were less than 85 years of age, no significant differences in agitation level was reported between the placebo and opioid phase. There was a decreased agitation level in 13 of the 25 patients who were greater than 85 years of age at the end of the opioid phase. This decrease in agitation persisted after opioid dose adjustments for sedation. The Jadad score for this study was 0.

In a randomized, double-blind trial, Sultzer et al.244 explored the relationship between behavioral improvement in patients with dementia who were treated on either haloperidol or trazodone. Twenty-eight patients in a geropsychiatry unit with dementia and agitation or aggressive behaviors were recruited. The intervention consisted of either haloperidol 1 mg or trazodone 50 mg. Dose escalation of one capsule was initiated if agitated symptoms worsened. In the haloperidol treatment group, improvement in behavioral symptoms was not associated with baseline delusional scores or with change in delusional scores over the course of the treatment. In the trazodone group, behavior symptom improvement was associated with improving depressive symptoms and neurovegetative signs. The investigators concluded that the use of trazadone in demented patients with mild depressive symptoms was associated with greater behavioral improvement. The Jadad score for this study was 2.

Two of the five studies exploring dementia focused on improving dementia care in nursing homes. Rogers et al.245 examined the effectiveness of a behavioral rehabilitation intervention for improving morning care activities of daily living in nursing home residents with dementia. Eighty-five residents participated in the study. Interventions consisted of activities in two different conditions. Patients in the usual care (condition 1) group received assistance in care by nursing home staff who were consistently assigned to care for them. Patients in the condition 2 group received skill elicitation intervention by a research therapist designed to identify and elicit retained ADL skills. The condition 2 group also received habit training intervention to continue to reinforce and solidify retained skills. ADLs monitored included dressing, bathing, and grooming. The experimental group residents reported an increase in the proportion of time engaged in nonassisted and assisted dressing and increased overall participation of ADL. There was also a concurrent decrease in disruptive behavior for the residents who received the intervention. The Jadad score for this study was 0.

Rovner et al.246 explored the efficacy of a dementia care program to reduce behavior disorders in nursing home residents with dementia. A total of 89 patients were accrued and randomized, and 81 subjects completed the trial. The intervention included an activity program during the day, psychotropic drug management, and educational grand rounds for staff where discussions of individual cases were made. Control treatment included usual nursing home care. Forty-two patients were randomized to the experimental group, and 39 to the control group. After a six-month follow-up, 12 of the 42 intervention patients exhibited behavioral disorders compared with 20 of the 39 control subjects. Control group residents were twice as likely to receive antipsychotic medications and to be restrained. There was more voluntary participation from the intervention group residents. The Jadad score for this study was 3.

Observational Studies and Symptoms

We identified 14 prospective observational cohort studies that addressed one or more of the symptoms and the site of care/condition/race characteristics we considered. Of these 14, four addressed pain management, two addressed delirium in cancer patients, three addressed behavioral problems in dementia patients, and five considered depression in the context of cancer of CHF populations. We highlight the findings of some of these studies here. More detail about the selected studies can be found in the Observational Evidence Table.

None of the studies identified here or in the previously discussed literature reported results separately by race or ethnic groups. Much of the research addressing end-of-life care in different racial/ethnic groups has been done in cross-sectional observational studies and thus was not considered here. More research is clearly needed about how different interventions affect different race/ethnic groups.

In the study by Goodwin et al.,247 the authors compared patients receiving palliative day care to those receiving usual palliative care services (i.e., in the hospital or home). The palliative day care model did not produce better outcomes than usual care. The authors cite limitations of quality of life measures and their current inability to capture all the dimensions of quality of life important to an individual as part of the reason that no differences were observed.

Three studies were selected because they dealt with the treatment of symptoms in a dementia population and where the setting was the nursing home. Nursing home use is associated with an increased incidence of dementia.248 Two of the studies249, 250 examined the use of pharmaceutical interventions (e.g., risperidone for behavioral disturbance in dementia and antibiotics for pneumonia) in improving symptoms and quality of life for dementia patients residing in nursing homes. These studies demonstrate that such interventions can have an important impact on the care of demented residents and that nursing homes can be the site of active intervention and not just custodial care for such residents.

Four studies focused on heart failure and depression.251–254 The exposure in each study was a diagnosis of depression; each demonstrated that depression was significantly associated with poor prognosis, worsening health status, poor functional status, and an increased utilization of health services. These studies indicate the importance of the diagnosis and treatment of depression in heart failure to improve a variety of clinical and quality of life outcomes.

D. Key Questions 2 and 3

2. What patient, family, and healthcare system factors are associated with better or worse outcomes at end of life?

3. What processes and interventions are associated with improved or worsened outcomes?

Elements associated with family experience, especially caregiving

Caregiver burden includes the full spectrum of potential concerns that families and other informal caregivers face in caring for someone with advanced illness. These concerns include but are not limited to mental and physical health, financial well-being including out-of-pocket costs and job loss, and interpersonal stresses. We did not generally assess bereavement, which we defined as after-death emotional concerns, but we otherwise considered the full impact of illness on caregiver well-being. If studies included bereavement as an outcome but also addressed other topics that were central to the review, they were included. However, studies on bereavement alone were excluded.

We evaluated six systematic reviews that potentially dealt with the subject of caregiver burden, addressed the project questions, and met implicit quality criteria. Three dealt with outcomes of caregivers for patients with dementia or other chronic illness; three others dealt with cancer patients or life-threatening illnesses. We went beyond the systematic reviews by including other interventions to reduce caregiver burdens at the end of life published after these systematic reviews or published at any time if not already addressed in a systematic review. In total, we reviewed an additional thirteen intervention studies. Finally, we explored the observational literature that addressed selected topics. Specifically, we identified prospective, observational cohort studies addressing any of our selected symptom topics and that also presented data separately by race, selected disease cohorts, or selected sites of care. Seventeen observational studies that met these criteria are discussed here.

The remainder of this section summarizes the systematic reviews, meta-analyses, intervention, and observational studies relevant to patient and caregiver burden. Summaries of the association of patient, family, and health system factors to caregiver burden and the effectiveness of interventions in improving caregiver burden are found at the conclusion of Chapter 3.

Systematic Reviews and Caregiver Burden

We identified eight systematic reviews and meta-analyses that were relevant to family or informal caregiving and met implicit quality criteria (see Table 5). Three dealt with outcomes of caregivers for patients with dementia or other chronic illness; five others dealt with cancer patients or other life-threatening illnesses.

Table 3. Systematic Reviews for Patient and Family Satisfaction with End-of-Life Care.

Table

Table 3. Systematic Reviews for Patient and Family Satisfaction with End-of-Life Care.

Table 4. Systematic Reviews for Symptoms: Pain, Dyspnea, Depression/Anxiety, Behavior in Dementia.

Table

Table 4. Systematic Reviews for Symptoms: Pain, Dyspnea, Depression/Anxiety, Behavior in Dementia.

Table 5. Systematic Review—Caregivers.

Table

Table 5. Systematic Review—Caregivers.

Acton & Kang (2001)255 reviewed 46 studies (experimental, quasi-experimental, and one group pre-post designs) published between 1982 and 1999. Family caregivers of dementia patients were studied; no further details were given on the patients receiving the care. The caregivers were not caring for “terminally ill” patients or patients at the end of life. Twenty-four studies testing 27 interventions were included in the quantitative analysis (three studies tested two different interventions). Interventions included psycho-education (n=10), education (n=5), respite care (n=4), counseling (n=4), multi-component interventions (n=3), and support group (n=1). A total of 1,254 participants (range: 11 to 180, with a mean of 51 participants per study) were included in the meta-analysis (866 =T; 388=C). The outcome assessed was caregiver burden. The analysis reported that the interventions had no effect on caregiver burden, and in some cases the effect of the intervention was negative or the control group scores improved more than those of the treatment group. Only one multi-component intervention and one respite intervention significantly reduced caregiver burden. Pooled analyses of treatments that evaluated subjective and objective burden separately showed that interventions had no significant beneficial effect on either type of burden.

Acton and Winter,256 in a review that partially overlapped the review conducted by Acton and Kang,255 examined 73 published and unpublished research reports (1991 to 2001) and included controlled trials and pre-post designs. All types of caregivers of patients with dementia and other diseases were included. The studies appeared to span patients with all degrees of severity; no “end of life” studies are specifically mentioned. Less than 50% of care receivers had dementia. A wide range of caregiver interventions designed to lessen the negative impact of caregiving or improve the positive aspects of caregiving for caregivers to patients with dementia or other diseases were included. Outcomes included burden, stress, anxiety, coping, life satisfaction, morale, perceived physical health, and rate of institutionalization. Studies were grouped by intervention type (education, support and education, counseling, respite, case management, and multi-component) and evaluated for strengths and weaknesses in design, sample, intervention, and outcomes. Overall, 32% of the study outcomes were changed in the desired direction after intervention.

Education interventions (23 studies, sample range: 2–95) focused on individualized, home-based programs delivered by nurses one-on-one over time periods ranging from one to eight months (resulting in higher intervention intensity than other education intervention strategies). About one-third of 90 outcome variables measured were reported to be statistically significant in the desired direction (depression, tension, anger, burden, negative affect). Knowledge, coping, and life satisfaction were significantly increased in three studies. Caregiver support and education intervention studies (14 studies, sample range: 26–53) were primarily delivered as individualized education to a caregiver. Only three studies were conducted in a group, community-based format and a fourth was a computerized intervention. Three of the 18 nursing outcomes were significantly changed in the desired direction (burden, stress, and decision confidence) and one study found stress and burden to be significantly increased. Counseling interventions, designed to help caregivers understand the behavioral symptoms exhibited by the demented care receiver and their own reactions to the behavioral symptoms, were tested in four studies (sample range: 5–39). One study reported significant changes in outcomes (increased knowledge and morale) after group counseling. Respite care, including eight studies evaluating day care and eight studies evaluating inpatient or in-home respite (sample range: 7–264) reported that the combination of use and duration varied considerably across studies, making it difficult to determine intervention effect. Intensity of interventions ranged from one to five days per week and duration ranged from one to four months. Case management (assessment, planning, coordination, collaboration, and monitoring by a professional case manager) was evaluated in six studies (sample range: 12–4,151). One study reported a significant reduction in stress. Twelve multi-component interventions (sample range: 5–86) were reviewed. Both positive and negative consequences of caregiving were measured as outcomes; most studies reported mixed results. Two studies found the rate of institutionalization significantly reduced. The results of the meta-analysis provide little support for the interventions studied.

Yin, Zhou, and Bashford (2002)257 examined 26 studies that had comparison groups (single-group pre-post test designs were excluded), published between 1985 and 2000. All types of caregivers were included. The mean age of the caregivers was 60 years; 79% were women and 86% were White. An average of 80% of the caregivers lived in the same household as the care receivers. About half of the care receivers had dementia (the rest had other types of chronic illnesses). The care receiver's mean age was 79 years old. There was no explicit identification of “terminally ill” patients or patients “at the end of life.”

The 26 studies included 18 addressing group caregiver interventions and 8 addressing individual caregiver interventions. The total sample size was 1,970 for the combined interventions and 472 for the individual interventions. Caregiver burden was the primary outcome in the group intervention studies (although only 10/18 used the same instrument—the Zarit Burden Inventory). No details on the individual intervention outcomes were provided. Studies were evaluated on effect size, study design, type of intervention, duration and frequency of intervention, method of assignment, type of instrumentation, time of posttests, and study site, characteristics of the study samples, and characteristics of researchers.

The weighted mean effect size for all studies was .41 (95% CI, 0.32–0.51), indicating a moderate beneficial treatment effect of group interventions on caregiver burden. Subgroup analysis indicated the mean effect size was larger for individual intervention studies (0.48) than for group intervention studies (0.26). The mean effect size of the quasi-experimental studies (0.89) was more than three times that of the true-experimental studies (0.26) but generally consistent with other research.

Higginson, Finlay, Goodwin, Cook, Edwards, Hood, Douglas, and Normand74 conducted an assessment of five systematic reviews (1977-1999, 43 studies) of palliative care team interventions on patients' pain, other symptoms, quality of life, satisfaction, referral to other services, and therapeutic interventions and on caregiver pre-post bereavement burden/morbidity and satisfaction with care. Disease severity, amount of family support, training and experience of team members in palliative care, whether the team had an occupational or physiotherapist, and team links to social services were not described in many of the caregiver-relevant studies and were thus excluded from the analysis. However, the general conclusions of each paper assessed were similar. Overall, the analysis indicated small beneficial effect of palliative care services on patient and caregiver outcomes, with the strongest support for home care services. Similar or improved outcomes were found for patient satisfaction, patient pain and symptom control, and family anxiety for hospice and palliative care services when compared to conventional care. There was a lack of good-quality evidence on which to base conclusions, and there was no evidence of an effect on other quality of life measures.

Gysels and Higginson72 conducted a systematic review of the effectiveness of different interventions targeted at healthcare professionals, the structure of healthcare delivery, or the care delivered to improve supportive and palliative care to cancer patients. In this review, 22 interventions targeted at improving care for families and informal caregivers (including bereavement) were identified. Interventions included home nursing care (four studies), respite services (three studies), social networks and activity enhancements (two studies), problem solving and education (three studies), and groupwork (ten studies). Nine of the interventions were targeted to caregivers only. Limitations of the data included a lack of outcome evaluation designs, small sample sizes and a reliance on intervention descriptions and formative evaluations. Only two quasi-experimental evaluations were included. Interventions, patient and caregiver characteristics, and outcome measures were not well described in the review. The evidence in this analysis appears to contribute more to understanding the feasibility and acceptability of these interventions than to their effectiveness.

Caregivers reported high satisfaction with home care services and described them as useful. However, the high levels of psychological morbidity and unmet need reported in these samples of caregivers using home nursing care in both cancer and palliative care indicated that such generic supportive nursing care does not meet all caregiver needs. Caregivers using inpatient and home hospice care reported a greater reduction in anxiety and higher satisfaction compared to conventional care in one RCT. Another study of home hospice found caregiver quality of life remained stable over four weeks. An RCT of a hospital at home for terminally ill patients in the last two weeks of life reported no significant difference between the intervention and standard care and low uptake of the intervention due to caregiver inability to cope. A formative qualitative evaluation of a community palliative care service (home care, day care, and respite as a single service) reported that caregivers valued the single point of contact, that it felt like a “home away from home,” and that it helped them overcome reluctance to access other services. A longitudinal RCT of home care nursing on caregiver psychosocial status of caregivers with and without physical problems of their own reported an improvement in psychosocial status for those caregivers with physical problems and at risk for psychological morbidity. A psycho-education program for cancer caregivers reported that perception of burden did not worsen even when caregiving tasks increased in intensity. Widely varying respite care services reported high satisfaction in caregivers.

Descriptive data were reported on respite care. One study, using a single group retrospective questionnaire (n=190), reported over 90% satisfaction among caregivers (though 33% felt the service had been offered too late). Social networks and “activation” programs for relatives of cancer patients aimed to promote increased social activity. One controlled trial was reviewed, with an age- and sex-matched comparison group (50 intervention and 45 comparison caregivers) and reported the intervention caregivers had significantly higher social activities during care receiver cancer treatments and significantly more involvement in personal activities preceding the death of the patient. One RCT of a one-to-one intervention designed to provide support, education, and build problem-solving and coping skills (n=38) reported the intervention effective only for a distressed sub-sample of cancer caregivers. This burdened sub-sample of caregivers was better at dealing with pressing problems following the intervention (n=11) compared to controls (n=18). Another one-to-one intervention on cancer pain education (n=50) reported significant improvements on knowledge/attitudes to pain, pain management, and caregiver burden. A multidisciplinary group support designed to alleviate caregiver stress from lack of knowledge is described, but no data are reported.

An RCT of group work interventions for patients and caregivers (n=25), designed to provide support and information to caregivers, reported that spouses who attended the intervention had significantly higher knowledge scores, achieved activity goals, coped better, and were more satisfied with care. However, psychosocial adjustment did not differ between the two groups. An observational study of group work on quality of life reported no differences in quality of life or coping strategies. A descriptive evaluation of a combined patients' (n=73) and caregivers' (n=54) group support for cancer care reported that the provision of information and education promoted understanding and facilitated coping, and the familiarity with the facts and feelings involved reinforced participants' confidence. A retrospective single group evaluation of a monthly support group for cancer patients and families reported that 26% of the respondents felt more anxious/worried and 29% felt sadder. Another observational study of group support on quality of life in cancer patients and their family (n=12) compared to a control (n=12) and an ongoing support group (n=8) found no significant differences in quality of life or coping strategies.

The review supports the small body of evidence on the effectiveness of interventions for caregivers to cancer patients. Despite caregivers' recognition of unmet need, they report some improved outcomes. However, one study indicated that these interventions could be detrimental to caregivers. Based on the analysis, the authors concluded that no single service model appears either acceptable or effective for the broad range of caregivers. Home care appears to produce high satisfaction. The authors concluded there is little evidence supporting the effectiveness of respite services, support groups, and one-to-one individual interventions. Groupwork interventions reported the most effect but they had low uptake and are acceptable primarily to caregivers with specific profiles. The evidence suggests there is a high rate of unmet need in caregivers using palliative care services that highlights the limited scope of some of these interventions.

Wilson's systematic review (in press)259 identified 11 studies and three systematic reviews of care/case management on end-of-life care. Of five research articles examining case management interventions on patient quality of life, two reported an increase in patient quality of life while three other studies reported no significant impact. However, the role of case management in the interventions varied significantly; in many cases, the case management function was not as care coordinator or problem solver. Three studies assessing the impact of case management on caregiver burden reported a beneficial impact on family caregivers. In general, case management was reported to be a means of reducing family caregiver burden, although only three studies were evaluated. In addition, the quality of the studies reviewed was poor, few RCTs were identified, and the exact interventions were inadequately described.

Additional Interventional Studies and Caregiver Burden

We identified 13 additional studies assessing interventions and caregiver burden.

One RCT242 (n=203) reported no effect for nurse coordination for cancer patients in the community. Specifically, there were no significant differences in time between last follow-up and death, in symptom experience, or in use of pharmacologic treatments. A few significant differences arose in caregiver reports of type, severity, and effectiveness of treatment of patients' symptoms in last week of life. The groups did not differ on the hospital anxiety and depression scale, social support, and quality of life, ADL assistance needs, unmet needs, financial impact, use of social services, and satisfaction w/ care.

A multi-site RCT147 compared enhanced home-based primary care (HBPC) in 16 VA hospitals compared to usual HBPC care. Enhanced care consisted of HBPC services plus systematic screening to identify high-risk patients, an emphasis on continuity of care, 24-hour telephone access, and the management of patients across organizational boundaries involving care management by HBPC physician serving as the primary care provider. Terminal patients in the intervention group significantly improved on eight health-related quality of life scales (emotional, social, bodily pain, mental health, vitality, general health), with the greatest improvement in emotional function. No difference was found in terminal patient satisfaction over the study period. Caregivers to terminal patients also reported significant health-related quality of life improvements (p<.05 overall) in all but two dimensions (vitality, general health), with greatest improvement in emotional function (13-point gain vs. usual care). Caregivers also reported significant gains in satisfaction with patient care (p<.001), except for one personal satisfaction item. An 8% reduction in hospitalizations and mean number of hospitalizations in enhanced HBPC program was reported in the first six months, but this was not sustained at 12 months. A 22% reduction in utilization was reported in those with the most disability. Patient and caregiver benefits were accompanied by a 6.8% increase in total costs of care at six months and 12.1% increase at 12 months. The Jadad score for this study was 3.

One RCT261 evaluated an intervention designed to improve end-of-life decision-making and reduce the frequency of a mechanically supported, painful, and prolonged process of dying, conducted in two phases. Phase I was a two-year prospective observational study with 4,301 patients, and phase II was a two-year RCT with 4,804 patients and their physicians (randomized by specialty group) to usual care or an intervention consisting of prognosis estimates to physicians, specially training nurses to improve communication and decision-making among seriously ill hospitalized patients, families, and healthcare teams. Phase I documented shortcomings in communication, frequency of aggressive treatment, and the characteristics of hospital death. Only 47% of physicians knew when their patients preferred to avoid cardiopulmonary resuscitation (CPR); 46% of DNR orders were written within two days of death; 38% of patients who died spent at least ten days in an intensive care unit (ICU); and, for 50% of conscious patients who died in the hospital, family members reported moderate to severe pain at least half the time. In the RCT, patients experienced no improvement in patient-physician communication (e.g., 37% of control and 40% of intervention patients discussed CPR preferences) or in five targeted outcomes, i.e., incidence of timing of written DNR orders (adjusted ratio, 1.02, 95% CI, 0.90–1.15), physicians' knowledge of their patients' preferences not to be resuscitated (AR, 1.22; 95% CI, 0.99–1.49), number of days spent in an ICU, receiving mechanical ventilation, or comatose before death (AR, 0.97; 95% CI, 0.87–1.07), or level of reported pain (AR, 1.15; 95% CI, 1.00 to 1.33). The intervention also did not reduce use of hospital resources (AR, 1.05; 95% CI, 0.87–1.07). The Jadad score for this study was 3.

An evaluation of a hospital-at-home intervention83 (n=198, 86% of 229 referred patients and 144 caregivers, 73% of 198 referred caregivers) on patient's quality of care, likelihood of remaining at home in their final two weeks of life, and general practitioner (GP) visits. The study reported no conclusive evidence that the hospital-at-home service for terminally ill patients increased the likelihood of remaining at home during the final two weeks of life. However, the service was associated with fewer GP out-of-hours visits. All respondent groups (GP, nurses, caregivers) rated the intervention favorably when compared to standard care but emphasized different aspects. Nurses rated services as better than standard care in terms of adequacy of night care and support for the caregiver; GPs positively rated the service in terms of the reduction of anxiety and depression in patients; and caregivers rated the service positively in terms of control of patient symptoms (pain and nausea). Overall, the authors concluded the service provided better quality of care. The Jadad score for this study was 3.

A cluster randomized trial of palliative care services for unspecified terminally ill patients and their caregivers84 (discussed in satisfaction) measured the place of death and satisfaction with care. This same trial reported impacts of the palliative care intervention on caregiver quality of life262 using a larger sample (517 caregivers) and reported similar results. Five of eight subscales of health-related quality of life scores declined between baseline and final follow-up (one to two months after patient's death). As expected, HRQOL returned to baseline by the end of the study. The intervention ameliorated declines in role limitation due to emotional problems and mental health, but scores for the other three subscales showed smaller and almost linear decline. There were extremely low response rates in both groups, which undermines the findings in this study. The Jadad score for this study was 3.

The Resources for Enhancing Alzheimer's Caregiver Health (REACH) project,263–268 a six-year multi-site research program funded by the National Institute on Aging (NIA) and the National Institute of Nursing Research (NINR), focused on testing the most promising home and community-based interventions for maintaining and improving the health and quality of life of caregivers if dementia patients. Fifteen well-defined interventions (nine active and six control group conditions) were implemented (Birmingham, Boston, Memphis, Miami, Palo Alto, and Philadelphia) and assessed common outcome measures. The interventions consisted of psychosocial and psycho-educational services, behavioral interventions, environmental modifications, and technology interventions. Three of the sites included a minimal support telephone contact control group and three sites included usual care control conditions. The study population (n=1,222) included African American, Cuban American, Mexican-American, and White American family caregivers of patients with Alzheimer's disease (AD) and related disorders (ADRD). A common set of measures was collected at all sites at baseline, 6 months, 12 months, and 18 months following random assignment to an intervention condition. Outcome measures included caregiver mental health or well-being and depression; social support; caregiver burden; religiosity; service utilization; caregiver and care recipient physical health and medication usage; and care recipient behavior and cognition.

The pooled effects of 15 site specific REACH interventions (nine active and six control group conditions) on caregiver burden and depressive symptoms following six months intervention and initial analysis of overall treatment effects by categories of caregiver race/ethnic identity, gender, educational level, and relationship to care recipient using an intent-to-treat model were reported.263 Three sites (Birmingham, Boston, Philadelphia) tested a single active intervention (skills-training, telephone-linked computer [TLC], environmental skill building program [ESP]). Three sites implemented two active interventions: Memphis (behavior and enhanced care), Miami (family-based structural multi-system in home intervention [FSMII] and FSMII combined with computer telephone integration system (CTIS), and Palo Alto (coping with caregiving class and enhanced support group). Two sites used modified usual care control groups (Boston and Philadelphia) in which caregivers received information packets only. One site (Memphis) provided information and referral and three other sites (Birmingham, Miami, and Palo Alto) utilized a minimal support control (MSC; information and empathetic listening).

Using meta-analysis, the pooled treatment effect for burden was statistically significant (p=.022), although the difference was small. Overall, caregivers in the active interventions across the REACH sites showed lower values in burden associated with patient behavior problems than controls. No intervention showed a statistically significant effect for caregiver burden, although all scores did improve for active interventions. In contrast to burden, the pooled treatment effect for CES-D was not statistically significant (p=.095). Only one site (Miami) reported a significant reduction in depressive symptoms (p=.034) in the combined family therapy plus technology treatment condition compared to controls. The family-therapy intervention did not have a significant effect on depressive symptoms by itself. Overall, the REACH interventions produced only a modest treatment effect, but this is consistent with results of other recent meta-analyses. The magnitude of the effect sizes for the combined active REACH interventions on caregiver burden (0.15 standard deviation units) and Miami's FSMII +CTIS intervention on depressive symptoms (0.23 standard deviation units) fall within the range of effect sizes reported by others (Sorenson, et al., 2002).269 The magnitude of change on burden for the REACH combined active intervention groups compared to control conditions was 10% (score range=0–96). This change is equivalent to the decrease or elimination of two very bothersome behaviors, such as repetitive vocalization or waking at night. The relatively small overall effects of REACH may be a result of the complex pattern of significant outcomes observed for various subgroups. Across sites, women and those with high school or less education who were in active interventions reported reduced burden compare to controls. In contrast, men and those with higher education levels did not show significant benefit from the interventions. Caregivers in active interventions who were Hispanic, those who were non-spouses, and those who had less education reported lower six-month depression scores than controls. These findings suggest that the combined interventions had an effect for those caregivers in most need of support.

The Miami REACH project264 investigated the efficacy of Structural Ecosystem Therapy (SET), based on the Brief Family Therapy intervention for treating behavior problems in dementia patients and SET+CTIS, a system designed to augment SET by facilitating linkages of the caregivers with their family and with supportive resources outside the home. The sample included 225 family caregivers (114 Cuban-American and 111 White American) of patients with Alzheimer's disease and related dementias (ADRD). Overall, there were significant differences by caregiver type, intervention, and ethnicity on depressive symptoms. Caregivers in the combined family therapy and technology intervention (SET+CTIS) experienced a significant reduction (five or more points for Cuban-American and White non-Hispanic daughters and Cuban-American husbands) in depressive symptoms at 6 months and at 18 months compared to all other intervention groups. Husband caregivers had lower CES-D scores than wife or daughter caregivers in all intervention groups, and Cuban-American caregivers (husbands and daughters) experienced the most benefit from the interventions, particularly the SET+CTIS.

The Memphis REACH265 project compared two structured, parallel interventions, Behavior Care (BC) and Enhanced Care (EC) in a 24-month clinical trial. Behavior Care interventions focused only on improving the caregiver's management of the care recipient's behavioral problems using 25 pamphlets addressing particular behaviors. Enhanced Care interventions focused on these same behavior problems but also on improving the caregiver's own well-being in response to the behavior problems through 12 additional pamphlets geared to caregiver well-being. Both models were delivered by a master's prepared health educator in an office setting. One-hundred sixty seven caregiver-patient dyads were randomized into BC (n=85) and EC (n=82). At two months, 7 of the original 17 active caregiver-patient dyads remained for analysis. Difference in completers vs. noncompleters was based on length of caregiving (shorter length of caregiving more likely to complete). Of final caregiver-patient dyads (n=167), 66 were Black American, 99 were White-Caucasian, and 2 were other race. Caregivers were predominantly women and tended to be spouses or daughters. Results showed that caregivers receiving only BC had significantly worse outcomes for general well-being and a trend toward depression compared to caregivers receiving EC. There was an overall improvement in both groups for bother associated with care recipient behaviors. No racial/ethnic differences were reported.

The Birmingham REACH project266 implemented a multi-component interventions intended to address the common needs of White and African-American family caregivers while remaining responsive to cultural issues. One hundred forty caregiver-patient dyads (White=70 and African American=70) were randomly assigned to either a skills training condition (ST) or a minimal support (MS) control condition. One hundred eighteen dyads completed the six-month assessment (White=70; African American=48). Significant differences between White and African American caregivers included: White caregivers more likely to be spouses and African American caregivers more likely to be non-spouses. White caregivers were significantly older and reported higher household occupational status than African American caregivers. African American care receivers had lower educational attainment and demonstrated greater cognitive impairment than White care receivers. Results demonstrated that both interventions were well received by caregivers. Caregivers in both groups and both races reported decreasing levels of problem behaviors and appraisals of behavioral bother, and increased satisfaction with leisure activities over time. On one measure of appraisal of distress related to behavior problems, White caregivers showed more improvement in the minimal support control condition, and African American caregivers showed the greatest improvements in the skills training condition. No significant effects were found for race, treatment group, their interaction, or time for depression or anxiety. A significant treatment by race by relationship interaction was found with the largest decreases in the number of problem behaviors found for White spouse in the MS condition and for African American spouses in the ST condition.

The Philadelphia REACH267 project examined the six-month effects of an Environmental Skill-Building program (ESB), as well as race, relationship, and gender on caregiver well-being and care recipient functioning. One hundred and ninety family caregivers of community-residing dementia patients completed the six-month follow-up. Caregivers were randomized to a usual care control group (UC) or intervention group (IG) that received five home contacts and one telephone contact by occupational therapists, who provided education, problem-solving training, and adaptive equipment. Baseline and six-month follow-up included self-report measures of caregiver objective and subjective burden, caregiver well-being, and care recipient problem behaviors and physical function. Compared with controls (n=101), intervention caregivers (n=89) reported less upset with memory-related behaviors, less need for assistance from others, and better affect. Intervention spouses reported less upset with disruptive behaviors; men reported spending less time in daily oversight; and women reported less need for help from others, better affect, and enhanced management ability, overall well-being, and mastery relative to controls. Statistically significant treatment differences were not found for hours helping with ADLs and IADLs, perceived change in somatic symptoms, White versus non-white caregivers, or care recipient outcomes.

The Boston REACH project268 examined the 12-month effects of a computer-mediated automated interactive voice response (IVR) intervention designed to assist family caregivers managing persons with disruptive behaviors related to Alzheimer's disease (AD). One hundred caregivers were randomized into treatment (n=49) and control conditions (n=51). The intervention provided caregiver stress monitoring and counseling information, personal voice-mail linkage to AD experts, a voice-mail telephone support group, and a distraction call for care recipients. Measures of the caregiver's appraisal of the bothersome nature of caregiving, anxiety, depression, and mastery were repeated at baseline, 6, 12, and 18 months. Results showed a significant intervention effects for participants with lower mastery at baseline on all three outcomes: bother (p=04), anxiety (p=01), and depression (p=.007). Wives exhibited a significant intervention effect in the reduction of bothersome nature of caregiving (p=.02). Wives and those with low mastery and high anxiety benefited most from the automated telecare intervention.

Observational Studies and Caregiver Burden

A number of prospective cohort and observational studies examined the impacts of caregiving on family caregivers. Two of these were discussed previously (see Satisfaction).27, 99 Seven studies evaluated the overall impacts of caregiving upon terminally ill patients. Covinsky, Goldman, Cook et al.,270 in a prospective cohort study (n=2,129) of outcomes, preferences, and decision-making in seriously ill hospitalized patients found that one-third (34%) of patients required considerable caregiving assistance from a family member. In 20% of cases, a family member had to quit work or make another major life change to provide care for the patient. Even though almost all patients had health insurance, loss of most or all of the family savings was reported by 31% of families, whereas 29% reported the loss of the major source of income. Patient factors independently associated with loss of the family's savings included poor functional status (OR 1.40; 95% CI 1.10–1.78), lower family income (OR 1.74; 95% CI 1.37–2.21 for those with annual incomes below $25,000 and young age (OR, 2.85; 95% CI 2.13–3.82 for those younger than 45 years of age compared to those 65 or older). Families of younger, poorer, and more functionally dependent patients are the most likely to report loss of most or all of the family's savings to a serious or fatal illness.

Emanuel, Fairclough et al.271 conducted a survey of 988 terminally patients and their caregivers in six randomly selected areas of the United States to determine how their needs for assistance were met and the frequency with which they received such assistance from family members and paid or volunteer caregivers. Of the 988 terminally ill patients, 59% were over the age of 65 years and 51.5% were women. The most frequent terminal illness was cancer (52%), followed by heart disease (18%), and chronic obstructive pulmonary disease (11%). Four percent of the sample were in an institution (nursing home, hospital, or residential hospice), the rest were living in the community. Seventy-two percent of caregivers were women and 96% of caregivers were family members. A need for assistance was reported by 87% of the patients, including help with transportation (reported by 62%), homemaking services (55%), nursing care (29%), and personal care (26%). Most patients relied completely on family members and friends for assistance. Only 15.5% of patients relied totally on paid assistance for more than half of the care they needed. Volunteers (unpaid helpers who were not family members) provided less than 3% of all care. In addition to medical care, dying patients often need many types of assistance. Family members, primarily women, provided the majority of assistance with non-medical care.

In a subsequent analysis of this data, Emanuel, Fairclough et al.50 found that 35% of the sample had substantial care needs and that those with substantial care needs were more likely to report that they had a subjective sense of economic burden (44.9% vs. 35.3%; difference 9.6 percentage points [95% CI, 3.1–16.1]; p=0.005). In addition, 10% of these families household income was spent on health care (28% vs. 17%; difference, 11 percentage points [CI 4.8–17.1]; p≤0.001) and they or their families had to take out a loan or mortgage, spend their savings, or obtain an additional job (16.3% vs. 10.2%; difference, 6.1 percentage points [CI 1.4–10.6]; p=0.004). Patients with substantial care needs were more likely to consider euthanasia or physician-assisted suicide (p=0.001). Caregivers of these patients were more likely to have depressive symptoms (p=0.01) and to report that caring for the patients interfered with their lives (p=0.001). Caregivers of patients whose physicians listened to patients' and caregivers' needs had fewer burdens. This study demonstrated that substantial care needs are an important cause of the economic and other burdens imposed by terminal illness.

Brazil, Bedard, Willison, and Hode272 examined the effects of palliative caregiving in the home for 151 family caregivers to terminally ill cancer patients. The majority of respondents were the female spouses (79%) of the patient. The numbers of caregivers providing assistance in specific functional activities were bathing (88%); mobility 81%); dressing and undressing (76%); toileting (67%); and assistance at night (64%). Forty-one percent of caregivers reported that they had been providing some form of care for over one year. Caregivers reported that physical demands in caregiving increased substantially during the last three months of the care recipient's life. As family caregivers provided more assistance in ADLs, they were at greater risk of reporting high caregiver burden.

Hodgson, Higginson, McDonnell and Butters273 prospectively collected patient and family well-being data on all patients referred for care over a six-month period in six home care services in Ireland. Five hundred and eight patients died while in care; 75% of these patients died at home. At referral, 32% of families had severe or overwhelming anxiety. During the last week of care, anxiety remained severe for 26% of care givers. Patient and family well-being were inter-related and there were significant interactions between family anxiety and patient physical and psychological symptoms and communication. Family anxiety at referral strongly predicts family anxiety at last week of life. Excluding family anxiety at referral, other predictors for family anxiety were patient symptom control, sex of patient, diagnosis, and patient age. These data suggest that while severe anxiety is not inevitable for all family members caring for a terminally ill patient, patient characteristics play a role in predicting family anxiety. Family anxiety is associated with patient age, sex, diagnosis and physical symptoms.

Outcomes at Transitions (Placement or Death)

Eight studies reviewed investigated the impact of two critical transitions faced by many caregivers (nursing home placement or the death of the care recipient) on caregivers. Some of these studies overlap with the topic of bereavement, which we did not explicitly address. We include only those studies that were also relevant to understanding other caregiver burdens. One overall impact of these was highlighted previously (under Satisfaction).114 Collins et al.274 prospectively examined changes in depression among family caregivers to dementia patients at three “transition” periods: nursing home placement, bereavement, and continuing residential care over three time periods (pre-event, and two post-event points). A convenience sample of family caregivers (n=142) was included in the analysis focusing on depression (n=46 residential caregivers, 49 institutional caregivers, and 47 bereaved caregivers). The mean depression levels for the total sample declined slightly over the three measurement periods but did not reach statistical significance. Depression appeared to decline among male caregivers and for bereaved caregivers over time but this change was not statistically significant. A more complex relationship occurred between gender and transition groups. Female residential and female institutional caregivers had higher combined levels of depression over time than female bereaved caregivers. In contrast, male residential and institutional caregivers had lower depression than bereaved male caregivers. Finally, male institutional or residential caregivers had significantly lower depression than their female counterparts. Thus, gender has an influence on mental health outcomes for caregivers who continue to provide care as well as those experiencing bereavement.

Grant, Adler et al.275 prospectively examined the extent to which the chronic stress of AD patient caregiving was alleviated by placement or death of the patient in 119 caregivers and 48 non-caregiving comparisons. Three assessments on caregiver mood, blood pressure, and symptoms were conducted at six-month intervals among caregivers who cared for the dementia patient at home for all three observations (n=38), who placed the patient at follow-up (n=28), whose spouse were placed and subsequently died (n=27), those whose spouses died at home (n=26), and 48 non-caregiving spouses. Caregivers who placed the care recipient in a nursing home or whose care recipient died showed significant improvement in depressive and physical symptoms at 6 and 12 months after the transition (placement or death) compared to caregivers who continued to provide care and the non-caregiving comparison group, both of whom had relatively stable depression scores over time. Caregivers who placed the patient at the later time reported fewer serious symptoms over time compared to no change in the other groups. However, both placement and death of patient were associated with higher blood pressure during transitions, and this continued up to 12 months after the transition. These data suggest that both placement and death of the demented relative can have beneficial effects on the mood and serious symptoms of the caregiver but that this effect can take 12 months to become evident. The blood pressure data suggest that a lengthy period of physiological readjustment may be necessary after placement or death of the AD spouse.

Hays, Kasl, & Jacobs276 prospectively examined depression, anxiety, and distress in 1,112 caregivers with seriously ill spouses who survived or died over a two-year period. For analysis, the sample was divided into five groups, depending on whether the subject's spouse had been hospitalized for a critical illness or elective surgery, whether the outcome of the illness was death, and when the death occurred. Depressive symptoms and feelings of helplessness/hopelessness were higher in caregivers whose spouse was hospitalized for critical illness, regardless of the outcome. In addition, distress related to the incidence of bereavement was significantly higher than that of the control group and endured for at least six months after the spouse's death. Anxiety levels did not change in response to either transition (hospitalization and/or death). Depressive symptoms and general anxiety were higher among widows and wives at the time of hospitalization compared to males while gender differences disappeared at two and six months for all bereaved caregivers, regardless of gender. Middle-age subjects reported more hopelessness/helplessness at baseline and six months compared to elderly subjects.

Schulz, Mendelsohn, Haley et al.277 prospectively examined the type and intensity of care provided by 217 family caregivers to persons with dementia during the year before the patient's death and assessed the caregiver' responses to the death. Overall, caregivers exhibited high levels of depressive symptoms while providing care to the relative with dementia (mean CES-D score: 15.8± 11.7; median, 13). Forty-three percent of caregivers had scores above 15. At the death of the relative, depressive symptom scores spiked to 22. However, within three months of the death of the relative, caregivers had clinically significant declines in the level of depressive symptoms, declining to a level similar to pre-bereavement levels (mean, 16.2±12.3; median, 14). Within one year the levels of symptoms were substantially lower than at baseline (mean 11.5±9.4; median, 9) (p=0.03). Caregivers who cared for and then placed their relative in a nursing home had mean scores for depression of 17.1±11.9 (median, 15) before placement and mean depression scores of 18.1±13.0 (median, 15) after placement. One year after placement, depression scores remained high and were significantly higher among caregivers of patients who had been institutionalized than among those caregivers of patients who had died (mean, 16.2 vs. 11.5; median, 14 vs. 9; p=0.02). Use of antidepressant medication and anxiolytic drugs increased after the death of the relative (16.6% and 19.4% before the death, 21% and 18% after the death). While the death of a close relative is generally viewed as a powerful source of psychological stress, the caregivers in this study showed remarkable resilience in adapting to the death of their relatives.

Volicer, Hurley, and Blasi278 conducted a survey of a nationwide sample of 156 family caregivers of demented individuals who had died during the preceding year. Twenty-two percent of patients died at home. The results indicated that end-of-life experiences of individuals with dementia differ according to setting of care. Patients cared for at home and receiving hospice care during the last 90 days had fewer symptoms vs. other groups and fewer signs of physical distress during the dying process. Hospice use did not affect caregiver burden but these patients stayed at home 23 days longer and were twice as likely to die at home than in an institution. Caregivers of patients dying at home had increased time dependence burden but other burden scores were similar among all groups. Caregivers with patients dying both at home and in an institution were less satisfied with care than those cared for in only one setting. No effect on burden was found for use of formal or informal assistance. Psychiatric symptoms in the patient increased caregiver burden and were the most common cause of institutionalization. Receipt of psychiatric care was associated with longer stay at home. Presence of advance directive decreased hospital stays and increased the likelihood of dying in a nursing home. These results indicate that quality end-of-life dementia care can be provided at home by family, with hospice and psychiatric care.

Martikainen and Valdonen279 prospectively examined the effects of the death of a spouse on caregiver mortality using census and death certificate data on all Finns who died between 1986 and 1991. Five thousand five hundred deaths of widowed individuals were examined to determine if income and education mitigated the negative effects of spousal death. The results indicated that both men and women experience excess mortality after the death of a spouse and that the relative excess mortality among the bereaved is broadly similar in all education and income subgroups analyzed. The absolute mortality difference between widowed and married persons, however, tends to be larger among less educated and, especially, low-income persons.

Markowitz, Gutterman, Sadik, and Papadopoulos280 investigated the relationship of caregivers' health-related quality of life to the burden of caring for patients with Alzheimer disease and resource utilization in a sample of 2,477 dementia caregivers. Compared with a normative, age-adjusted sample, the dementia caregivers had lower mental and physical scores (for the latter, only those 54 years of age or older). Increased caregiver mental functioning was associated with caregiver support and perceived quality of patient care, fewer hours of caregiving, and fewer patient behavioral symptoms.

Caregiving for Non-Cancer, Non-Alzheimer's Disease

There has been little research done on palliative caregiving and non-cancer deaths, other than in patients and families with dementia. Two studies reviewed examined the needs of terminally ill non-cancer patients and their caregivers. McCarthy, Addington-Hall, and Ley281 examined the needs, services, and outcomes of care for 600 non-cancer deaths (heart disease) from the Regional Study of Care of the Dying, a population-based investigation of dying people based on reports of their main informal caregivers after the death. Just under half (47%) of caregivers felt they had not been able to get all the information regarding the deceased's illness that they had wanted or when they had wanted it. Thirty-seven percent of caregivers said they had known the deceased was likely to die and 26% said they had “half-known,” whereas 26% of deceased patients were reported to have known and 25% were reported to have “probably” known that they were likely to die. Of those deceased patients who were reported to have known or probably known they were likely to die, most were reported to have had to work this out for themselves: only 8% were said to have been told by a GP or hospital doctor. Moreover, only 44% of caregivers were told of the terminal prognosis. Half of the patients (54%) died in hospitals, 30% at home, and 4% in other places. Patients under age 75 were less likely to die in an institution and more likely to die at home than patients 75 or older. Women aged 75 or older more frequently died in residential or nursing homes than males. One-quarter of the deceased were reported to have expressed a wish to die sooner; more women than men were said to have expressed such a wish (30% vs. 17%, p<0.01). Moreover, decedents who were aged 75 or older were 2.6 times more likely to have expressed a wish to die sooner; those with four or more symptoms perceived as “very distressing” were 2.3 times more likely; and those who had a poor quality of life wee 1.9 times more likely to expressed such a wish. These results indicate that healthcare providers rarely discussed prognosis with heart disease patients, even though the five-year survival rate of chronic heart disease is about 50%, comparable with many types of cancer. Better palliative care, with concern for symptom control and psychological care should be available to all dying patients. Open communication about death and dying is needed to allow for a patient-centered end of life.

Evangelista, Dracup, Doering et al.282 surveyed 103 heart failure patient/caregiver dyads to investigate whether caregiver characteristics were related to the emotional well-being of heart failure patients. Overall, patients had significantly lower (poorer) emotional well-being scores than caregivers (p<.001). However, both gender and age influenced well-being. Female patients and caregivers had lower emotional well-being compared to males; however, the difference was only statistically significant for patients (p=<.018). Male and younger patients had higher (better) scores than female and older patients (p<.05). Patients' age, gender, and caregivers' emotional well-being accounted for 54% of the variance in patients' emotional well-being. These findings suggest that caregiver emotional well-being is associated with the well-being of the heart failure patient. A focus on supporting caregivers and providing them with methods to support their loved ones would be beneficial to patients.

E. Key Question 2 and 3

2. What patient, family, and healthcare system factors are associated with better or worse outcomes at end of life?

3. What processes and interventions are associated with improved or worsened outcomes?

Elements associated with healthcare system performance, especially continuity of services

Introduction

We used a multidisciplinary systematic review of the overall literature on continuity of care as a conceptual framework for our review.283 Based on a systematic review of the literature through 2001 and feedback from an expert workshop, these investigators defined continuity as “the degree to which a series of discrete healthcare events is experienced as coherent and connected and consistent with the patient's medical needs and personal context.” This review, and a second systematic review particularly interested in the concept of measurement,283, 30 identified several key elements of continuity across disciplines: informational, management, and relational. Although we make distinctions between these aspects of continuity in an effort to bring some clarity to the literature in our discussion, in practice, interventions and their measures overlap these boundaries to varying degrees.

We evaluated nine systematic reviews that potentially dealt with the subject of continuity. All nine addressed the project questions and met implicit quality criteria. We went beyond the systematic reviews by including other interventions to improve continuity at the end of life published after these systematic reviews or published at any time if not already addressed in a systematic review. In total, we reviewed an additional 20 intervention studies. Because reviews and interventions related to heart failure were distinctive, we discuss them after more general interventions targeting continuity.

The remainder of this section summarizes the systematic reviews, meta-analyses, intervention, and observational studies relevant to continuity of care. With regard to observational literature, we identified prospective, observational cohort studies addressing continuity and that also presented data separately by race, selected disease cohorts, or selected sites of care. Because there were so few studies that met the design criteria, we also identified several other of the largest studies that addressed important aspects of those questions. Nineteen observational studies met these criteria. All observational studies are discussed at the conclusion of this section. We first summarize information related to continuity in general, and then a series of reviews and studies of patients with congestive heart failure. Summaries of the association of patient, family, and health system factors to continuity and the effectiveness of interventions in improving continuity are found at the conclusion of Chapter 3.

Systematic Reviews

We found seven systematic reviews (Table 6). These systematic reviews are briefly summarized in Tables 6 and 7 and discussed in the text in more detail. Two of the reviews72, 74 and a published paper (by the same authors) graded and summarized other systematic reviews.284

Table 6. Systematic Reviews for Continuity/Coordination.

Table

Table 6. Systematic Reviews for Continuity/Coordination.

Table 7. Summary of Results from Systematic Reviews Relevant to Continuity.

Table

Table 7. Summary of Results from Systematic Reviews Relevant to Continuity.

Gysels et al.,72 the review that explicitly addressed the issue of coordination, identified 11 relevant individual trials. Ten experimented with organizational changes in the usual available care. Many of these studies address palliative and home care interventions and are also addressed in systematic reviews on these topics. Not all studies in Gysels et al.72 or Wilson (Health Canada)259 were relevant to the end of life, so relevant interventions have been extracted and non-cancer or more recent interventions have been added.

With regard to improving continuity, we identified an extensive systematic review of the literature on improving supportive and palliative care for adults with cancer.72 It found that specific interventions, such as structured symptom assessment, needs assessments, improved medical record documentation and sharing, coordination of services, information or education, support, and preparing patients for physician visits, may all improve utilization or various patient-centered outcomes or utilization. Multi-component interventions including these and other elements, such as home care, care protocols, nurse availability, team care, and involvement in discharge planning, often as a part of a palliative care intervention, have also shown some evidence of effectiveness. However, a meta-analysis found that benefits of palliative care interventions, although affecting several domains, appear to be relatively small. Multi-component interventions are often not targeted toward particular outcomes but attempt to address multiple domains, and are also often ill defined. Nevertheless, these studies provided some evidence for reduced hospitalizations from these interventions. The review recommended that further intervention research describe theoretical models, attempt to separate out effects of different components, and evaluate the processes of care in order to better understand how interventions affect outcomes.

Palliative care, home care, and hospice interventions include continuity as an integral component. These interventions vary greatly in the composition of the team and nature of the intervention (e.g., consultation vs. direct care). Some of the systematic reviews assessed communication interventions, support of chemotherapy, and primary care. A meta-regression74 reported that palliative care had small benefits in many areas but did not show benefits for home death. Results were consistent by the type of service. Less evidence is available for hospital at home.

A systematic review of case management at the end of life259 used the Case Management Society of America's definition of case management, “a collaborative process which assesses, plans, implements, coordinates, monitors, and evaluates options and services to meet an individual's health needs through communication and available resources to promote quality, cost-effective outcomes.” Eleven research articles were selected for review out of “over 200” identified, although it is unclear how they defined relevance. The applicability of this descriptive review to this report is unclear because studies of different quality were mixed together, many of the included studies do not seem relevant to the end of life (for example, patients with early Alzheimer's dementia or discharges from a general medical service), and few articles overlapped with our review. In the potential patient-related domains identified with the review, they found conflicting results for well-being and quality of life, patient satisfaction, hospital utilization, home death, and cost-effectiveness. Impact on family caregiver burden was promising but still had limited impact. In general, although they found anecdotal accounts of end-of-life case management, few research articles compared case management to other delivery models, and end-of-life populations or issues were often not included.

A systematic review using only the word “continuity” with end-of-life terms found 13 relevant research articles, 6 monographs, 14 non-research articles, and 2 primary websites.286 The review was very limited because of the single search term, and articles were almost all observational and quite heterogeneous.

Additional Interventional Studies and Continuity

Information/record continuity

The SUPPORT intervention287, 288 and other studies that focused mainly on advance care planning are fully described in that section. Latimer et al.229 randomized 61 patients to a patient care traveling record, and did report a marginally significant improvement in pain control (described in the Pain section above). An RCT giving patients audiotaped recordings of their multidisciplinary oncology consultations reported only higher “usefulness” of the clinic (described in detail under the Satisfaction section above).90

Management continuity

We identified two studies examining the effects of ethics consultations in the ICU.91, 289 The second study was a multi-site trial based on the earlier study in a single institution. Although the interventions were not standardized across institutions, in general, ethics consultations involved coordinating care by interviewing “those involved in the patient's care who bore on the issues under consideration.” Issues were framed in easily understood ethical terms with the involved parties; consultations and recommendations were documented in the medical record; and ethics consultants provided ongoing follow-up. These studies reported that, only in the subgroup of patients who did not survive to discharge from the hospital, patients receiving the intervention had fewer hospital and ICU days and days receiving ventilation than patients in the control group. Further details of these studies are described in the Satisfaction section.

Another study in the ICU reported that patients cared for by an attending physician who focused on continuity of care had lower lengths of stay than other patients in the ICU.290 This single site quality improvement report involved two ICU clinical nurse specialists and an ICU physician who adhered to a structured communication with the family and the nurse specialists who also provided psychosocial support to families. The Jadad score for this study was 0.

A multi-site RCT of team-managed home-based primary care in the VA147 focused on continuity and reported results separately for terminally ill patients. The study reported improvements in multiple domains and is described in detail in the Satisfaction section. Another RCT of nurse coordination in the community, described in the Satisfaction section, reported no effect in multiple domains.242

An RCT of hospital-at-home reported no difference in the location of death (58% controls, 67% intervention) in an intent-to-treat analysis.291 Hospital at home provides practical home nursing support continuously for up to two weeks, typically at the very end of life. However, only 61% of patients offered admission to the home hospital actually enrolled; among those who used hospital at home, the frequency of home death was higher (78% vs. 58%). In addition, this RCT compared the service to standard care, which included a variety of hospice and cancer support services. The Jadad score for this trial was 2. An RCT of a hospital palliative care team, described in the Satisfaction section, did not report an effect on satisfaction.85 A pre-post comparison of a Kaiser Permanente palliative care program, also described in the Satisfaction section, also reported significant improvements in satisfaction.87

DeCourtney et al.292 developed a multidisciplinary quality improvement intervention to deliver end-of-life care services to remote Alaska Native communities, and in a pre-post evaluation of site of death records reported that home deaths had increased from 33% in 1997 to 77% in 2001. This approach (Helping Hands) relies on nurse case management in extremely remote villages and includes coordination with remote physicians, bush air support for nurse visits and medical supply provision, and remote communication using radio and phone. It mobilizes village youth to provide practical hands-on support to patients and families when medical professionals are not available.

Raftery et al.146 performed a randomized controlled trial in the UK of the cost-effectiveness of a district coordinating service for terminally ill cancer patients. The nurse coordinators were based in the community; their role was to assess the need for different services, advise on how to obtain services or contact the agencies themselves when needed, and ensure that services were provided and of good quality. They acted as liaisons with other nurses but did not provide any clinical care themselves. Outcomes were all related to costs. The intervention did reduce health service costs, hospital days, and nurse home visits but did not affect indirect or direct costs borne by patients. The Jadad score for this study was 2.

Selwyn et al.293 evaluated the impact of a multidisciplinary palliative care consultation service for patients with AIDS in an urban teaching hospital in an uncontrolled pre-post design. One of the components of this program includes coordinating care with primary HIV providers and community support and healthcare resources. They report follow-up data on 115 patients followed until death or problem resolution by an interdisciplinary team that worked across settings and attempted to integrate palliative care with usual AIDS care. Full problem resolution was 73% for care decision-making, 59% for conflict resolution, and 7%–67% for symptoms from the MSAS. 55% died while in care, 29% went to a nursing home with hospice, and 14% went to a nursing home for chronic care.

Stockelberg et al.294 evaluated the impact of home nursing for 17 patients with hematological malignancies, and found that providing support and transfusions in the home avoided most ambulatory visits.

We identified two RCTs of continuity-related interventions for the frail elderly. Melin et al. conducted a study of elderly patients who had from 1–5 ADL impairments.295 Intervention and control patients had a variety of diagnoses and a mean age of approximately 80. Intervention consisted of a team of physicians and nurses who made regular home visits for assessment and treatment as well as 24 hour phone support. Approximately 25% of the 249 patients (150-I, 99-C) died during the study. Relative to controls, the intervention group improved in IADL, social activities, and more were living at home at the conclusion of the study (79% vs. 63%). Hospital use was similar, but controls had more long-term care use and intervention patients used more home care days.

Mann et al. conducted296 an RCT of 104 home-based frail elderly patients (52-I, 52-C) who were randomized to functional and home environmental assessment. These patients averaged 73 years of age and approximately 60% in both groups had been hospitalized in the previous six months. Intervention patients received assistive devices such as canes or walkers and environmental interventions such as ramps. Control participants declined on 7/9 functional measures during six months of follow-up compared to the intervention group. Pain scores were also lower at the conclusion of the trial in intervention group. There was no difference in total costs—although intervention participants spent more on devices and modification, and control groups spent more on nursing homes and nurse visits.

Relational continuity

In a randomized, crossover trial, 214 patients receiving palliative chemotherapy completed a HRQOL questionnaire at three successive outpatient visits.222 Physicians and patients were briefly educated about the intervention and were given a graphic summary of the questionnaire before each consultation. HRQOL-related issues were discussed more frequently in the intervention than in the control group, and physicians identified more patients with moderate-to-severe problems in feelings, social activities, and fatigue in the intervention than in the control group. Significantly more patients in the intervention group than in the control group received counseling from their physician on managing their health problems, the level of patient satisfaction with emotional support was higher, and significantly more patients showed improvement in mental health and role functioning. Seventy-nine percent of patients believed that the HRQOL summary increased their physician's awareness of their health problems. The Jadad score for this study was 3.

Specific populations

Heart failure

Many of the studies discussed in this area (in the context of systematic reviews or interventions) excluded ‘terminally ill’ patients and did not address other palliative domains. However, because these studies enrolled patients with advanced CHF who had high mortality and given the prognostic uncertainty in this condition, we determined them to be relevant for the purposes of the review. We identified five systematic reviews related to continuity and coordination in heart failure. One was a previous review that has now been updated297 and one older review only included 7 articles;298 these are not included here. These reviews all addressed the literature somewhat differently, with different definitions and inclusion criteria but some overlap; they are summarized briefly in the table below.

In general, these reviews concentrated on utilization outcomes and mortality. McAlister et al.'s review of 29 RCTs300 reported that trials that incorporated follow-up by a multidisciplinary team, enhancing patient self-care activities, or used telephone contact and advised patients to see their physician when needed all reduced heart failure hospitalizations. Reductions in mortality and all-cause hospitalizations were more variable. All interventions incorporated patient education.

Results for patient-centered outcomes were more equivocal. Windham et al.302 and McAlister et al.300 reported that approximately half of studies that examined HRQOL or functional status showed improvements. Phillips et al.299 pooled QOL scores for 6 RCTs and reported a statistically significant difference: 26% improvement in the intervention groups compared to 14% in the control groups. Patient satisfaction was measured in only three of 32 studies evaluated in one review.301 One review descriptively compared the characteristics of the 15 effective to the 17 ineffective case management interventions and concluded that education and close monitoring for CHF symptoms by nurses or care managers were important components. Eight of the 32 studies included a social worker as part of the intervention.301

Additional Interventional Studies and Continuity in CHF

We identified seven additional interventions related to continuity and coordination in CHF.303–305, 306–309, 310

Stewart et al. reported the outcome of a nurse case management intervention that included structured, intensive education, both patient and family activation, and coordination of care activities between both the primary physician and cardiologist.303, 311 Twenty deaths (10%) occurred in six months of follow-up. Rates of unplanned readmission were lower in the intervention group 68 vs. 118, p=0.031), and costs were correspondingly lower. Quality of life improved among survivors in both groups. The investigators reported the effectiveness of the intervention on the most high risk patients and noted a mortality benefit as well in this subset analysis.311 The Jadad score for this trial was 1.

Goldberg et al. randomized CHF patients to either a telephonic monitoring system linked to an electronic scale and trained CHF nurse case manager vs. usual care.312 Patients with prognosis < 6 months were excluded as were those with advanced renal disease (dialysis or Cr. > 4.0) and other specific cardiac conditions although 13% (37/280) of enrolled patients died in six months of follow-up (lower in the intervention group 8% vs. 13%). There was no difference in re-hospitalization and quality of life improvements were similar in both groups. The Jadad score for this trial was 1.

Jaarsma et al. randomized 179 patients (84-I, 95-C) to a nursing education and one time telephone follow-up of CHF patients that was intended to increase self-care during the ten days after hospital discharge.307 Exclusion included serious comorbidity. In nine months of follow-up, 38/179 (21%) patients died. Self-care behaviors attenuated strongly in control and intervention patients after discharge, although there was a small difference in persistence of self-care behaviors at nine months. This study failed to show a difference in utilization. The Jadad score for this trial was 1.

Philbin et al. conducted a hospital-level quality improvement cluster randomized trial.313 Five intervention hospitals attempted to implement a critical pathway for CHF management. The intervention also included staff professional education. CHF survivors were followed for six months post-hospitalization. All-cause mortality in these patients averaged about 20% among survivors at six months. This inpatient-only intervention noted no post-discharge benefits on mortality, utilization, or quality of life. The Jadad score for this trial was 1.

Goodyer et al.314 performed an RCT of a three-month intensive medication counseling intervention in patients with chronic, stable heart failure. The study did not report mortality. The intervention significantly improved compliance and decreased edema and subjective breathlessness in the intervention group compared to the control group. The control group also had significant decreases in scores for energy and physical mobility that were not seen in the intervention group. The Jadad score for this trial was 1.

Heidenreich et al. reported an uncontrolled pre-post evaluation of an automated weight and vital signs monitoring device linked to physician and nurse manager alerts coupled with nurse education. Forty percent of these patients had moderate or greater reduction in left ventricular function. Survival was estimated as 82% at 12 months. Pre-post utilization and costs were lower in the intervention group, although quality of life was unchanged.

Gorski and Johnson reported a post-evaluation of a quality improvement intervention without a control group306 that suggested benefits in self-care and utilization.

Observational Studies in Continuity

With regard to the association of ethnicity with continuity, although a prospective cohort study found that blacks received less intensive care in the hospital,315 others have found that black nursing home residents tend to receive higher-intensity care and are more likely to die in the hospital.316 Several studies provide potential explanations for these disparities. One study found that black residents tend to be concentrated in nursing homes with fewer available resources, which is associated with more hospitalizations.317 Another study318 examining non-English-speaking patients found that patients of different ethnicities had poorer understanding of their prognosis than English-speaking patients. Another319 found that nurses spent less time at the bedsides of non-white dying patients.

In related literature about settings of care and continuity, we identified several studies that identified problems related to transitions in care in nursing homes.316, 320, 321 Other observational literature (see Satisfaction and Advance Care Planning) highlighted the challenges of transitions and/or continuity involving other or multiple settings.114 This literature also highlights the possible role of advance care planning (ACP) in reducing transfers.316 One study highlighted an association between more intensive staffing and primary care in nursing homes and decreased risk for transfers.322 Several studies suggest a higher preference for death at home than is typically observed.323, 324

With regard to disease, the observational literature underscores the risks of discontinuity in patients with CHF. One study highlighted the social factors, especially single marital status associated with a risk for re-admission.325 Several other studies demonstrate that the risk of re-admission (up to 50%) is particularly associated with age and comorbidity, and in such unselected patients is even higher than in trials using more selected patients conducted in CHF.325–327 We found no studies of other particular disease states or comparative studies of risk by disease.

We identified several studies notable for particular mention related to the subject of continuity. Fisher et al.'s large national study of Medicare cohorts (described in the Satisfaction section above) descriptively highlighted the fact that ten or more physicians were involved in the care of 37% of chronically ill patients during the last six months of life.115 A retrospective study of approximately 9000 decedents using administrative data demonstrated a strong association between higher physician primary care continuity measured using the Modified Continuity Index (MMCI) and lower emergency department use in the last six months of life as well as greater likelihood of home death.328, 329

F. Key Questions 2 and 3

2. What patient, family, and healthcare system factors are associated with better or worse outcomes at end of life?

3. What processes and interventions are associated with improved or worsened outcomes?

Elements associated with decision-making, especially advance care planning

Advanced directives (ADs) (including “living wills” and “instructional directives”), are formal, legally endorsed documents that state instructions for care (e.g., circumstances in which life-sustaining treatment is to be provided or forgone) or that name a proxy/surrogate decision-maker (e.g., “durable power of attorney” and “healthcare proxy”) in the event of future decisional incapacity. The federal Patient Self-Determination Act (PSDA) (OBRA-1990) and legislation on practice in all 50 states provided legal enforcement for ADs that followed certain procedures. More recent efforts to elicit patient preferences have moved toward advanced care planning, which denotes a broader set of activities. Advance care planning requires a well-informed patient or surrogate to make decisions about future care so that treatments undertaken during a future period of decisional incapacity will still be in accord with the patient's preferences.

For this report, studies involving advance-care planning, advance directives, living wills, and “do not resuscitate” (DNR) orders were included if their targets were patients or families (rather than only reporting about clinicians). We evaluated four systematic reviews that potentially dealt with the subject of ACP. We went beyond the systematic reviews by including other interventions to improve ACP at the end of life published after these systematic reviews or published at any time if not already addressed in a systematic review. In total, we reviewed an additional 21 intervention studies. All observational studies that met our criteria are discussed at the conclusion of this section.

The remainder of this section summarizes the evidence from systematic reviews, intervention studies, and observational studies relevant to advance care planning, including ADs, living wills and DNR orders. Summaries of the association of patient, family, and health system factors with ACP and the effectiveness of interventions in improving ACP are found at the conclusion of Chapter 3.

Systematic Reviews

Four systematic reviews were identified that met implicit quality criteria (see Table 9), and reflected three separate reviews addressing the topic of advance care planning.

Table 8. Systematic Reviews Relevant to Continuity/Coordination in Heart Failure.

Table

Table 8. Systematic Reviews Relevant to Continuity/Coordination in Heart Failure.

Table 9. Systematic Review for Advance Care Planning.

Table

Table 9. Systematic Review for Advance Care Planning.

Table 10. Systematic Review of Outcome Variation.

Table

Table 10. Systematic Review of Outcome Variation.

In the first of these, Baggs330 systematically reviewed the literature for evidence on end-of-life care for older adults (over 44 years old) in ICUs and their families and caregivers, focusing in part on literature involving nursing. Ninety-one studies were evaluated, reviewing the literature on interventions, prospective cohort studies, and retrospective analyses. Baggs' review found that the characteristics of appropriate care for dying patients, particularly patients with DNR orders, are not clear. She also identified that advanced age, severity of condition, and DNR orders correlate with limited life-sustaining treatment. Two studies found that resource use decreased following placement of DNR orders. One study found that life-supporting interventions were withdrawn or withheld in 45 percent of ICU patients before their death. On the whole, age was not a factor in limiting care among patients once admitted to the ICU, but age has been a factor in limiting admission to the ICU for the elderly and some studies showed that the elderly appear to receive less aggressive care than do younger patients.

Furthermore, Baggs' review showed that end-of-life hospital care often involves inadequate communication and disagreement between patients and their families and physicians and nurses. Several studies, including the SUPPORT study, found that families and clinicians often have an incomplete understanding of patients' preferences. This inadequate communication can cause some elderly to receive technical interventions rather than their preferred comfort care prior to death. Since dying ICU patients are generally not capable of communicating their preferences, achieving accord on a plan of care is often challenging. Ongoing communication is important because many patients want relief of symptoms without prolonged dying, decision-making that requires both families and providers, and psychological support. Communication is important because, while some of those in ICUs experience a feeling of safety and security, many experience discomfort, cognitive impairment, and anxiety. In the SUPPORT project,287 families reported that in the last three days of life, dying patients were most often in pain, experiencing dyspnea and fatigue. Baggs' review of clinical interventions also showed that hospital end-of-life care does not rely upon evidence-based guidelines.

The Hanson et al.331 systematic review examined the literature between January 1990 and March 1996, on whether interventions aimed at physicians and/or patients affected AD completion and subsequent end-of-life care. Six of the eight randomized studies reviewed, each targeting only patients, increased the rates of AD completion or proxy choices when they combined written materials on ADs with one or more discussions with physicians or a social worker about ADs. Five other studies, that targeted only physicians in single sites, used physician education in combination with reminders and feedback. This combined approach was successful in increasing AD completion and/or advance care planning discussions with patients. Three additional studies reviewed involved both physicians and patients. For example, SUPPORT showed that having an intervention to improve advance care planning did not alter the use of life-sustaining treatment or other outcomes.287 From this review of a total of 14 studies, Hanson et al.331 concluded that, for end-of-life populations, educational interventions for both patients and physicians combined with repeated treatment preference discussions between physicians and patients as well as accessible documentation of the patient's treatment preferences, could reduce the use of life-sustaining interventions at the end-of-life.

A systematic review of the literature and meta-analysis by Higginson et al.332 focused on the impact of hospital-based palliative care teams on patient outcomes and on families. Some of the findings related to advance care planning, particularly reporting a modest effect on hospital length of stay, but the authors note that the study designs do not allow confidence in this assessment or generalization to other settings. The 13 studies incorporated in the Higginson review do not overlap with this report. This Higginson review also reported that hospital-based palliative care teams are somewhat beneficial in terms of reducing the length of hospital stay and having a small positive effect on addressing patient's symptoms.

In the other systematic review, Walsh et al.333 examined the literature on conveying “bad news” concerning a life-threatening diagnosis or death, including publications between 1994 and August 1997. Ten RCTs examined how bad news is delivered and how the impact in doing so influences patients' knowledge of their diagnosis, psychological adjustment, and satisfaction. In eight RCTs, the impact of communication interventions on patient recall and information needs was inconclusive; half showing an increase in short-term knowledge and the others showing no effect. Of the eight RCTs examined for the effect of communication practices on psychological adjustment, seven reported no significant differences, but one RCT found lower adjustment levels and higher anxiety after receiving the “bad news” intervention. The authors noted that these findings were in contrast to an older meta-analysis of 45 studies that had found a positive effect. Patient satisfaction in six RCTs was also inconclusive; three showing an increase and the other three showing no difference.

Additional Intervention Studies of Advance Care Planning

We found an additional 21 intervention studies that were not discussed in the systematic reviews. A trial of ACP in nursing homes was discussed above in the Satisfaction section.92 Studies varied in scope, methodology, duration, and outcome. These studies involved the following approaches to research involving advance care planning: 1) providing education about and the opportunity to complete an advance directive or participate in ACP discussions to patients; 2) having clinicians, patients, and families/surrogates discuss the patient's treatment preferences and prognosis for patients who are dying or are likely to die; 3) having clinicians receive consultations to assist their decision-making process with patients; and 4) determining if efforts to document ADs affect treatment. Of these 21 studies, six were RCTs. Two of the RCTs examined ACP discussions in hospitals, three in outpatient settings, and one in nursing homes; the other RCT assessed the impact of physician-initiated AD discussions in elderly outpatients who were not severely ill. Among these studies, the few reporting differences by race/ethnicity and gender are discussed below.

Within SUPPORT, Marbella et al.288 examined the accord of patients and surrogates as to the patient's preferences, comparing the 386 paired patient and surrogate responses of the intervention group for whom trained nurses spent extra time with patients and families to explain and answer questions about prognoses and potential treatments with the 331 patient and surrogate pairs in the control group. No difference in concordance between patients and families followed the intervention. Among all patient-surrogate pairs, there was slightly less accord if the patient was older or the surrogate was not a close family member. Race and gender did not have a significant impact, though the trend was toward more accord in non-white patients and with female patients. The SUPPORT intervention study has a Jadad score of 3.

The Landry et al.334 and Dexter et al.335 RCTs focused on interventions aimed at increasing the number of adult and elderly patients with advance directives in outpatient settings. In the Landry et al. RCT,334 95 patients with no known life-threatening diagnoses were randomized to the intervention group from the 187 patients in an outpatient clinic. The intervention group participated in an educational seminar without their clinicians and received written materials on ADs, while the control group only received the written materials. AD completion was found to have increased in both groups, but the AD completion rates in the intervention group doubled. In the Dexter et al. RCT,335 1,009 patients age 75 and older, with no known life-threatening illness, were randomized to an instruction directive group, proxy directive group, instruction directive and proxy directive group or to the control group. Primary care physicians were given reminders on the computer-generated encounter form in the patient record to have advance directive discussions with the intervention group patients. Comparatively, there was a significant increase in AD discussions in the intervention groups, half of which resulted in the patient completing an AD. The Jadad score for both studies is 3.

In the Smucker et al.336 RCT, 100 patients 65 years or older, in an outpatient clinic, were randomized to physician-initiated discussions on ADs or to discussions on health promotion and assessed on their subsequent emotional or attitudinal response. Patients in the intervention group did not experience adverse emotional or attitudinal effects, and those who had ACP discussions with their primary care clinicians were more satisfied, findings that were even more pronounced when patient's had higher educational levels and a long term relationship with their physician. This study has a Jadad score of 1.

In a pilot RCT, 61 ambulatory geriatric patients between the ages of 65 to 92 were randomized, either to the control group, which only received a healthcare proxy form, or to the intervention group, which received the form and participated in an ACP discussion with a skilled nurse. For the 31 intervention patients, discussions of ACP involved a program called “Respecting Choices,” while 30 patients had no involvement in “Respecting Choices.” Patients and their surrogates concurred as to the patient's preferences and more patients opted for less aggressive interventions in the intervention group. Interventions such as “Respecting Choices,” where options are presented along with an actual living will form, have been tested in numerous studies, but generally only at one site of care.337 This study has a Jadad score of 2.

We also identified an additional 15 non-randomized intervention studies aiming to improve ACP, six with a comparison group and nine without a comparison group. These studies fell into the following categories: five studies that used specially trained clinicians to discuss ACP, including the diagnosis and prognosis of severely ill ICU patients at high risk of dying with assessment of changes in utilization; three studies that used palliative care teams for ACP discussions for hospitalized patients; one study that used clinician education and institutionalizing forms to assess patients who died; three studies that provided ACP discussions for patients receiving outpatient care; and three studies of end-of-life discussions in patients with advanced illness receiving non-hospice home care.

In one of the active communication studies in ICUs by Lilly et al.,338 advance care planning discussions with dying patients did have some impact, unlike SUPPORT. As part of the intervention, indirect caregivers (e.g., social workers and care coordinators) and some nurses met with the patient following the intensive communication session with direct caregivers, but a standardized meeting template was used to convey information from the intensive communication meetings.338 The improvements documented in the initial intervention were sustained over the ensuing four years. However, this study did not monitor whether the patients admitted to intensive care changed over time. In both of these studies, presenting information about clinical status and expected outcomes to patients having advance directives and their families was associated with increases in decisions to forgo some therapeutic interventions.338

Similar to the findings of the Schneiderman et al. RCT91 described in the Baggs systematic review330 (and also above in the Satisfaction section), two of the communication studies examined the impact of specially trained teams on changes in utilization prior to death. In the first of these two, clinicians of 31 of 99 ICU patients receiving mechanical ventilation discussed the patient's preferences and prognosis with a team of two ethical consultants before decision-making became problematic for the clinicians. These consultations resulted in an increase in decisions to forgo life-sustaining treatment and shorter lengths of stay in the ICU.339 In the second study, Campbell and Guzman340 enrolled MICU patients with global cerebral ischemia after cardiopulmonary resuscitation and multiple organ system failure with or without ADs. These patients and their surrogates were given the opportunity to discuss the patient's prognosis and treatment options with a palliative care team. As a result, patients opted for palliative care more often than for pursuing all possible therapeutic interventions.

The last of this group of ICU communication studies assessed the impact of a healthcare team that met daily to select how 83 ICU patients should be treated, whereupon the team would seek agreement from the family, particularly if it was recommended that treatment be withheld or withdrawn. While the investigators did find an increase the incidence of withholding or withdrawing treatments, the study did not report whether the withholding or withdrawal of treatment was in accord with the patient's AD or treatment preferences.341

Three studies assessed improving communication of patient's preferences in hospitals and assessing the impact on patient outcomes. Jack et al.342 enrolled 50 cancer patients referred to the hospital palliative care team for symptom control, and compared findings to a control group of 50 patients. Subsequent analysis focused on the positive effect on the patients' understanding of their diagnosis and prognosis. Patients in both groups improved their understanding of their diagnosis and prognosis, but the intervention group had higher levels of understanding. The Butler et al.343 study assessed the impact of institutionalizing a standardized DNR order form on the number of patients who died in the hospital with a DNR form. Compared to the period before the form was institutionalized (94 patients), there was an increase in the documentation of the 62 patients' treatment preferences and in patient involvement in treatment decisions while hospitalized. In the Monteleoni and Clark study,344 the impact of a palliative care team communicating with attending physicians about the patient's treatment preferences was assessed. In assessing rates of feeding tube placement before the time when they initiated their intervention, the study found no difference in placement of feeding tubes in patients having ADs, including ADs that documented refusal of artificial nutrition.

We found three studies that investigated the impact of ACP communication with patients with advanced illness receiving outpatient care. In a controlled interventional study, the effect of AD education on 50 COPD rehabilitation patients was compared to 43 patients not involved with the educational workshop. Following the AD education, the rate of intervention patients with a completed living will increased from 52% to 72% and the rate with completed durable powers of attorney increased from 34% to 86%, as well as an increase from 16% to 52% in patients discussing their life-support decisions with a physician.345 Another of these studies involved using palliative care consultations for nephrologists caring for 19 patients with no comparison group. These consultations were intended to assist the nephrologists when they helped patients to select treatments, including symptom control and assistance with coping with their burden of illness. These consultations increased discussions of advanced care planning between the patient's nephrologists and surrogates, though the numbers of advance directives did not increase (also discussed above in the Satisfaction section).88 Both of these studies found that less than half of the patients had ACP discussions prior to the intervention, and their physicians or surrogates did not understand the patients' preferences.

The third ACP communication intervention in outpatient settings involved 204 patients receiving clinical services from two hospitals. The health status and psychological well-being of all study patients were assessed. Of the 104 intervention patients offered an AD form, 67% completed the AD form. Furthermore, signing the AD form did not adversely affect the patients' overall health (both physical and psychological).346

From another vantage point, three studies without comparison groups reported facilitating ACP discussion with severely ill home care patients. An evaluation of the Comprehensive Home-based Options for Informed Consent about End-stage services (CHOICES) program, enabled by the comprehensive nature of managed care in Medicare + Choice, 208 enrolled patients with advanced chronic illness elicited treatment preferences (including modifying ADs for patients with ADs), arranged appropriate services, and eventually facilitated entry into hospice care. The focus of CHOICES was to understand a patient's multifaceted needs and to fashion treatment options to fit the preferences of the patients. During the study period, the length of time in hospice significantly increased among the 208 patients, more patients died at home, and enrolled patients spent less time in hospitals.347 The second home care study investigated the effectiveness of discussions about end-of-life care in a patient's home. After such discussions, almost all the 84 adult patients with life-limiting illnesses were willing to have an advance directive and wanted end-of-life care at home.348 Both studies facilitated increased utilization of hospice care and dying at home. The third home care study was a small investigation documenting the treatment preferences among 31 AIDS patients with no comparison group. The investigator, who assisted the patients with defining their AD, found that standardized AD forms did not fully capture the patient's treatment preferences.349

Prospective Cohort Observational Studies on Advance Care Planning

We found an additional 22 prospective cohort studies that were not discussed in the systematic reviews and that address aspects of advance care planning not covered by the aforementioned intervention studies. These employed various methodologies, including using different instruments and respondents. Eight of these were prospective cohort studies drawn from the SUPPORT intervention study. From these studies, three major themes emerge. First, the impact of efforts to increase AD communication, completion, and documentation was evaluated positively by participants, but has not been shown to be effective in altering treatment patterns. Second, patient preferences often change over time and as illness progresses. Third, communicating with families and involving them, as well as patients, in advance care planning is important when possible.

Two SUPPORT studies by Teno et al.350, 351 found that, in most instances, ADs did not affect end-of-life decision-making,350 especially when family members and physicians had a clear preference and the patient's vague advance directive generally disagreed.351 In SUPPORT, the intervention did succeed in having more advance directives present in the medical record, but very few advance directives provided clear instruction that was applicable to the situation. Very few patients and physicians talked about preferences, even when patients actually prefer forgoing resuscitation.352 However, communication and physician understanding of patient preferences improved when physicians have a longer relationship with the patient, have an accurate understanding of the patient's six-month survival, and discuss resuscitation preferences with the patient.353

Assuming that clinicians have the responsibility of initiating advance care planning discussions, efforts to help them increase AD communication, completion and documentation have been mixed. Assessment of the impact of the AD discussions among 686 patients initiated by primary care physicians resulted in patients being more satisfied with their physicians, and no significant differences were associated with race/ethnicity and gender.354 One Curtis et al.168 study of 31 AIDS patients also found that, when these conversations did take place, patients were more satisfied, but more so when clinicians knew their patients and their ADs. However, non-Hispanic Whites and those with higher incomes were more satisfied. Furthermore, a study of 642 hospitalized cancer patients found that generalists and oncologists discussed similar topics and demonstrated similar prescribed treatments, which were influenced by the perception of the patient's preferences and prognosis.355

Given the importance of patient-physician communication, barriers exist for physicians to initiate end-of-life care discussions. Another Curtis et al.356 study of 57 AIDS patients found that patient-physician communication barriers often stemmed from the clinicians' lack of education about end-of-life care and lack of time for these discussions, as well as having the opinion that the need for end-of-life care discussions had not been evident. A study of 255 patients in a Swiss hospital found that, even among patients with DNR orders, physicians tended to make DNR decisions when they perceived patients as having a poor quality of life; a perception that is often lower than the patient's measured quality of life.357

As time passes, patient's preferences may change. In a study of 50 adults age 65 and older, selection of health impairment states worse than death were similar to the selection of specific life-sustaining treatment preferences; preferences that did change over time.358 Some of these changes may be associated with specific health events, but the evidence is equivocal. In a study of palliative chemotherapy treatment decisions in 203 cancer patients, changes in health related quality of life resulted in patients' opting to modify or discontinue treatment when there was evidence of tumor progression or treatment toxicity; not when the patient's health related quality of life deteriorated.359 Another study of patient preferences for place of death among 98 cancer patients found that, when patients understood their disease and prognosis, almost two-thirds did not want to die in a hospital and their needs could be met at home.323 Similarly, in a study of 80 cancer patients, patient preferences for information and involvement in decision-making often changed between consultations. Generally, female patients wanted more information than males, and patients with worsening conditions more often wanted physicians to make treatment decisions. Some differences in patient preferences for information were associated with which of the two study physicians was seen on a particular visit, so preferences might well relate to physician behavior during patient-physician communication.360

Over the course of illness, patients' preferences may also not reflect their actual prognosis or subsequent utilization. Findings from one of the SUPPORT studies found that patients with advanced colon and lung cancer estimated longer survival times than their actual prognoses, and this correlated with a greater likelihood of wanting life-extending interventions.361 Another SUPPORT study found that decisions to withhold or withdraw life-sustaining treatments was not associated with the patient's race/ethnicity,362 but those with advanced age and men were more likely to have dialysis withheld or withdrawn.363 Another SUPPORT report found that a patient's prognosis and preferences appear to affect the timing of documenting DNR orders.364 A related study, assessing hospital utilization among 241 patients with advanced illness, found that patients' prior preferences were not associated with actual hospital and life-sustaining treatment utilization.365 Similar findings were observed in another study of 65 nursing home residents, where patients' prior treatment preferences were found not to reflect subsequent utilization, particularly when a patient's health deteriorated and family and physician chose to limit further interventions.366

The involvement of families in advance care planning and decision-making is also important, but not a standard of practice. In a study of families and friends of 600 heart disease decedents, almost half had limited information on the decedents' illness, yet half of the decedents were reported as having known their prognosis and likelihood of death—many of whom wanted to die soon because of uncontrolled symptoms.281 In another study of 102 ICU patients in a Paris hospital, physicians did not communicate with half of the ICU patients' families when they were from another country, spoke a different language, were not the spouse of the patient, and did not have a healthcare background.367 There may be communication barriers among clinicians, between clinicians and patients and their families, and between patients and families. In an assessment of end-of-life care communication and cancer patients in England, Ireland, and Italy, there were communication problems in 30% to 40% of instances between patients and their families, compared to communication problems in 10% to 20% of instances associated between clinicians and clinicians, or patients and their families. There were also more communication problems when patients died in inpatient hospice care, not when patients died at home.104

Additional Cross-Sectional and Retrospective Observational Studies and Advance Care Planning

Beyond the findings of the intervention and prospective cohort studies, we reviewed an additional 74 cross-sectional observational studies and 57 retrospective studies. Among the cross-sectional studies, we found that 21 had study populations less than 100, 7 reported outcome differences by race/ethnicity, and 23 reported outcome differences by gender. Among the 57 retrospective studies, we found that 8 had study populations less than 100, 11 reported outcome differences by race/ethnicity, and 5 reported outcome differences by gender. The majority of these retrospective studies used large secondary datasets. Among the cross-sectional and retrospective studies, there were significant differences: in study subjects (e.g., health status, diagnosis, prognosis, and proximity to death), setting of care, survey design and data collection instruments, and study methodology, Educational studies have included providing written AD materials or an informational videotape, providing education on the importance of advance directives to patients and/or providers (e.g., using AD discussion guides and written information for consumers, identification of a central or uniform place for the advance directive, and educational seminars), or counseling activities (e.g., supplying written information on ADs and providing an opportunity to complete an AD or assisting patients with life-threatening diseases to make decisions about medical care). Hammes and Rooney368 reported remarkable effects from a community-wide implementation of AD education, with 85% of all who died in La Crosse, WI, having a written AD at the time of death, virtually all of which were available and followed. Showing that at least one large healthcare delivery system can similarly increase the rate of advance care planning, the Veterans Health Care System designated advance care planning for six conditions as being a mandated goal for each of their geographically based networks. The VA system increased advance care planning for veterans with the designated serious illnesses by 15% system-wide in just three months.369 Two recent reports from Oregon where a special form called the Physician Orders for Life-Sustaining Treatment (POLST) is in common use show that three-quarters of all out-of-hospital decedents369 and of all nursing facilities370 have the POLST completed and available.

G. Summary regarding outcome variations among populations (by patient, family, and health system characteristics)

We identified one systematic review that addressed the issue of outcome variations by race/ethnicity and by settings of care—the two issues that we chose to focus on due to time and resource limitations. Because this review did not map easily to the topics we chose, we discuss it here in introducing a summary of our own findings.

Wilson et al. identified studies relevant to cultural variation in outcomes. The eleven studies were all observational in nature, and were relevant to advance directives, healthcare preferences, communication, and decision-making. These studies generally observed that religion and acculturation as well as ethnicity figured prominently in explaining cultural differences. African-American and Hispanic status was associated with preferences for life-sustaining treatment in several studies. Hispanic status was also associated with impaired communication secondary to language — not only about advance directives, but also pain. Several studies highlight cultural differences not only in preferences for treatment, but also in the process of decision-making with respect to disclosure and the family's role vis-à-vis the patient's autonomy with non-white patients more likely to prefer non-disclosure and group vs. individual decision-making.

The same review also highlighted studies related to home, nursing home, and hospital care. These reviews did not explicitly examine differences, but the individual reviews are useful by comparison. The study identified 11 studies relevant to hospital care at the end of life, 20 studies related to home death, and 22 articles related to long-term care death. With respect to hospitals, the review highlighted the fact that many patients experience a hospital death and that from the population perspective, bed supply is one of the more important determinants of site of death. Women and non-white patients were at higher risk of end-of-life hospitalization, according to several studies. Less than a third of hospitalized patients made advance care plans in several studies and high-intensity care was common, even among patients with dementia. This review highlighted the role of nursing education in palliative care—studies reported an association between education or experience in end-of-life care and management or attitudes. With respect to nursing home issues, the review highlighted descriptive studies of difficulties in pain management, personal care, communication, and caregiver support. Several studies reported an association between hospice use in the nursing home and family perceptions of better nursing home care compared to families whose loved ones did not receive hospice. With regard to home death, the review highlighted literature describing a discrepancy between preference for home death and its low rate of occurrence. Home support was associated with increased likelihood of death and care at home, although home death was also associated with emotional and practical stress on caregivers.

We identified a number of studies highlighting important healthcare system associations. Several of the highest-quality population-based observational studies that have compared the performance of systems of care have found relative deficiencies in symptom management, physician communication, emotional support, and being treated with respect in hospital and nursing home environments compared to hospice at the end of life. Observational studies of symptoms did not inform differences among settings, but did point out associations between treatment and symptoms of behavioral disturbance in dementia. Observational literature on continuity of care highlighted particular issues with continuity related to each setting of care—including hospital readmission, nursing home transfers, and multiple providers when patients are living with advanced illness.

We found little evidence to inform whether or not there are racial/ethnic differences in satisfaction, although we found evidence that racial/ethnic considerations could affect expectations regarding the quality of care, especially with regard to advance care planning and treatment preferences. A number of observational studies describe Hispanic, African-American, and other group preferences for indirect or non-disclosure, group rather than individual decision-making, and use of life-sustaining treatments. The highest-quality observational studies of pain and other symptoms also provided little information on racial/ethnic differences in pain, dyspnea, and depression and anxiety. Observational studies in caregiving and continuity similarly provided little information on racial/ethnic differences, although intervention studies of caregiving did more so (this is discussed in the summary of effectiveness of interventions). A few studies of advance care planning showed modest improvement in patient-surrogate accord with non-white race, and preferences for avoiding planning ahead for persons with poorer health or lower education. However, effect sizes were modest and studies were small and conducted in non-generalizable populations.

The observational literature was generally uninformative with regard to important differences by disease. To the extent that it does highlight differences, it is mostly a function of the fact that research on certain topics is commonly pursued in specific diseases. For example, the caregiving literature highlights problems with caregiving in dementia, although a few studies of advanced CHF also highlight the stress experienced by caregivers. For the most part, with regard to all topics, this shows our need to expand our understanding of how disease status might be associated with the kinds and chronology of needs that patients and families face.

H. Summary regarding the effectiveness of interventions

Satisfaction

Although the evidence is mixed, the preponderance of the interventional and observational literature supports the effectiveness of palliative care for improving both patient and caregiver satisfaction. Subjective measures of the end-of-life care experience include both satisfaction and quality-of-care measures, and these tools overlap significantly. Satisfaction or quality-of-care instruments that assess focused aspects of end-of-life care have been most useful in demonstrating the effects of interventions. Nonspecific satisfaction instruments or studies that use measures not specifically adapted for or developed for palliative care settings have often demonstrated ceiling effects on satisfaction. Possibly for that reason, intervention effects on satisfaction have been somewhat inconsistent.

Measures of satisfaction that are more specific and strongly related to explicit intervention aims or processes (e.g., communication, pain control, practical support, and enhanced caregiving) have demonstrated greater sensitivity to change and support a process-outcome relationship among these variables. The relationship of other processes or attributes of care (e.g., treatment of symptoms other than pain, spiritual support, continuity and coordination of care) to satisfaction is less evident in the literature although it is supported qualitatively. The ability to demonstrate relationships of these aspects of care to satisfaction may be partially related to challenges defining spiritual support as an intervention and measuring spiritual support and continuity of care.

Pain, Depression and Anxiety, and Behavioral Symptoms in Dementia

The evidence base supporting the effectiveness of interventions for cancer pain is quite strong, but better descriptive information is needed about the experience of pain at the end of life in conditions other than cancer. In cancer populations, experiments testing different opioids, different dosages of the same opioid, or different means of opioid delivery did not produce statistically significant results as highlighted in both reviews and intervention studies. These studies were among the strongest in terms of study design. Few CAM interventions had a positive impact on pain relief; acupuncture and massage produced short-term pain relief in cancer patients. Along with descriptive studies, studies of pain treatment in non-cancer conditions needs further study. None of the review studies and only four of the intervention studies included non-cancer patient samples in their studies; none of these studies were on a single disease. Studies of non-pharmacologic interventions are small and of varied quality.

Morphine and other opioids may have a beneficial impact on dyspnea; one meta-analysis and three small but promising intervention studies reported mostly positive results for cancer and COPD. No large studies have examined interventions to relieve dyspnea in cancer or non-cancer conditions, or attempted to describe the experience of dyspnea, despite the fact that dyspnea is a characteristic symptom of several important end-of-life conditions (e.g., advanced cancer, COPD, CHF). Dyspnea in advanced CHF appears to be the most understudied among these conditions. The evidence from the reviews and individual intervention studies presents relatively negative results for the role oxygen therapy plays in the management of dyspnea in cancer patients. Exercise interventions may have a positive effect on those with severe COPD and heart failure but these have not been tested in cancer patients. In small, short-term studies, acupuncture, acupressure, and relaxation therapy showed some clinical benefits.

Effective interventions have targeted the pharmacologic treatment of depression in cancer, but relatively few studies have evaluated shorter-acting drugs, or the treatment of depression in non-cancer conditions. We reported on one extensive review of the intervention literature regarding depression in cancer patients. Of the seven interventions considered, five focused on the treatment of depression and/or anxiety in cancer patients as well. The other review and two intervention studies focused on other disease cohorts (one study focused specifically on depression in heart failure patients, the other on mixed disease). SSRIs have been shown to be very successful in treating depression in palliative care populations. Behavioral and CAM interventions have demonstrated mixed results.

The existing literature on dementia has focused primarily on Alzheimer's disease. Given the considerable amount of time one can live after a diagnosis of dementia, these studies are somewhat limited in the context of this review because it is not clear how many of them include a population clearly near the end of life. The literature addresses many symptoms for the dementia patient population: aggressive/disruptive behavior, agitation, wandering, and mood were the most common. These studies suggest that a variety of non-pharmacologic therapies may be effective for behavioral symptoms in dementia. Pharmaceutical interventions were the subject of only a few studies and with mixed results. There are many more methodological limitations in the literature on dementia making it difficult to make definitive statements about the best treatment for these patients.

Caregiving Burden

In general, a variety of interventions were studied for a broad range of caregivers (e.g., spouse, adult children, others), primarily caregivers to dementia patients255–257 and to terminal cancer patient caregivers,72, 74, 258, 259, 286 usually as a supplement to clinical palliative care services being provided to the terminally ill patient. Most studies, whether on dementia or end-of-life caregiver interventions, focused on caregiver burden (objective and subjective burden) as the main outcome measure, but outcomes also included psychological distress (stress, depression), anxiety, coping skills, life satisfaction, health related quality of life, satisfaction with services or care, morale, rate of home death, rates of institutionalization, and costs.

There were generally two kinds of interventions used to address caregiver burden: individual and group interventions. The interventions included education, counseling, support groups, home health, hospice, or palliative care services to caregivers singly, or in some combination. For the most part, intervention studies have reported inconsistent results. Larger treatment effects have been found for individual interventions,257 yet group interventions predominate the literature (Knight, Lutzky, & Macofsky-Urban, 1993). In addition, only multi-component interventions and some respite services have shown positive (though small) impacts on caregiver burden. The inconsistencies in the literature may be attributable to the differences in the caregiver outcome measurement, research design, and analytical methods used.

The caregiving interventional literature provides some information about ethnic or racial differences in caregiving experience. Caregiver race was significantly associated with the effect size of some interventions. Several studies in our review demonstrated such differences with regard to African-American and White caregivers. Race was significantly associated with caregiver intervention impact (p<.001), indicating that the treatment was more effective for non-white caregivers.257 Hispanic men and Hispanic and White daughters experienced a higher impact from Birmingham REACH intervention than other groups.264

Overall, palliative care teams do appear to have a small but beneficial effect on patient outcomes. In contrast, small effects have been found related to caregiver outcomes. There is also no significant effect of palliative care teams on home death rates, no matter what the make-up of the team.72, 74

Continuity

The preponderance of systematic reviews and interventions supports the efficacy of interventions to improve continuity in the context of palliation of cancer. In addition, we found some lower-quality evidence that palliative HIV care could improve continuity. Interventions embody a variety of successful approaches including aspects of management, informational, and interpersonal continuity as well as comprehensive integrated care such as palliative care services. We found evidence for the effectiveness of interventions targeting care at multiple levels—provider, patient, provider/patient interface, and multiple settings but particularly home and hospital. Our review is limited in that it identified no evidence related to improving continuity across multiple sites of care.

Although we identified many effective interventions for improving continuity in CHF care, few of these explicitly addressed or reported patient-centered palliative outcomes (e.g., dyspnea, advance care planning, caregiving impact). However, successful interventions share features of successful interventions in general including longer intervention periods, coordination among providers, and regular, structured home assessment. Many CHF interventions specifically excluded patients who were ‘terminally ill,’ limiting their generalizability. We identified no palliative interventions targeting other conditions and continuity of care—other than in the context of unselected populations that were more commonly focused on cancer care. Most interventions have targeted re-admission to the hospital or other kinds of high-cost care, but interventions are needed to understand how to improve other aspects of continuity as well.

Advance Care Planning

The usual practice of advance directives and advance care planning is supported by little reliable scientific evidence of efficacy in improving outcomes. Improved communication and planning has some tendency toward improved patient and family satisfaction, and certainly anecdotes and small series point to patient and family frustration and disappointment with seriously flawed communication. Nevertheless, high-quality research designs have not often been applied to these questions and, when applied, have shown quite modest effects, even on increasing the rate of making decisions in advance. Whether improved advance care planning actually improves the experience for patients and their families has only thin and equivocal evidence.

However, studies provided several key insights involving advance care planning. First, advance care planning has to reflect changing preferences and circumstances; patients' preferences change over the course of their illness. Second, when clinicians and families understand and agree with patients' preferences and prognosis, patients are more likely to experience preferred outcomes. Third, physical and psychosocial support for patients and their families is needed and can improve communication and decision-making among clinicians, patients, and families. Fourth, interventions limited to one type of strategy and one site of care, as well as those that have few study subjects, are not likely to change care patterns or have long term impact. For about half of the studies reviewed here, only one site of care was used and a small number of patients were enroll.

PubReader format: click here to try

Views

  • PubReader
  • Print View
  • Cite this Page

In this Page

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...