Perceived Success in Addressing End-of-Life Care Needs of Low-Income Elders and Their Families: What Has Family Conflict Got to Do With It?

Transcription

1 Vol. 41 No. 1 January 2011 Journal of Pain and Symptom Management 35 Original Article Perceived Success in Addressing End-of-Life Care Needs of Low-Income Elders and Their Families: What Has Family Conflict Got to Do With It? Betty J. Kramer, PhD, and James A. Yonker, MS School of Social Work (B.J.K.) and Department of Sociology (J.A.Y.), School of Medicine and Public Health, University of Wisconsin-Madison, Madison, Wisconsin, USA Abstract The purpose of this study was to examine the predictors of perceived success in addressing the end-of-life care needs of low-income older adults and their family members. Perceived success is defined as the clinician s subjective assessment of the extent to which end-of-life care needs of the patient and family have been met by the interdisciplinary team. The results are drawn from a larger longitudinal multimethod case study designed to understand how end-of-life care is provided to a diverse group of frail elders in an innovative, fully integrated, managed care program. Data were generated from 120 social work surveys detailing care experiences and outcomes particular to 120 elder deaths. Significant predictors of perceived success for addressing patient needs included patient care needs (b ¼ 0.17, P # 0.05), race (b ¼ 0.19, P # 0.05), patient preferences elicited (b ¼ 0.29, P # 0.01) and honored (b ¼ 0.20, P # 0.05), and family conflict (b ¼ 0.24, P # 0.01). Significant predictors of perceived success for addressing family needs included family care needs (b ¼ 0.30, P # 0.001), team and administrative resources (b ¼ 0.19, P # 0.01), patient preferences honored (b ¼ 0.16, P # 0.05), quality of relationship with patient (b ¼ 0.27, P # 0.001) and family (b ¼ 0.23, P # 0.01), and family conflict (b ¼ 0.31, P # 0.001). This study provides preliminary evidence of differential correlates and predictors of perceived success for addressing patient and family needs, highlighting the detrimental influence of family conflict. Future research is needed to better understand the kinds of assessment and intervention protocols that might prevent or ameliorate conflict and enhance structures and process-of-care variables to facilitate more successful outcomes. J Pain Symptom Manage 2011;41:35e48. Ó 2011 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved. Key Words End of life, palliative care, care quality, structures of care, process of care, perceived success, family conflict, low-income elders, family, managed care Address correspondence to: Betty J. Kramer, PhD, School of Social Work, University of Wisconsin- Madison, 1350 University Avenue, Madison, WI 53706, USA. ejkramer@wisc.edu Accepted for publication: April 20, Ó 2011 U.S. Cancer Pain Relief Committee Published by Elsevier Inc. All rights reserved. Introduction Although end-of-life care research has grown dramatically over the past decade, the paucity of scientific knowledge continues to /$ - see front matter doi: /j.jpainsymman

2 36 Kramer and Yonker Vol. 41 No. 1 January 2011 constrain efforts to improve care delivery systems, 1e4 particularly for elders. 5,6 Although hospice programs have made great strides to provide comprehensive interdisciplinary care to persons at the end of life, 5 regulatory and other barriers limit service utilization of elders with chronic, progressive, and ambiguous comorbid conditions. 7 To optimally meet the complex end-of-life care needs of these elders and their families, innovative health care systems are needed that integrate care across settings, tailor services to meet emerging needs, provide comprehensive interdisciplinary community care, and proactively anticipate death as an eventual outcome. 6,8 One innovative elder care model program is the Wisconsin Partnership Program (WPP) (developed from its parent, the Program for Allinclusive Care for the ElderlydPACE). WPP provides comprehensive, community-based, interdisciplinary, and consumer-directed care for nursing home-eligible low-income elders with complex chronic illness. It also offers integrated financing of long-term care (i.e., primary and acute) under Medicare and Medicaid capitation arrangement, provider consistency, 9 and comparable consumer satisfaction with PACE. 10 Although WPP was not originally designed to care for the dying, the vast majority of elders are enrolled until death. As the movement to develop integrated, community-based, managed care programs for poor and frail elders grows and other states seek to replicate and expand on innovative models, 11 it becomes important to understand how structures and processes of care in these settings influence success in addressing the end-of-life care needs of patients and families. Our preliminary research with the Care Wisconsin Partnership Program (CWPP; originally named Elder Care Partnership), the first WPP site, identified several challenges to providing end-of-life care. 12 One of these was competing needs of complex family systems, 12 and more specifically, family conflict that was present in 55% of the deaths. 13 Little attention has been given to understanding how family conflict might affect the ability of team members to address the end-of-life care needs of the elder and the family. The purpose of this study was to examine the extent to which family conflict will predict perceived success in addressing end-oflife care needs of elders and families after controlling for structure- and process-of-care domains. Perceived success may be defined as the clinician s subjective assessment of the extent to which end-of-life care needs have been met by the interdisciplinary team. Quality improvement initiatives often use perceived success as an indicator for evaluating program outcomes. 14 Evaluation of care outcomes is by definition subjectively determined, and health care providers have unique and valid perspectives on this evaluation. 15 Interdisciplinary team members have the responsibility to address the physical, psychological, social, spiritual, and cultural needs of the patient and family, which are considered essential domains of a good death, 15,16 and to assess and monitor the extent to which these needs are met. Team members directly observe what transpires at the end of life as a result of their care. Given the many challenges in measuring quality of dying, 17 clinician perceptions of care quality offer valid and feasible alternatives to assess care outcomes. 15,18 Conceptual Framework Fig. 1 details the conceptual framework guiding this study, which was based on an adaptation of the comprehensive conceptual model put forth by Stewart et al. 19 detailing domains for evaluating the quality of end-of-life care and how it affects patients and families. Personal and Social Environment. Contextual factors in the personal and social environment may have an effect on structure, process, and outcomes of care. 19 Research on contextual factors that may influence outcomes is limited although there is evidence that race, socioeconomic status, diagnosis, and care needs are very important to understanding variability in outcomes for dying patients. 1 Structure and Process of Care. There is a wide range of structure-of-care variables that might influence outcomes. 19,20 Because elders in the CWPP program may die in any number of care settings while still receiving interdisciplinary team care, we are particularly interested in how place of death and team and administrative resources will influence perceived success. We anticipate that perceived success in addressing end-of-life care needs will be greater among

3 Vol. 41 No. 1 January 2011 Perceived Success in Addressing End-of-Life Care Needs 37 PERSONAL AND SOCIAL ENVIRONMENT STRUCTURE AND PROCESS OF CARE Place of death Elder age Elder race Care needs of elder & family Team and administrative support Patient preference elicited and honored FAMILY CONFLICT OUTCOMES OF CARE Success in addressing end-of-life care needs Diagnosis Quality of relationship with patient and family Fig. 1. Conceptual model of factors affecting success in addressing care needs at the end of life. those who have greater team and administrative resources/support and who care for patients who die in a hospice setting. According to the conceptual model put forth by Stewart et al., 19 two major process-of-care domains of interest are those related to decision-making processes with the patient and family, in the form of care preferences being elicited and honored, and interpersonal and relationship processes, measured by the quality of the relationship with the patient and family. Family Conflict. Family conflict has been defined as interpersonal tension or struggle between two or more persons (family members) whose opinions, values, needs, or expectations are opposing or incompatible. 13 Nurses in a hospital-based palliative care program identified family conflict as hindering the effective provision of family care, 21 and hospice teams described added challenges they face in providing care when responding to challenging family dynamics. 22 Similarly, physicians viewed conflict as a major barrier to providing quality end-of-life care. 23 Interdisciplinary team members in the CWPP program reported that family conflict restricts care planning and implementation, jeopardizes quality of endof-life care, and increases elder, family, and/ or team member distress. 13 It is hypothesized that after controlling for contextual and structure- and process-of-care domains, conflict will predict lower levels of perceived success in addressing the end-of-life care needs of elders and their families. Outcomes of Care. A major goal for improving end-of-life care is to successfully address the needs of both patients and family members, and family-centered care is considered essential to quality palliative and end-of-life care. 16,24 As such, two central outcome domains focus on the quality of life of the patient and family. Quality-of-life indicators for patients typically include the extent to which their needs are fulfilled in terms of their physical or psychological symptoms and the extent to which their spiritual and cultural beliefs and needs have been addressed. 25 For families, these indicators include support for psychological or emotional response to illness, grief or bereavement, and caregiver support. 25 Continuous effort to more adequately address these needs is an essential goal of quality improvement initiatives in palliative and end-of-life care. 26,27 This study focused on the following primary research question: To what extent do structure- and process-of-care variables and family conflict contribute to perceived success in addressing end-of-life care needs of low-income elders and their family members served in an innovative, fully integrated, managed care program? This study tested the following research hypotheses, which were generated from the literature review and conceptual framework described above. Hypothesis 1. Structure-of-care (i.e., place of death and team and administrative resources) and process-of-care variables (i.e., patient preferences elicited and honored, and quality of

4 38 Kramer and Yonker Vol. 41 No. 1 January 2011 relationship with patient or family members, respectively) will be associated with perceived success in addressing the end-of-life care needs of patients and families. Hypothesis 2. After controlling for personal and social contextual variables and structureand process-of-care variables that are associated with perceived success, family conflict will be significantly associated with lower levels of perceived success. Methods Design This project was part of a larger longitudinal case study. The case study is a mixed-methods approach to research and evaluation that has a distinguished history in the social sciences. 28 Case study stresses the uniqueness of a case (or cases) 29 and is the preferred design when interest is on a program or a contemporary phenomenon within its real-life context. 30 Although our broader aim was to understand the phenomenon of end-of-life care within the CWPP program, the results reported here focus on predictors of perceived success in addressing elder and family member end-of-life care needs derived from a quantitative survey completed by social work team members as described below. Setting and Sample Elder Care of Dane County, a not-for-profit organization, has provided community-based services for elders since CWPP, the largest program operated at this site, was chosen over the three other WPP sites in the state of Wisconsin because it is the largest in terms of the number of participants served and has been operating for the longest period of time. CWPP employed 10 interdisciplinary teams consisting of a nurse practitioner (NP), a Master s level social worker, and two registered nurses (RNs) at the time of this study. Teams contract with the elder s physician, who, jointly with the NP, manages the participant s medical care. The interdisciplinary team works with the elder to create a comprehensive, personalized, biopsychosocial service plan and is responsible for coordinating medical, psychological, and long-term care service needs. Although there is flexibility in team member roles, nurses generally monitor health, identify home care needs, and provide health and prevention education and social workers address emotional, social, and mental health needs, provide resources and supportive counseling to the elder and their family members, and coordinate mental health, financial, housing, and community services. Teams meet weekly to review care plans, and at least one member of the team accompanies the elder to medical appointments. Frequency of contact with elders and team members varies tremendously depending on the elder s functioning and needs and generally increases dramatically near the end of life. In contrast to hospice program patients short length of survival, 31,32 the average duration of enrollment to death in CWPP is three years. As such, team members tend to develop very close relationships with the elders served and come to know involved family members quite well. Since its 1995 inception, CWPP has enrolled approximately 470 elders, with 289 current enrollees. Participants have annual incomes below $10,000, average 29 medical diagnoses, seven to eight major chronic medical conditions, and three or more activities of daily living dependencies. Seventy-two percent of disenrollments are because of death, averaging three deaths per month. Data Collection To ensure consistency in protocols and honor agency interdisciplinary team and advisory committee recommendations regarding optimal and most feasible methods for survey completion, social work staff were identified as the designated clinicians to take responsibility for ensuring completion of surveys, eliciting data regarding the care experiences and outcomes particular to 120 deaths. All social workers participated in a three-hour training session to review protocols for survey completion, addressing issues of reliability and validity. Social workers were given clear instructions for involving other team members to optimize the quality of data. They were instructed to consult with either an RN or an NP (or both) to confer on health-related information and the social worker s ratings of care needs, family conflict, and perceived success. If there were discrepancies in ratings, they were asked to obtain consensus through discussion of any discrepancies

5 Vol. 41 No. 1 January 2011 Perceived Success in Addressing End-of-Life Care Needs 39 with team members and to record which team members assisted in completing the survey. Fifty-nine percent consulted with either an RN or the NP, and 41% consulted with both team members. A total of 14 social workers completed surveys for the 10 teams, accounting for turnover resulting from position transfers, retirements, promotions, or maternity leaves. The total number of surveys each social worker completed varied based on the number of deaths that occurred on each team, which was influenced to some extent by the age of the team (i.e., the number of teams had grown as the census increased over the years) and the duration of employment. Social workers on newer teams who were employed for shorter periods of time completed anywhere from one to nine surveys, whereas others who had been employed for several years and were caring for elders for longer periods of time completed anywhere from 12 to 21 surveys. Surveys were both retrospective (i.e., based on recall and case record review of most recent elder deaths; n ¼ 75; 62%) and concurrent (i.e., completed as deaths occurred; n ¼ 45; 38%). For retrospective reports, team members reviewed actual case records to minimize recall bias. Outcomes. Perceived success of meeting patient needs was measured with a 17-item scale consisting of ratings regarding the extent to which the social worker believed that the team was successful in achieving the goals relevant to each condition that was addressed. A 5-point continuum was used for each item to indicate the degree of success (1 ¼ not at all to 5 ¼ extremely successful). An average score was computed. The 17 items included pain, dyspnea, incontinence, constipation, nausea, food/fluids, skin integrity, medication management, personal care, safety, depression, agitation, anxiety, psychological and emotional response, spiritual needs, funeral plans, and cultural beliefs/rituals. Cronbach s a was Perceived success of meeting family needs was measured with a three-item scale consisting of an identical rating system (1 ¼ not at all to 5 ¼ extremely), and an average score was computed. The three items included psychological and emotional response of the family, caregiver/support system involvement and support, and grief and/or bereavement. Cronbach s a was Family Conflict. To assess the influence of family conflict that was present at the end of life, respondents were asked to respond to the question How important was family conflict to the care of this patient during the last six months of life? using a 5-point scale (1 ¼ none/did not apply to 5 ¼ extremely important). Responses of 1 indicated that there was no family conflict, and responses in the range of 2e5 indicated the degree of conflict. This single-item indicator did not ask clinician s to specify the type or nature of conflict; however, as a follow-up, other sources of data from this study elaborated on team members perceptions of various forms of conflict, contributing factors, and consequences, which are reported elsewhere. 13 Structure and Process of Care. Respondents indicated place of death, and these were coded into dummy variables (0 or 1) for each place of death: home, community-based residential facility (CBRF), hospital, nursing home, and inpatient hospice. Team and administrative resources were measured with a 5-item scale in which respondents were asked How helpful were each of the following toward assisting you in providing quality end-of-life care to this participant? Items included: team support, team communication, team consensus about care plan, effectiveness of team meetings, and administrative support. Respondents used 5- point scales to indicate levels of helpfulness (1 ¼ not at all to 5 ¼ extremely helpful), and an average score was computed. Cronbach s a was For patient preferences elicited, respondents used a 5-point response set (1 ¼ not at all to 5 ¼ completely) to answer three questions that were scaled and averaged. Questions centered on the extent to which advance care planning discussions were initiated with the patient and the family, respectively, regarding the participants wishes and the extent to which the wishes were known to the team. Cronbach s a for this three-item scale was Preferences honored were assessed with the question Overall, to what extent were the health care wishes of this participant honored and carried out? (1 ¼ not at all to 5 ¼ completely). Quality of relationship with the patient and family was measured with single-item indicators asking the respondent to use a 5-point scale (1 ¼ extremely poor to

6 40 Kramer and Yonker Vol. 41 No. 1 January ¼ excellent) to rate the quality of the relationship that you had with this participant and primary family contact, respectively. Personal and Social Environment. Respondents indicated the age of the patient at the time of death and their race (non-hispanic white, African American, Native American/Eskimo, Southeast Asian, Hispanic, and other). Race responses were condensed and dummy coded (0 ¼ non-hispanic white; 1 ¼ minority) for the analysis. For end-of-life care needs of the patient and family, respondents rated the 17 items for patients and three items for family members previously described according to the degree of need that was addressed by the team during the last six months of life using a 5-point rating system (1 ¼ not at all to 5 ¼ extreme). Cronbach s a s were 0.74 and 0.69 for patient and family care needs, respectively. Finally, respondents indicated the primary diagnosis that ultimately contributed to the cause of death, coded into the overarching categories of cardiovascular, renal, cancer, pulmonary, gastrointestinal, neurological, multisystem failure, and unknown. These were further coded into dummy variables (0 or 1) for each category. Data Analysis The analysis has four parts. First, we present descriptive statistics to describe the patient population and the results of bivariate analyses conducted to determine which diagnoses and place of death variables would be included in the analyses. Second, we present the descriptives and bivariate correlations for the analytical variables. Third, to test the first hypothesis, we examine correlations to determine the association between the structure- and process-of-care variables and perceived success. Finally, to test the second hypothesis, we use hierarchical multiple regression analysis to control for the personal and social environment and structure- and process-of-care variables. Given the limited sample size and the number of independent variables, only the contextual and structure- and process-of-care variables that were significantly correlated with perceived success were chosen for the analysis to allow for a more parsimonious model. Omitted variables were analyzed as covariates in an exploratory regression analysis, and aside from race, their presence did not alter appreciably the relationships between independent variables and perceived success. Therefore, race was included in the model. The first block included the personal and social environment variables of age, race, care needs, and cancer diagnosis. The second block included two places of death variables, team and administrative resources, patient preferences elicited and honored, and quality of relationship with the patient and family. The final block included family conflict. Results Background Characteristics of Elders Served and Perceived Success Table 1 describes demographic characteristics of the deceased elders representing 62% of the total and most recent deaths in the program. Elders ranged in age from 57 to 103 years, with a mean age of 80 years (standard deviation [SD] ¼ 9.6 years). The majority was female (68%), nearly half were widowed (47%), and fewer than 35% had completed high school. No relationship was found between perceived success and marital status, gender, or education of the patient. Twenty percent of elders were not Caucasian, disproportionate to the percentage of racial minority elders living in Dane County (w2%). In terms of location of death, most elders died in the nursing home (30%), followed by hospital (28%), home (18%), and group home (16%). Only 8% received hospice care. Contrary to our expectation, hospice care was not correlated with greater perceived success (r ¼ 0.12, P ¼ 0.21). As will be described below, two places of death variables that were correlated with perceived success were hospital death (r ¼ 0.27, P # 0.01) and group home death (r ¼ 0.19, P # 0.01). The top four causes of death were categorized as cardiovascular disease (37%), pulmonary disease or infection (20%), multisystem failure (11%), and neurological disease/dementia (10%), reflecting common causes of late-life death attributed to advanced chronic diseases. 33 Interestingly, cancer was the only cause of death that was correlated with perceived success at the trend level (r ¼ 0.17, P # 0.10), suggesting that team members perceived that they were less successful addressing the end-of-life care needs

7 Vol. 41 No. 1 January 2011 Perceived Success in Addressing End-of-Life Care Needs 41 Table 1 Demographic and Background Characteristics of Deceased Elders Patient Characteristic Mean (SD) Total (n ¼ 120) Percent Age at death 80 (9.6) Education No formal education Elementary school Junior high school Some high school High school grad Some college College graduate Graduate school Unknown Gender Male Female Race Caucasian African American Native American/Eskimo Southeast Asian Hispanic Other Marital status Single/never married Married Divorced or separated Widowed Place of death Home Group home Nursing home Hospital Inpatient hospice Primary cause of death Cardiovascular disease Pulmonary disease/ infection Multisystem failure Neurological disease/ dementia Cancer Renal disease Gastrointestinal disease Unknown of cancer patients compared with patients who died with other late-life chronic conditions. Structure- and Process-of-Care and Perceived Success: Hypothesis 1 We hypothesized structure- (i.e., place of death and team and administrative resources) and process-of-care variables (i.e., patient preferences elicited and honored, and quality of relationship with patient or family members, respectively) will be associated with perceived success in addressing the end-oflife care needs of patients and families. Table 2 presents the means and SDs of the analytic variables as well as the zero-order correlations among them. As shown in Table 2, in addition to the contextual variables of age, care needs of elder and family, and cancer diagnosis, significant correlations were found between perceived success and all of these structure- and process-of-care variables. Perceived success was higher when patients died in a group home (r ¼ 0.19, P # 0.05), there were greater team and administrative resources available to the staff (r ¼ 0.28, P # 0.01), patient preferences were both elicited (r ¼ 0.52, P # 0.001) and honored (r ¼ 0.46, P #.001), and quality of relationships were better between the respondent and the patient (r ¼ 0.27, P # 0.001) and family (r ¼ 0.23, P # 0.05). Perceived success was lower when patients died in a hospital setting (r ¼ 0.27, P # 0.01). Hypothesis 1 is supported. Family Conflict and Perceived Success: Hypothesis 2 We hypothesized that after controlling for personal and social contextual variables and structure- and process-of-care variables that are associated with perceived success, family conflict will be significantly associated with lower levels of perceived success. Table 3 presents the results of the regression analysis for success in addressing patient needs, explaining 46% of the total variance in perceived success. In the first step of the model, personal and social environmental variables explained 14% of the variance in perceived success, and all three of the patient characteristics were significant predictors. In the second step, structure- and process-of-care variables explained an additional 28% of the variance in perceived success in addressing patient needs, the two most important being patient preferences elicited (b ¼ 0.31, P # 0.01) and honored (b ¼ 0.22, P # 0.05). In the third step, family conflict significantly explained an additional 4% of the variance in perceived success (b ¼ 0.24, P # 0.01). Significant main effects in the prediction of perceived success for addressing patient needs in the final model included patient care needs (b ¼ 0.17, P # 0.05), race (b ¼ 0.19, P # 0.05), patient

8 42 Kramer and Yonker Vol. 41 No. 1 January 2011 Table 2 Correlations and Descriptive Statistics for Analytic Variables Variables Perceived success patient needs Perceived success family needs 0.53 c Patient age at death 0.20 a Patient race a Patient end-of-life care needs 0.18 a Family end-of-life care needs 0.23 a 0.51 c c Cancer diagnosis Hospital death 0.27 b b Group home death 0.19 a b b Team/administrative resources 0.28 b 0.38 c b 0.40 c 0.35 c Patient preferences elicited 0.52 c 0.45 c c a c Patient preferences honored 0.46 c 0.41 c b b 0.24 b 0.23 a 0.47 c Relationship quality with patient 0.27 b 0.36 c c 0.28 b Relationship quality with family 0.23 a 0.51 c c a a Family conflict 0.29 b 0.30 c 0.31 c 0.20 a c 1.00 Mean SD Range a P # b P # c P # Table 3 Hierarchical Regression Analysis for Perceived Success of Addressing Patient Needs b Values Variables Step 1 Step 2 Step 3 Personal and social environment Patient age at death 0.23 a Patient race 0.19 a a Patient end-of-life care needs 0.21 a a Cancer diagnosis Structure and process of care Death in hospital Death in CBRF Team and administrative resources Patient preferences elicited 0.31 b 0.29 b Patient preferences honored 0.22 a 0.20 a Quality of relationship with patient Quality of relationship with family Family conflict 0.24 b Total R R 2 change F for R 2 change 4.35 b 7.06 c 8.13 c a P # b P # c P # (two tailed). preferences elicited (b ¼ 0.29, P # 0.01) and honored (b ¼ 0.20, P # 0.05), and family conflict (b ¼ 0.24, P # 0.01). Hypothesis 2 is supported. Table 4 presents the results of the regression analysis for success in addressing family needs, which explained 61% of the total variance in perceived success. In the first step of the model, personal and social environmental variables explained 26% of the variance in perceived success, but only family care needs was a significant predictor (b ¼ 0.49, P # 0.001). In the second step, structure- and process-ofcare variables explained an additional 28% of the variance in perceived success in addressing patient needs and all of the structure- and process-of-care variables except for place of death and patient preferences elicited were significant predictors. As hypothesized, in the third step, family conflict significantly explained an additional 7% of the variance in perceived success (b ¼ 0.31, P # 0.001). Significant main effects in the prediction of perceived success for addressing family needs in the final model included family care needs (b ¼ 0.30, P # 0.001), team and administrative

9 Vol. 41 No. 1 January 2011 Perceived Success in Addressing End-of-Life Care Needs 43 Table 4 Hierarchical Regression Analysis for Perceived Success of Addressing Family Needs b Values Variables Step 1 Step 2 Step 3 Personal and social environment Patient age at death Patient race Patient end-of-life care needs 0.49 c 0.25 b 0.30 c Cancer diagnosis Structure and process of care Death in hospital Death in CBRF Team and administrative 0.16 a 0.19 b resources Patient preferences elicited Patient preferences honored 0.19 a 0.16 a Quality of relationship with 0.19 a 0.27 c patient Quality of relationship with 0.33 c 0.23 b family Family conflict 0.31 c Total R R 2 change F for R 2 change 9.36 c 8.92 c c a P # b P # c P # (two tailed). resources (b ¼ 0.19, P # 0.01), patient preferences honored (b ¼ 0.16, P # 0.05), quality of relationship with patient (b ¼ 0.27, P # 0.001) and family (b ¼ 0.23, P # 0.01), and family conflict (b ¼ 0.31, P # 0.001). Hypothesis 3 is supported. Discussion To our knowledge, this is the first study to investigate the role that family conflict may have in contributing to perceived success in addressing the end-of-life care needs of low-income frail elders and their families. As hypothesized, this study documents that family conflict has a detrimental influence on the extent to which social workers perceive that the interdisciplinary team was successful in addressing needs for both the patient and family. Even after controlling for contextual and structure- and processof-care variables that were correlated with the outcome, family conflict significantly explained additional variance in perceived success. This supports our prior qualitative research findings that family conflict contributes to poorer outcomes. 13 Professionals often feel ill prepared to address conflict because there is little educational training provided in family systems concepts 34 or family conflict prevention and intervention. 35 Given the strong relationship between end-of-life conflict and a history of family conflict, 36 there may be little staff can do to completely prevent conflict from arising. However, it seems prudent to consider routine assessment and screening protocols to identify families at risk 34,35,37 and to identify strategies that teams may find helpful to employ as early on as possible. For example, it would be fruitful to consistently implement more comprehensive intake assessments, map out all family members and potential decision makers, and better understand existing client-family relationships, communication patterns, family roles, and any historical information that may have bearing on how different family members may respond to the patient s needs (e.g., prior deaths in the family and how they were handled). The routine facilitation of family meetings to build trust, exchange information, and promote reciprocal understanding of care needs and preferences may enhance communication among the family, patient, and team members, facilitate more open lines of communication, and potentially minimize future potential sources of conflict based on misinformation or miscommunication. Investing time and effort in developing the relationship with the family beginning at intake may ease difficult decisions and transitions during the later active stages of dying. Additional principles recently put forth for interdisciplinary palliative care teams for managing conflict include maintaining the palliative perspective, flexibility, neutrality, transparency and professionalism, avoiding splitting and demonizing, setting necessary limits, and employing suitable interventions. 22 This study supports our assertion regarding the need for future research to develop and test interventions to provide greater emotional support to the family. and to facilitate shared decision making and enhance open communication among at risk families. 36 Few familybased interventions have been developed or empirically tested. One promising intervention is the Family Focused Grief Therapy Model 37 that has been used successfully to address grief in conflicted families during palliative care and bereavement. Whether such models might be adaptable or feasible in consumer-directed managed care programs for poor elders and

10 44 Kramer and Yonker Vol. 41 No. 1 January 2011 their families is an empirical question. We have recently identified two predictors of family conflict that may be targeted for intervention development, including communication constraints and asserting control. 36 Brief interventions such as family meetings or individual counseling for family members and training families in stress management, mindfulness techniques, and compassion practices might be other interventions worthy of consideration and empirical testing. Aside from family conflict, the other most compelling findings from this study concern the importance of structure- and process-ofcare variables that differentially influence perceived success in addressing patient and family member needs at the end of life. In the final regression models, after controlling for personal and social environment, honoring patient preferences was the only shared predictor of perceived success in meeting patient and family needs. A key component of high-quality end-of-life care is honoring patient s treatment preferences. 38 As a consumer-directed program, the primary goal of CWPP is to identify and carry out the patient s goals for care. As such, the extent to which this is accomplished should naturally influence perceived success regarding meeting patient needs. It should be noted, however, that honoring preferences and perceived success are not interchangeable concepts. In fact, team members may not be able to honor the patient s wishes but still feel that they were successful in addressing needs. For example, many frail elders with advanced chronic disease wish to die at home but must live in a group home or nursing home because of their frail health, functional status, resource constraints, lack of in-home caregivers, and/or the team members ethical concerns regarding safety. Even in these lessthan-ideal situations, team members may perceive that they were successful in addressing needs related to symptom management or safety. For the family, we know that, for the most part, family members of frail elders want to honor the elder s wishes. 39 One might expect that if the patient s wishes are being honored, it might ease the complications related to psychological, emotional, and bereavement needs of the family so that teams would more successfully address these needs. The only distinctive predictor of perceived success in the structure- and process-of-care domains for the patient was the extent to which the patient s preferences were in fact elicited, suggesting that teams are more successful in addressing needs of patients when they have actually engaged in conversations to elicit wishes in advance. Although this finding is to be expected, it is important to acknowledge that eliciting and honoring patient preferences are not necessarily easy to accomplish. Many elders with advanced chronic diseases may lack decisionmaking capacity or the ability to articulate their wishes. 40 Barriers to honoring care preferences was the third most frequently reported challenge in providing quality end-of-life care in the CWPP program, resulting from elders diminished capacity to comprehend care options or consequences and their complex care needs, the incapacity of support systems, and the realities of balancing safety and risk assessment and costs. 12 Recommendations for addressing barriers to honoring care preferences that have been put forth include a variety of strategies and approaches related to advance care planning, communication, and anticipation of ethical dilemmas. 13 These include making early efforts to establish rapport and engage family members in advanced care planning and endof-life care discussions to ensure that family members are aware of the patient s preferences regarding a variety of scenarios; enhancing communication with collateral service providers to minimize problems during transitions in care (e.g., with hospital staff); using creative approaches such as reminiscence or life review to explore values and prior experiences that could be used as a springboard toward discussions regarding future scenarios and how they would like them to be handled; anticipating ethical dilemmas and developing and integrating risk assessment protocols; making earlier referrals to ethics committees; and educating staff regarding the legal pathways for protecting patient s when family disregard their wishes. 13 In addition, as increasing evidence accumulates regarding the potential value of the Physician Orders for Life-Sustaining Treatment (POLST) as an approach for ensuring that patient s preferences are elicited and honored, 41 and because this model has been used successfully in the PACE program, 42 community-based managed care programs serving frail low-income

11 Vol. 41 No. 1 January 2011 Perceived Success in Addressing End-of-Life Care Needs 45 elders may consider working more collaboratively with primary care physicians to integrate POLST as part of the care plan. In contrast to the predictors of success for addressing patient needs focusing on eliciting and honoring care preferences, perceived success in addressing the needs of the family member was distinctively influenced by team and administrative resources and relationship quality between the social worker and both the patient and family. Social workers believed that the interdisciplinary teams were better able to address family needs when their relationships were of higher quality and there were higher levels of team and administrative support. Although limited research has been done on relationship quality and team and administrative resources, a recently published qualitative descriptive study reported similar findings regarding the importance of relationship quality between staff and residents, shared goals and expectations among staff, and administrative support in supporting quality end-of-life care and outcomes among hospice-enrolled residents in assisted living facilities. 43 A few observations regarding the distinctive nature of consumer-directed models that may have bearing on the relationships between social workers and family members in particular are worth noting. As opposed to hospice models of care, consumerdirected models are challenged by the reality that they are not designed, mandated, or financed to meet the needs of family members. Rather their mandate is consumer directed. So the extent to which elders want their family members included, are cognitively able to make their own decisions, do or do not have family members who are actively involved in their care or living in the area, as well as several other factors, will influence how and in what ways team members will have an opportunity to develop relationships with family and to address and respond to family member needs. We know from other results from this case study that competing needs of complex family systems were especially challenging to teams. Examples of these challenges include precarious psychological well-being of family members including depression, schizophrenia, drug or alcohol use, anxiety., and as noted by one social worker, the difficulty of engaging family members with a history of drug addiction, incarceration, homicide, violence,.. 12 Certainly, we would expect that better relationship quality with the patient and family would lead to enhanced success in addressing needs and that teams working well together and who are well supported by program administrators will have better resources to meet their needs. We found substantial support for the conceptual model proposed by Stewart et al. 19 as a heuristic device for understanding success in addressing end-of-life care needs of patient and family, evidenced by significant increases in R 2 values with the addition of each model component and statistically significant relationships between perceived success and variables across all conceptual categories. Although structure- and process-of-care variables investigated represent a variety of domains considered essential for quality care at the end of life, 16,24 little research has investigated the ways in which these variables may differentially influence outcomes in different care settings and across populations. Yet, understanding which structure- and process-of-care variables are most important to care outcomes may inform program development and quality improvement initiatives. There are several limitations of this study design and measurement that should be taken into consideration in interpreting these results. First, as a case study, results are not presumed to be generalizable but are generated for their potential value for what they might teach us about the particular case. 28 Certainly, these results may inform program development and quality improvement initiatives in this setting. We suggest that there may be value in understanding what the CWPP program is experiencing as it may promote understanding of similar cases of general issues encountered in other programs 29 that are attempting to replicate this program or attempting service integration and employing interdisciplinary teams to serve a similar low-income, frail, elder population. Second, our reliance on social work retrospective accounts may be subject to bias because of issues of recall, reinterpretation of the past based on intervening events, and current experience of grief and bereavement. 25 We found no substantial difference between mean perceived success scores for retrospective vs. concurrent reports (t ¼ 1.56, P ¼ 0.12). We recognize that for some of the variables (e.g., patient and family member s needs, quality of relationship), it would have been more ideal to interview

12 46 Kramer and Yonker Vol. 41 No. 1 January 2011 patients and family members directly to ascertain their perspectives; however, the complexity of such a design was not feasible in this setting. Some evidence suggests that professional ratings may be more congruent with patient reports than with family member reports for particular ratings such as physical symptoms. 25 To assess perceived success, we certainly needed to rely on the professional s perspective; however, it is unknown whether these reports are congruent with objective indictors of patient and family outcomes. Ethical and methodological considerations curtail the amount of information researchers may obtain from patients and their families during the dying process, making studies of family conflict at the end of life in real time especially challenging. Third, we acknowledge that our measures are limited. We worked in collaboration with agency personnel to develop an instrument that would minimize respondent burden for already burdened staff. As such, our goal was to keep our measurements as concise as possible, and as a result we used several single-item indicators for ease of completion. Our single-item family conflict measure assessed the clinician s global assessment of the degree of influence of conflict but failed to measure objective indicators or information on the actual nature or form of the conflict. As an outgrowth of this study, we have since developed a brief family conflict scale that has been used successfully in a recent study to examine predictors of conflict at the end of life as reported by family members of deceased lung cancer patients. 36 Fourth, we had 14 social workers completing surveys with team members, all of whom had varying levels of experience working on the teams. Staff turnover is always a challenge in agency-based research. Inspection of means revealed that individual social workers did not rate their own cases as more or less successful as the group as a whole, with average individual scores all within 0.5, above or below the overall mean. Fortunately, all of the social workers consulted with one or two other team members ensuring consistency in reporting. Finally, the cross-sectional design of our study seriously limits any definitive conclusions regarding causality, which could only be assessed with prospective studies using time-ordered data. Given these limitations, it is important that these findings be viewed as a beginning step in understanding the role that family conflict and structures and processes of care may play in contributing to perceived success in addressing the needs of elders and family members at the end of life. In conclusion, this study was guided by a conceptual framework to provide an understanding of the determinants of perceived success in addressing end-of-life care needs in an innovative, fully integrated, managed care program serving frail elders. This study provides preliminary evidence of the importance of family conflict and structure- and process-of-care variables that may differentially influence team members ability to address the needs of elders and family members. Although this study provides preliminary evidence of differential correlates and predictors of perceived success, and the detrimental influence that family conflict has on this outcome, improving quality of end-of-life care for later life families will require a much better understanding of the kinds of assessment and intervention protocols that might prevent or ameliorate conflict and enhance structure- and process-of-care variables to facilitate more successful outcomes. Disclosures and Acknowledgments This research was supported by a grant from the John A. Hartford Foundation. The authors declare no conflicts of interest. Appreciation is extended to the Care Wisconsin Partnership Program Staff, and Administration, and to End-of-Life Committee members who provided ongoing support and consultation. Special thanks to the social workers who completed the surveys of deceased elders. References 1. NIH State-of-the-Science Conference Statement on improving end-of-life care. NIH Consens State Sci Statements 2004;21:1e Cohen SR, MacNeil C, Mount BM. Well-being at the end of life: Part 2. A research agenda for the delivery of care from the patient s perspective. Cancer Prev Control 1997;1:343e George LK. Research design in end-of-life research: state of science. Gerontologist 2002;42: 86e98.