Symptom Experience of Dying Long-Term Care Residents

Symptom Experience of Dying Long-Term Care Residents Laura C. Hanson, MD, MPH, w J. Kevin Eckert, PhD, z Debra Dobbs, PhD,§ Christianna S. Williams, PhD, Anthony J. Caprio, MD,w Philip D. Sloane, MD, k and Sheryl Zimmerman, PhD z OBJECTIVES: To describe the end-of-life symptoms of nursing home (NH) and residential care/assisted living (RC/ AL) residents, compare staff and family symptom ratings, and compare how staff assess pain and dyspnea for cognitively impaired and cognitively intact residents. DESIGN: After-death interviews. SETTING: Stratified random sample of 230 long-term care facilities in four states. PARTICIPANTS: Staff (n 5 674) and family (n 5 446) caregivers for dying residents. MEASUREMENTS: Interview items measured frequency and severity of physical symptoms, effectiveness of treatment, recommendations to improve care, and staff report of assessment. RESULTS: Decedents’ median age was 85, 89% were white, and 77% were cognitively impaired. In their last month of life, 47% had pain, 48% dyspnea, 90% problems with cleanliness, and 72% symptoms affecting intake. Problems with cleanliness, intake, and overall symptom burden were worse for decedents in NHs than for those in RC/AL. Treatment for pain and dyspnea was rated very effective for only half of decedents. For a subset of residents with both staff and family interviews (n 5 331), overall ratings of care were similar, although agreement in paired analyses was modest (kappa 5 0.043–0.425). Staff relied on nonverbal expressions to assess dyspnea but not pain. Both groups of caregivers recommended improved application of treatment and increased staffing to improve care. CONCLUSION: In NHs and RC/AL, dying residents have high rates of physical symptoms and need for more-effective From the Cecil G. Sheps Center for Health Services Research, wDivision of Geriatric Medicine, Department of Medicine, kDepartment of Family Medicine, and zSchool of Social Work, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina; zDepartment of Sociology and Anthropology, Erickson School of Aging Studies, University of Maryland Baltimore County, Baltimore, Maryland; and §School of Aging Studies, University of South Florida, Tampa, Florida. Presented at the American Academy of Hospice and Palliative Medicine National meeting, February 8, 2006, Nashville, Tennessee. Address correspondence to Laura C. Hanson, MD, MPH, Division of Geriatric Medicine, CB 7550, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599. E-mail: [email protected] DOI: 10.1111/j.1532-5415.2007.01388.x JAGS 56:91–98, 2008 r 2007, Copyright the Authors Journal compilation r 2008, The American Geriatrics Society palliation of symptoms near the end of life. J Am Geriatr Soc 56:91–98, 2008. Key words: symptoms; long-term care; palliative care T he number of Americans dying in long-term care (LTC) settings is increasing, from 18% of deaths in 1986 to 23% of deaths in 2001.1,2 Their experience of symptoms is a major determinant of quality of life near the end of life. Prior research has demonstrated that at least one of these symptomsFpainFis undertreated for residents of nursing homes (NHs).3–7 Prevalence of pain for NH residents ranges from 45% to 86%, and 15% have persistent pain over 2 to 6 months of follow-up.8–10 In addition to pain, dying NH residents experience dyspnea, problems with cleanliness, delirium, fatigue, and poor intake.11–13 It is plausible that these symptoms are equally common for residents who die in residential care/ assisted living (RC/AL) settings, although little research examines this question.14 Only one previous study has examined end-of-life symptoms in RC/AL. In this study, staff ratings of residents’ overall discomfort during the last month of life were similar in NH and RC/AL facilities.15 LTC residents are highly dependent on staff and family caregivers to assess and seek treatment for symptoms. Little is known about how staff and family caregivers perceive the severity of symptoms or effectiveness of treatment. Also unknown is the degree to which family and staff agree on residents’ symptoms and how staff evaluate symptoms for cognitively impaired residents. Therefore, a study was designed to provide systematic data on end-of-life symptoms in LTC. Structured after-death interviews were conducted with staff and family caregivers for decedents in 230 LTC facilities to describe the prevalence and severity of physical symptoms during the last month of life for residents of NH and RC/AL facilities, compare staff and family caregiver ratings of residents’ symptoms and effectiveness of treatment, and compare how staff assess pain and dyspnea for cognitively impaired and cognitively intact residents. As a secondary aim, staff and family respondents were also asked to suggest ways to improve symptom care in these settings. 0002-8614/08/$15.00 92 HANSON ET AL. METHODS In a stratified random sample of LTC facilities, consecutive decedents were identified, and structured after-death interviews were completed with staff and family caregivers. Participants Study participants were identified within a stratified random sample of 230 LTC sites in Florida, New Jersey, North Carolina, and Maryland participating in the Collaborative Studies of Long-Term Care.16 Sites included 199 RC/AL facilities and 31 NHs. Investigators identified residents who had lived in the facility at least 15 days of their final month of life and who died in the facility or less than 3 days after transfer to another healthcare setting. Consecutive decedents were identified from July 2002 until December 2004. Primary family and staff caregivers for these decedents were eligible for the study. The family caregiver was defined as the person most involved in care during the last month of life. The staff caregiver was defined as the staff member who knew the resident best and provided direct care, care supervision, medication administration, family communication, or care coordination. Staff caregivers could be interviewed about more than one decedent. Trained interviewers contacted each caregiver for a telephone interview. Interviews averaged 40 minutes for staff and 60 minutes for family caregivers. Sixty-three percent of staff and 40% of family interviews were completed within 3 months of death, and 97% of both types were completed within 6 months. Interviewers collected data from staff or family on 792 decedents, or 78% of 1,020 eligible deaths, during the study enrollment period. Staff interviews were completed for 677 decedents, or 66% of eligible deaths. Family interviews were completed for 451 decedents, or 49% of the 925 eligible deaths for which a family caregiver could be identified. Three staff interviews had incomplete data on symptoms; they were excluded from analyses, for a final sample size of 674 staff interviews. Five family interviews had incomplete data on symptoms and were excluded from analyses. To meet the second aim of this study, a subset of 331 paired staff and family caregiver interviews on the same decedent were used. Deaths for which no interview data could be obtained occurred more often in a NH (68% nonrespondents vs 56% respondents, P 5.04) and more often in larger facilities (101 vs 87 beds, P 5.04). Decedents without family caregiver interviews were younger (84 vs 86, P 5.03) and more often African American (12% vs 7%, P 5.01). Participants gave verbal informed consent to be interviewed, and the University of North Carolina School of Medicine institutional review board approved all study procedures. Demographic and Symptom Measures Staff respondents provided data on their demographics and job role, as well as decedents’ sex, race or ethnicity, and place of death. Staff reported on decedents’ functional and cognitive status in the last 3 months of life and all known causes of death. Family caregivers answered questions about their own sex, race or ethnicity, and relationship to the decedent. JANUARY 2008–VOL. 56, NO. 1 JAGS Staff and family caregivers answered parallel questions about the presence, frequency, severity, and treatment of four symptoms: pain, dyspnea (shortness of breath), problems with personal cleanliness, and decreased intake of food and water. For pain and dyspnea, interviewers asked, ‘‘Did [RESIDENT] experience [SYMPTOM] at any time during the last month of life?’’ If the respondent indicated no pain or dyspnea, the interviewer clarified whether this resulted from effective treatment or the absence of the symptom. If pain or dyspnea was present, the interviewer asked about severity on a typical day during the last month of life and asked the respondent to rate effectiveness of treatment on a 4-point Likert scale. To measure problems with cleanliness during the last month of life, interviewers asked about the resident’s need for assistance to maintain cleanliness of skin and hair, mouth, and bottom area during the last month of life. Respondents rated the frequency of each type of cleanliness problem and the effectiveness of overall care for cleanliness on 4-point Likert scales. To measure problems with intake, interviewers asked, ‘‘During the last month of life, did [RESIDENT] take less food or water by mouth than necessary for [HIM or HER]?’’ and rated the frequency of this problem on a 4-point Likert scale. Staff respondents reported whether anorexia, choking, nausea, or depressed alertness caused reduced intake. Family rated the effectiveness of supportive care to promote intake. Finally, for all symptoms, interviewers asked brief open-ended questions of family and staff respondents to seek suggestions to improve treatment. A summary symptom burden score was constructed to permit comparison of family and staff caregiver perceptions of overall symptom burden. The symptom burden score ranges from 0 to 36 and includes 9 points for each of four symptoms: pain (severity), dyspnea (severity), personal cleanliness (frequency of needing help for cleanliness of skin and hair, mouth, and bottom area), and nutrition and hydration (frequency of taking less than necessary). To weight symptoms equally within the score, each of the four symptoms could contribute 0 to 9 points to the overall score. For pain and dyspnea points were based on severity (0 5 none, 3 5 mild, 6 5 moderate, 9 5 severe or horrible). For problems with intake, points were based on frequency (0 5 never or almost never, 3 5 less than half the days, 6 5 more than half the days, 9 5 daily or almost daily). The three cleanliness items (frequency of needing assistance for skin and hair, mouth and teeth, and bottom area) were each based on frequency (0 5 never or almost never, 1 5 sometimes, 2 5 most of the time, 3 5 daily or almost daily), and were summed for a possible 9 points. Staff Report of Symptom Expression According to Cognitive Status Interviewers asked staff open-ended items to learn more about how they evaluate pain or dyspnea. When staff indicated that either symptom was present, the interviewer asked, ‘‘How did [RESIDENT] typically express [HIS or HER] pain (shortness of breath)?’’ The interviewer also asked, ‘‘Under what conditions, circumstances, or situations did the pain (shortness of breath) most often occur or worsen?’’ Staff reports of these observations were stratified JAGS JANUARY 2008–VOL. 56, NO. 1 according to residents’ cognitive status. Resident were classified as cognitively impaired if staff reported that they had dementia in the last 3 months of life or were unable to speak or write in a meaningful way during the last month of life. Analysis Data were double entered using a Microsoft Access database (Microsoft Corp., Redmond, WA), and descriptive and comparative statistics were computed using SAS 9.1 (SAS Institute, Inc., Cary, NC). To test whether symptom presence and severity differed between NH and RC/AL residents, generalized estimating equations (GEEs) were applied to logistic models, adjusting for clustering of residents within staff respondents and staff within facility, using GEE empirical standard error estimates and an exchangeable correlation matrix.17 To compare paired staff and family responses, McNemar’s test was used for categorical variables (e.g., presence of any pain) and a paired t-test for the symptom burden score. Agreement between paired family and staff responses was assessed using kappa for categorical measures and the intraclass correlation coefficient for the symptom burden score. To analyze responses to open-ended items, three investigators (LH, KE, CW) used consensus coding to categorize text. Items coded in this manner included questions about how pain and dyspnea were expressed, circumstances of worst symptoms, and ways to improve treatment of each symptom. Code development was iterative and inductive, arising from investigators’ independent review and coding of all text followed by discussion to achieve consensus. Investigators then compared draft categories and their application to individual items until they reached consensus on codes and their definitions. Coders were blinded to the cognitive status of residents. To test the reliability of the coding approach, one additional investigator (DD) applied the coding scheme to 30 responses for each of the open-ended items in the interviews. Application of the codes was compared with coding of the same 30 items by a second investigator (LH) to establish interrater reliability. Percentage agreement was above 70% for all items except responses to, ‘‘What was the best thing about how this facility provides care for eating and drinking?’’ This item was dropped from further analysis based on poor reliability. To test whether the method of expression of pain and dyspnea differed between cognitively intact and impaired residents, GEEs applied to logistic models were used as described above. RESULTS The 674 decedents with a complete staff caregiver interview formed the study sample. Just over half the deaths were from NH facilities (54%), and deaths were evenly distributed from the four states. Nearly all deaths occurred in LTC facilities (85%). Decedents’ average age at death was 85; 69% were female, and 89% were white, non-Hispanic. Seventy-seven percent were cognitively impaired during their last month of life. During the final 3 months of life, more than 60% were dependent for transfer and ambulation (Table 1). According to staff, dementia and problems with food and water intake 93 SYMPTOM EXPERIENCE IN LONG-TERM CARE Table 1. Characteristics of Decedents According to Staff Report (n 5 674) Characteristic Age at death, mean  standard deviation Female, % Race or ethnicity, % White, non-Hispanic African American Hispanic Function in 3 months preceding death, % Cognitively impaired Unable to transfer Unable to walk or wheel Unable to feed self Cause of death, % Dementia Problems eating and drinking Heart disease Infection Cancer Chronic lung disease Site of death, % Long-term care facility Hospital Other How long death expected by primary staff caregiver, % Not expected 0–7 days 8–14 days 15–28 days 428 days Value 85.4  9.5 69 89 9 2 77 65 61 39 68 62 40 28 19 18 85 14 1 33 34 15 4 15  Categories were not mutually exclusive; staff could report more than one impaired function or cause of death for a decedent. were common causes of death. For two-thirds of the residents, staff reported that they expected the death for less than a week or did not expect the resident to die. A total of 331 unique staff caregivers provided data on 674 decedents. Individual staff caregivers provided data on one decedent (59%), two decedents (18%), or three or more decedents (23%). Nearly all staff respondents were women (93%). Fifty-eight percent described themselves as white, non-Hispanic, and 29% were African American. Staff roles of respondents were nursing aide (33%), licensed practical nurse (31%), registered nurse (22%), and administrator (10%). Staff cared for decedents for an average of 21 months before death. Of 331 family caregiver respondents who were paired to staff interviews about the same decedent, 72% were women, and 92% were white, non-Hispanic. Forty-seven percent were daughters or daughters-in-law, 19% were sons or sons-in-law, and 8% were spouses of the resident. Presence and Severity of Symptoms Staff respondents reported that nearly half of LTC residents had pain during the last month of life; on a typical day 94 JANUARY 2008–VOL. 56, NO. 1 HANSON ET AL. during the last month of life, 8% of dying residents experienced mild pain, 24% moderate pain, and 15% severe or horrible pain. For 41% of dying residents, pain was never a problem, and 12% had no pain, because it was effectively treated (Table 2). Staff reported that nearly half of LTC residents had dyspnea during their last month of life; on a typical day during the last month of life, 13% of dying residents had mild, 24% had moderate, and 11% had severe or very severe dyspnea. For 49% of dying residents, dyspnea was never a problem, and 3% had no dyspnea, because it was effectively treated. Nearly all LTC residents needed assistance to maintain cleanliness of skin and hair during their last month of life (90%) and to keep their mouth (84%) or bottom area clean (84%). Twenty-two percent of residents also had a skin ulcer during their last month of life (data not shown). During the last month of life, staff perceived that 72% of dying residents had diminished intake of food and water. JAGS Of all dying residents, 28% had problems affecting intake always or almost always during their last month. Thirtythree percent of the decedents had an episode of dehydration, and 32% lost a noticeable amount of weight before death. Symptoms affecting intake included anorexia (65%), decreased alertness (29%), choking (16%), and nausea (8%); these symptoms were not mutually exclusive. End-of-Life Symptoms in NHs Versus RC/AL Facilities End-of-life symptoms experienced by NH residents were compared with those of residents in RC/AL facilities. Frequency of pain and dyspnea were similar, as was staff report of effectiveness of treatment for these symptoms, although dying NH residents were more likely to have problems with cleanliness and with intake of food and water during their final month of life (Table 2). During the final month of life, residents in both types of facilities had similar rates of dehydration (36% vs 30%, P 5.11) and of weight loss (28% Table 2. Staff Report of Presence and Severity of Symptoms in the Last Month of Life According to Facility Type NH (n 5 363) n (%) Presence and Severity of Symptom Pain: presence or severity None, never a problem None, treatment effective Mild Moderate Severe or horrible Dyspnea: presence or severity None, never an issue None, treatment effective Mild Moderate Severe or horrible Cleanliness care needed: always or almost always With skin With bottom area With mouth Frequency of inadequate food and water intake Never or almost never Sometimes Most of the time Always or almost always Episode of dehydration Noticeable weight loss Anorexia Less alert Choking Nausea Symptom burden score, mean  standard deviation§  RC/AL (n 5 311) P-Value Overall (N 5 674) n (%) 126 (36) 54 (16) 20 (6) 100 (29) 48 (14) 139 (46) 26 (9) 33 (11) 58 (19) 46 (15) .53w .43z 265 (41) 80 (12) 53 (8) 158 (24) 94 (15) 176 (49) 13 (4) 46 (13) 88 (25) 33 (9) 149 (49) 5 (2) 41 (14) 69 (23) 40 (13) .58w .25z 325 (49) 18 (3) 87 (13) 157 (24) 73 (11) 350 (96) 338 (93) 338 (93) 254 (82) 230 (74) 224 (72) o.001 o.001 o.001 604 (90) 568 (84) 562 (84) 75 (21) 70 (19) 90 (25) 127 (35) 126 (36) 101 (28) 252 (70) 129 (36) 61 (17) 24 (7) 19.8  6.7 106 (34) 81 (26) 60 (19) 64 (21) 93 (30) 111(36) 182 (59) 68 (22) 47 (15) 26 (8) 16.8  7.6 o.001 181 (27) 151 (22) 150 (22) 191 (28) 219 (33) 212 (32) 434 (65) 197 (29) 108 (16) 50 (8) 18.3  7.2 .11 .24 .004 o.001 .47 .52 o.001 For difference in proportions between nursing home (NH) and residential care/assisted living (RC/AL). Based on generalized estimating equations (GEEs) applied to logistic models (or linear models for pain severity, dyspnea severity, frequency of inadequate food or water intake, and symptom burden score) and adjusting for subject clustering by facility and staff respondents within facility, using GEE empirical standard error estimates and exchangeable correlation. w For any pain or dyspnea versus none. z For severity of pain or dyspnea scored as follows: 1 5 none (treatment effective or never a problem), 2 5 mild, 3 5 moderate, 4 5 severe. § Based on a possible score of 0–36, higher score indicating higher symptom burden. JAGS JANUARY 2008–VOL. 56, NO. 1 SYMPTOM EXPERIENCE IN LONG-TERM CARE 95 Table 3. Comparison of Staff and Family Report of Symptoms and Treatment (331 Paired Interviews) Symptom Measure Pain Any Severe Treatment very effective Dyspnea Any Severe Treatment very effective Personal cleanliness Skin cleanliness care, always or almost always Care very effective Nutrition and hydration Problems with food or water intake Intake problems always or almost always Care very effectivew Yes Staff, % Yes Family, % P-Value Agreement, % 46 16 51 40 16 53 .06 .90 .78 60 81 48 0.191 0.271 0.043 45 13 58 43 12 54 .50 .56 .64 67 85 49 0.338 0.311 0.027 89 91 89 71 1.00 o.001 89 67 0.425 0.004 73 26 NA 68 32 55 .10 .09 NA 70 62 0.289 0.064 Kappa  McNemar test, paired comparison. w This item not asked of staff. NA 5 not applicable. vs 36%, P 5.24), although NH residents more often experienced anorexia (70% vs 59%, P 5.004) and decreased alertness (36% vs 22%, Po.001) that resulted in decreased intake. The overall symptom burden was higher for NH decedents than for residents dying in RC/AL facilities (19.8 vs 16.8, Po.001). Staff and Family Ratings of Symptoms and Treatment Of the residents who had pain in the last month of life, staff reported that nearly all (98%) had medication ordered and 35% had nonmedication treatments. Of the residents who had shortness of breath, 64% had medication ordered, and 85% had nonmedication treatments such as supplemental oxygen. For problems with intake of food and water, staff reported that 5% of residents who took less food or water by mouth than necessary had a feeding tube, 9% received intravenous fluids, and 21% were prescribed an appetite stimulant. Most residents with intake problems received modified diets (87%) or nutritional supplements (82%) as part of their treatment. Using the subset of 331 deaths for which staff and family respondents answered parallel questions about symptoms, ratings of symptoms and of treatment effectiveness were compared (Table 3). Similar proportions of staff and family caregivers reported that the residents had pain, shortness of breath, problems with cleanliness, and problems with intake. On items rating treatment effectiveness, just over half of staff and family respondents reported that treatment for pain and dyspnea were very effective, but fewer family caregivers rated cleanliness care as very effective (71% vs 91%, Po.001, McNemar test paired comparison). Percentage agreement and kappa statistics were also calculated to assess concordance between family and staff respondents on each item; raw agreement ranged from 48% to 89%, and kappas ranged from 0.043 to 0.425. The highest level of agreement in paired comparisons was reached for the item measuring the need for help with cleanliness; all other items had only fair to poor concordance. The lowest level of agreement was evidenced for items rating treatment effectiveness. Of a possible 36 points, the average overall symptom burden score was 18.1 when rated by staff respondents and 17.5 when rated by family respondents (P 5.23, paired t-test), although in paired comparisons, the difference between staff and family symptom burden scores ranged as high as 25 points on the 36-point scale. Sixty-four percent of pairs had scores that differed by more than 4 points. The intraclass correlation between staff and family scores was 0.269. Thus, staff and family respondents gave similar overall appraisals of symptom burden for residents dying in LTC but did not demonstrate strong agreement on symptom burden experienced by individual residents. Staff Report of Symptom Expression Staff reported differences in how they knew that pain was present for cognitively intact and cognitively impaired residents. They said that 76% of cognitively intact residents but only 42% of cognitively impaired residents expressed pain verbally (Po.001). For cognitively impaired residents in pain, they more often relied on nonverbal expressions of discomfort such as changes in breathing patterns (39% vs 15%, Po.001) or moaning or crying out (27% vs 14%, Po.001) (Figure 1). They reported that pain was most often present or worsened at specific times of day or with movement. Staff were also more likely to report verbal expressions of dyspnea for cognitively intact than cognitively impaired residents (13% vs 5%, Po.05), although they more frequently evaluated dyspnea based on behaviors, regardless of cognitive status (Figure 2). Staff primarily assessed dyspnea based on breathing patterns (68% vs 69%, P 5.71). They assessed dyspnea based on response to 96 JANUARY 2008–VOL. 56, NO. 1 HANSON ET AL. 80 % of cognitively intact residents (n=80) 70 % of cognitively impaired residents (n=234) 60 50 40 30 20 10 0 Verbal* Breathing** Nonverbal** Mood** Figure 1. Staff report of residents’ expressions of pain according to cognitive status. Po.001, Po.01 for difference between cognitively intact and cognitively impaired residents, based on generalized estimating equations (GEE), applied to logistic models and adjusting for subject clustering by facility and staff respondents within facility, using GEE empirical standard error estimates and exchangeable correlation. movement for 32% of residents regardless of cognitive status. Staff also perceived that dyspnea to worsened in the context of disease progression and with movement. Improving Care for Symptoms Finally, interviewers asked staff and family respondents what should be done to improve care for pain, dyspnea, cleanliness problems, and problems with nutrition and hydration. For pain, 87% of staff and 66% of family were 70 % of cognitively intact residents (n=80) 60 % of cognitively impaired residents (n=234) 50 40 30 20 10 0 Verbal* Breathing Physical Mood Figure 2. Staff report of residents’ expressions of dyspnea by cognitive status. Po.05 for difference between cognitively intact and cognitively impaired groups, based on generalized estimating equations (GEEs), applied to logistic models and adjusting for subject clustering by facility and staff respondents within facility, using GEE empirical standard error estimates and exchangeable correlation. JAGS uncertain or had no suggested improvements. Suggestions to improve pain management included more-effective application of current treatments, additional staff time or special staff expertise, and improved communication and coordination between care providers. For dyspnea, 89% of staff and 73% of family were uncertain or had no recommendations for improvement. Respondents most often recommended more-effective application of existing treatments, enhanced staff skills, and increased staff presence to improve treatment of dyspnea. More than 90% of staff respondents were uncertain or had no recommendations for improvements in care for problems with cleanliness or intake near the end of life. In contrast, only 59% and 62% of family respondents reported no suggestions for improvement in these aspects of endof-life care. Their primary suggestions for improved quality of care were improved staffing and more-effective application of existing treatment. DISCUSSION This study reports the end-of-life symptom experience of a large and representative sample of individuals who die in LTC settings. Half of dying residents have pain and shortness of breath, three-fourths have problems affecting intake of food and water, one-third have weight loss or dehydration, and nearly all have problems staying clean in their final month of life. Symptom burden is somewhat greater for residents of NHs than for residents of RC/AL. These symptoms merit attention in the clinical palliative care of patients who die in LTC. Staff caregivers in LTC may require training to enhance the quality of care for pain and dyspnea. Increased access to hospice or palliative care consultation might improve symptom management; alternatively, increased NH professional staff time for residents with symptom needs such as pain or dyspnea might also improve care. Two descriptive studies compared NH decedents with hospice with those without hospice and reported improvements in pain management,18,19 although another found that NH staff and NH care with hospice provided similar quality of care when family members anticipated death, suggesting that emotional preparation for death is an essential component of hospice care.20 Cleanliness and supportive feeding, alternatively, are personal care services provided by nursing assistants. Thus, seriously and terminally ill LTC residents may need new reimbursement mechanisms to allow for increased direct care staffing to promote comfortable and dignified dying. This study also has implications for measurement of quality of end-of-life care. The high prevalence of dementia among LTC residents means that surrogatesFfamily or staff caregiversFwill often decide when symptoms are present and whether care is effective. It was found that family and staff provided similar overall estimates of the quality of care for symptoms, although they did not often agree on the symptom experience of an individual resident who was dying. The reason for this discordance is unknown, and neither can be considered a criterion standard surrogate. Family caregivers may know the dying person better, but staff may see them more frequently and have better symptom assessment skills. Such differences in perspective may result in conflicts over the best care for JAGS JANUARY 2008–VOL. 56, NO. 1 individual residents. Future research is needed to determine whether standardized symptom assessment methods will improve the agreement about symptoms and improve effectiveness of treatment. Symptoms are inherently a self-reported, subjective phenomenon, yet many dying patients lose the ability to report and rate their own symptoms as death approaches. This is particularly true in LTC settings, where more than half of residents have dementia. It was found that staff often use self-report of pain for cognitively intact residents but rely on behaviors to assess pain for cognitively impaired residents and to assess dyspnea regardless of cognitive status. Although this approach is intuitive, observation of pain behaviors may contribute to undertreatment of pain, and perhaps other sources of discomfort, for patients with dementia.6,21 Investigators have validated verbal and observational scales for pain in dementia.22,23 One study compared three self-report scales and one observational scale to rate pain in dementia patients. This study found modest correlation between pain ratings on self-report scales and on the observational scale; furthermore, the observational scale tended to underestimate pain relative to self-report. 24 Another study found that 25% of residents with dementia in NHs and 40% in RC/AL facilities reported pain, with 62% agreement between resident and staff reports of pain.25 One implication of these findings is that LTC staff should be trained in symptom assessment using self-report whenever possible, even for residents with dementia. These findings should be interpreted with consideration of limitations and strengths of after-death interviews. This method generates information on all deaths within a population and provides generalizable information on the dying experience, but recall bias or grief may affect after-death interviews. This study purposefully compared two surrogate respondents and found few systematic differences despite relatively poor agreement on the symptom experience of individual residents. Research in other end-of-life populations shows that healthcare providers underestimate patient pain and family caregivers may perceive symptoms as more severe.26–31 The two surrogate perspectives may be more similar in the LTC population, because both may rely on interpreting nonverbal expressions of discomfort. A second study limitation is the lower response rate for family caregiver interviews and the under-representation of African Americans and other minority populations. There were small but potentially important differences between residents whose deaths were represented by family interviews and those who were not. Potential bias introduced by nonresponse could affect analyses of staff and family agreement, whereas results based on staff interviews alone would not be affected. One in four older Americans currently receives his or her end-of-life care in a LTC facility, and this setting will become more common for terminal care as the population ages. The high prevalence of end-of-life symptoms has implications for direct-care staffing ratios, for staff training, and for targeted use of hospice and specialized palliative care services to augment symptom management. Physicians and staff caregivers will need evidence from carefully designed trials of pain and dyspnea treatment to guide practice. Furthermore, policy and reimbursement strategies SYMPTOM EXPERIENCE IN LONG-TERM CARE 97 are needed to enhance staffing and to encourage improved supportive care for problems with cleanliness and intake while dying. Lastly, although family and staff caregivers do not express a systematic bias in overall ratings of care for symptoms, they do not agree on the symptom experience of individual dying residents. Future research should test the use of structured clinical symptom assessments for their ability to improve agreement and thereby improve the quality of symptom management in LTC. ACKNOWLEDGMENTS Financial Disclosures: This study was supported by Grants R01 AG13863 and K02 AG00970 from the National Institute on Aging. All authors have declared their freedom from any financial conflicts of interest related to the content of this manuscript. Author Contributions: Hanson, Eckert, Sloane, Zimmerman: study concept and design. Dobbs, Zimmerman: acquisition of subjects and data. Hanson, Caprio, Dobbs, Eckert, Sloane, Williams, Zimmerman: analysis and interpretation of data, preparation of manuscript. Sponsor’s Role: The sponsor had no role in the design, methods, subject recruitment, data collection, analyses, or manuscript preparation. REFERENCES 1. Teno JM. Facts on Dying [on-line]. Available at www.chcr.brown.edu/dying/ factsondying.htm Accessed June 12, 2005. 2. Brock DB, Foley DJ. Demography and epidemiology of dying in the U.S. with emphasis on deaths of older persons. Hosp J 1998;13:49–60. 3. Field MJ, Cassel CK, eds. Approaching Death: Improving Care at the End of Life. Institute of Medicine. Washington, DC: National Academy Press, 1997. 4. National Consensus Project: Clinical Practice Guidelines for Quality Palliative Care [on-line]. Available at www.nationalconsensusproject.org Accessed December 15, 2005. 5. Keay TJ, Fredman L, Taler GA et al. Indicators of quality medical care for the terminally ill in nursing homes. J Am Geriatr Soc 1994;42:853–860. 6. Morrison RS, Siu AL. A comparison of pain and its treatment in advanced dementia and cognitively intact patients with hip fracture. J Pain Symptom Manage 2000;19:240–248. 7. Wu N, Miller SC, Lapane K et al. Impact of cognitive function on assessments of nursing home residents’ pain. Med Care 2005;43:934–939. 8. Bernabei R, Gambassi G, Lapane K et al. Management of pain in elderly cancer patients. JAMA 1998;279:1877–1882. 9. Ferrell BA. Pain evaluation and management in the nursing home. Ann Intern Med 1995;123:681–687. 10. Teno JM, Weitzen S, Wetle T et al. Persistent pain in nursing home residents. JAMA 2001;285:2081. 11. Reynolds K, Henderson M, Schulman A et al. Needs of the dying in nursing homes. J Palliat Med 2002;5:895–901. 12. Brandt HE, Deliens L, Ooms ME et al. Symptoms, signs, problems and diseases of terminally ill nursing home patients. Arch Intern Med 2005;165:314– 320. 13. Hall P, Schroder C, Weaver L. The last 48 hours of life in long-term care: A focused chart audit. J Am Geriatr Soc 2002;50:501–506. 14. Zimmerman S, Gruber-Baldini AL, Sloane PD et al. Assisted living and nursing homes: Apples and oranges? Gerontologist 2003;43:107–117. 15. Sloane PD, Zimmerman S, Hanson L et al. End-of-life care in assisted living and related residential care settings: Comparison with nursing homes. J Am Geriatr Soc 2003;51:1587–1594. 16. Zimmerman S, Sloane PD, Eckert JK et al. An overview of the Collaborative Studies of Long-Term Care. In: Zimmerman S, Sloane PD, Eckert JK, eds. Assisted Living. Baltimore: Johns Hopkins Press, 2001, pp 117–143. 17. Diggle PJ, Heagerty P, Liang K-Y et al. The Analysis of Longitudinal Data. Oxford: Oxford University Press, 2002. 18. Baer WM, Hanson LC. Families’ perception of the added value of hospice in the nursing home. J Am Geriatr Soc 2000;48:879–882. 19. Miller SC, Gozalo P, Mor V. Hospice enrollment and hospitalization of dying nursing home patients. Am J Med 2001;111:38–44. 98 HANSON ET AL. 20. Munn JC, Hanson LC, Zimmerman S et al. Is hospice associated with improved end of life care in nursing homes and assisted living facilities? J Am Geriatr Soc 2006;54:490–495. 21. Parmelee PA. Pain in cognitively impaired older persons. Clin Geriatr Med 1996;12:473–487. 22. Stolee P, Hillier LM, Esbaugh J et al. Instruments for the assessment of pain in older persons with cognitive impairment. J Am Geriatr Soc 2005;53: 319–326. 23. Krulewitch H, London MR, Skakel VJ et al. Assessment of pain in cognitively impaired older adults: A comparison of pain assessment tools and their use by nonprofessional caregivers. J Am Geriatr Soc 2000;48:1607–1611. 24. Pautex S, Michon A, Guedira M et al. Pain in severe dementia: Self-assessment or observational scales? J Am Geriatr Soc 2006;54:1040–1045. 25. Williams CS, Zimmerman S, Sloane PD et al. Characteristics associated with pain in long-term care residents with dementia. Gerontologist 2005;45:68–73. JANUARY 2008–VOL. 56, NO. 1 JAGS 26. Sengstaken EA, King SA. The problems of pain and its detection among geriatric nursing home residents. J Am Geriatr Soc 1993;41:541–544. 27. Burrows AB, Satlin A, Salzman C et al. Depression in a long-term care facility: Clinical features and discordance between nursing assessment and patient interviews. J Am Geriatr Soc 1995;43:1118–1122. 28. Higginson I, Wade A, McCarthy M. Palliative care: Views of patients and their families. BMJ 1990;301:277–281. 29. Laitinen P. Elderly patients’ and their informal caregivers perceptions of care given: The study-control ward design. J Adv Nurs 1994;20:71–76. 30. Addington-Hall J, McPherson C. After death interviews with surrogates/ bereaved family members: Some issues of validity. J Pain Symptom Manage 2001;22:784–790. 31. Neumann PJ, Araki SS, Gutterman EM. The use of proxy respondents in studies of older adults: Lessons, challenges, and opportunities. J Am Geriatr Soc 2000;48:1646–1654.