Scales for Evaluation of End-of-Life Care in Dementia

Alzheimer Disease and Associated Disorders Vol. 15, No. 4, pp. 194 200 2001 Lippincott Williams & Wilkins, Inc., Philadelphia Scales for Evaluation of End-of-Life Care in Dementia * Ladislav Volicer, * Ann C. Hurley, and * Zuzka V. Blasi *Geriatrics Research Education Clinical Center, E.N. Rogers Memorial Veterans Hospital, Bedford, Massachusetts, U.S.A.; Boston University School of Medicine, Boston, Massachusetts; Northeastern University, School of Nursing, Boston, Massachusetts, U.S.A.; and Education Development Center, Newton, Massachusetts, U.S.A. Summary: Systematic evaluation of end-of-life care in dementia has been hampered by a lack of instruments to specifically address those issues that are unique for persons who are dying with dementia. This study evaluated psychometric properties of three scales designed to measure outcomes of care of persons suffering from terminal dementia. A survey of family caregivers whose loved one died during the past year was conducted using a questionnaire that included questions regarding satisfaction with care, physical and emotional symptoms that occurred during the last 90 days of the care recipient s life, and comfort during the dying process. Three scales were developed based on responses from 156 questionnaires: Satisfaction with Care at the End-of-Life in Dementia (SWC-EOLD), Symptom Management at the End-of-Life in Dementia (SM-EOLD) with Physical and Psychological Symptoms subscales, and Comfort Assessment in Dying with Dementia (CAD-EOLD) with four subscales: Physical Distress, Dying Symptoms, Emotional Distress, and Well Being. The three scales developed and evaluated in this study can be used as outcome measures in studies investigating effectiveness of interventions aimed to improve end-of-life care for individuals with dementia. Key Words: Dementia End-of-life care Evaluation comfort Satisfaction. With the prolongation of the human life span and development of technological advances for management of life-threatening conditions, end-of-life care is becoming an increasingly important component of the health care system. Although everybody must die eventually, the dying process may be prolonged by medical interventions that are focused at extending life. These interventions are not always desirable because they may lead to considerable suffering of the patient and cause family stress (Covinsky, Goldman, et al. 1994). Unfortunately, there is a tendency of the medical system to concentrate Received November 20, 2000. Revised August 16, 2000. Accepted April 6, 2001. Address correspondence and reprint requests to Dr. Ladislav Volicer, GRECC (182B), E.N. Rogers Memorial Veterans Hospital, 200 Springs Road, Bedford, MA 01730, U.S.A. This work was supported by the Alzheimer s Association, USPHS grant P30 AG13846, and Department of Veterans Affairs. on life-prolonging interventions instead of providing pain control and spiritual support during the dying process (The SUPPORT Principal investigators 1995). Therefore, it is important to monitor the quality of endof-life care and strive for its improvement. Measuring quality of health care is difficult and endof-life care poses additional obstacles. Optimal end-oflife care may differ from individual to individual depending on their religious and philosophical values. Thus, quality of end-of-life care, similar to the quality of life concept, is subjective and depends on a person s own evaluation. Such an evaluation is impossible if the patient is cognitively impaired. Quality of end-of-life care for individuals dying with severe and terminal dementia must be evaluated by their caregivers who use verbal and non-verbal cues to monitor the patient s condition. Evaluations of existing care practices and of care improvement initiatives require operational definitions and the ability to measure characteristics of specific end-of- 194

SCALES FOR EVALUATION OF END-OF-LIFE CARE IN DEMENTIA 195 life care components. General principles regarding quality of care at the end of life, which include 10 domains, were published (Lynn, 1997) and modified into 12 domains when applied to dementia care (Teno, Landrum, et al., 1997). Although there are measurement tools available for some of these domains, e.g. family burden, these instruments do not specifically address issues of end-oflife care. Therefore, we developed a questionnaire regarding care provided to a demented individual during the last 90 days of life. We conducted a survey of family caregivers whose care recipients had died during the last year and used that data to develop specific instruments for the evaluation of end-of-life care in dementia. METHODS Information was obtained by asking caregivers of relatives with dementia who had died within the last year to return an anonymous questionnaire. This project was approved by the appropriate Institutional Review Boards. All chapters of the Alzheimer s Association (n 201), Veterans Administration Geriatric Research Educations Clinical Centers (GRECCs) (n 16) and the National Institute on Aging Alzheimer s Disease Centers (ADCs) (n 27) were contacted to ascertain their ability and willingness to distribute questionnaires to caregivers whose care recipient had died within the last year. Thirty eight Alzheimer chapters, 12 GRECCs, and 5 ADCs agreed to collaborate with us and to distribute the questionnaires, and requested 938 copies of the questionnaire from us. The caregivers were given the questionnaire by the collaborators and asked to return the questionnaire to us in an enclosed stamped return envelope. We attempted to ascertain how many questionnaires were actually distributed to the caregivers by asking the collaborators for this information, but only 29 of the 55 collaborators responded to this inquiry. According to their information, approximately 61% of questionnaires (n 572) that we distributed were mailed or given to caregivers. This is most likely an overestimate, because the collaborators who distributed smaller proportions of questionnaires were probably less likely to respond to our inquiry regarding number of distributed questionnaires. At the end of a 9-month collection period, we received 156 questionnaires from caregivers. If the estimated number of distributed questionnaires is correct, our response rate was 27.3%. Questionnaire Content The questionnaire asked caregivers about their care recipient s last 90 days of life and included demographic information, social support, caregiver burden, family impact, service utilization, advance directives, and resource consumption. Dependence in activities of daily living (ADL) was measured from caregiver report by the six item Katz Index of ADL (.87) (Katz, Ford, et al., 1963). The overall severity of dementia was measured by the Bedford Alzheimer Nursing Scale - Severity Subscale (BANS-S), a seven-item scale ranging from 7 (no impairment) to 28 (most severe impairment) (Volicer, Hurley, et al., 1994). This scale combines ratings of interaction ability (speech, eye contact), functional deficits (dressing, eating, ambulation), and occurrence of pathological symptoms (sleep-wake cycle disturbance, muscle rigidity/contractures) (.80). The BANS-S is more sensitive to detecting disease progression beyond the severe stage than scales that measure only cognitive or functional deficits and is directly related to the density of neurofibrillary tangles observed after autopsy (Volicer, Hurley, et al., 1994) and there is a good agreement of BANS-S scores obtained from nursing staff and family members (Hurley, Johnson, et al., 1996). Three blocks of questions were included to ascertain the caregiver s perception of quality of care. Block one (n 15) addressed satisfaction with care and was based on the Medicaring questionnaire using a four-point scale (Skolnick 1998). Answers to each statement ranged on a four-point scale from strongly agree (4) to strongly disagree (1) and included a not applicable selection. Negative questions were reverse-coded so that, for all responses, higher numbers represented more satisfaction. Not applicable responses were recoded to missing. The second block was constructed as an array of physical and emotional symptoms and signs common at the end of life for persons with dementia that occurred during the last 90 days of the care recipient s life. We included symptoms proposed by the American Geriatrics Society (Lynn, 1997): pain, shortness of breath, fatigue, depression, fear, anxiety, skin breakdown. A panel of experts recommended addition of difficulty swallowing, calm, happiness, agitation, and resistiveness to care. For each symptom, the respondents were asked to rate its frequency with choices: never (5), once a month (4), two to three days a month (3), once a week (2), two to three days a week (1), and every day (0). The third block was constructed as an array of the symptoms and conditions commonly observed during the dying process in persons with dementia (Smith, 1998): discomfort, pain, serenity, restlessness, shortness of breath, peace, choking, gurgling, difficulty swallowing, emotional support, fear, anxiety, calm, crying, and moaning. For each symptom, the respondents were asked to rate its intensity during the dying process with choices: not at all (3), somewhat (2), a lot (1).

196 L. VOLICER ET AL. Data Analysis Respondents had to answer all items in a block of questions for their questionnaires to be used for examining psychometric characteristics of each of the three scales. We computed measures of central tendency and dispersion of responses for each item, item-item correlations and corrected item-total correlations. We eliminated items that were answered does not apply by a significant number of respondents. In the interest of brevity, we also eliminated five items from the Satisfaction with Care scale that did not significantly influence the scale reliability. For each individual scale, we performed an exploratory factor analysis with varimax rotation to establish possible subscales. s were also deleted if the response rate for that item was low, if the corrected item-total correlation was low (<.3), and if item deletion significantly increased the internal consistency of the scale or a subscale. Our goal was to select items that were related conceptually but not duplicative to avoid redundancy, assure internal consistency, and have a parsimonious tool. Reliability of the scales and subscales was determined by calculation of coefficient alpha. Distribution of values in each scale was compared with normal distribution by calculating skewness and kurtosis. RESULTS Demographic Characteristics Caregiver respondents were mostly white (97%) and females (75%), and their mean level of education was 14.4 ± 2.7 years. Before the care recipient died, most of the caregivers were married (92%), 5% were divorced, 2% single, and 1% widowed. Care recipients were almost equally males and females (55 and 45%), and also predominantly white (98%). Most of the care recipients were husbands of the caregivers (45%), 24% were mothers, 19% wives, and 10% fathers. At the time of death, the mean age of care recipients was 81.2 ± 7.5 years ranging from 55 to 103 years. Most care recipients had severe dementia during the last 90 days of their lives; 42% were completely dependent in all activities of daily living and their mean score on the Bedford Alzheimer Nursing Scale-Severity was 20.4 ± 4.5 (four care recipients had a score of 28 that indicates maximal impairment). Satisfaction With Care at the End of Life in Dementia (SWC-EOLD) Most items included in the scale were considered applicable by more than 94% of the respondents with the exception of the statement More should have been done to keep my loved one free from pain, which was answered as Does Not Apply by 12.6% of the respondents. Because this item also strongly correlated with item All measures were taken to keep my loved one comfortable (r.56, p < 0.001) it was deleted. Factor analysis of the scale items showed that the scale consisted of only one factor. Four other items had high corrected item-total correlations (r >.8) and were deleted. The frequency of the 10 remaining items, and their itemtotal correlations are listed in Table 1. The possible scores of this scale range from 10 to 40 with a higher score indicating more satisfaction. The scores in our sample of completed questionnaires (n 130) ranged from 13 to 40, with a mean of 30.5, standard deviation of 6.0, and alpha.90. The distribution of scores (Fig. 1) was not significantly different from normal distribution (skewness -.411 ±.212, kurtosis -.160 ±.422). TABLE 1. Satisfaction with care at the end of life in dementia (SWC-EOLD) n Strongly disagree 1 Frequencies (%) Disagree 2 Agree 3 Strongly agree 4 Corrected item-total correlation I felt fully involved in all decision making 150 3.3 10.7 41.3 44.7.63 I would probably have made different decisions if I had had more information* 143 23.8 51.0 16.1 9.1.61 All measures were taken to keep my care recipient comfortable 154 3.9 9.7 46.1 40.3.63 The health care team was sensitive to my needs and feelings 149 3.4 8.7 53.7 34.2.69 I did not really understand my care recipient s condition* 149 34.2 46.3 16.1 3.4.33 I always knew which doctor or nurse was in charge of my care recipient s care 153 6.5 21.6 48.4 23.5.54 I feel that my care recipient got all necessary nursing assistance 153 4.6 19.0 47.7 28.8.79 I felt that all medication issues were clearly explained to me 151 4.6 17.2 46.4 31.8.77 My care recipient received all treatments or interventions that he or she could have benefited from 153 6.5 18.3 47.7 27.5.76 I feel that my care recipient needed better medical care at the end of his or her life* 152 27.0 48.7 17.8 6.6.75 *Reverse coded for calculation of the total score.

SCALES FOR EVALUATION OF END-OF-LIFE CARE IN DEMENTIA 197 FIG. 1. Distribution of Satisfaction with Care at the End of Life in Dementia (SWC-EOLD) scores. Symptom Management at the End of Life in Dementia (SM-EOLD) Answers to items of this scale indicated the condition of the care recipient during the last 90 days of his/her life. An item Fatigue was dropped because it was infrequently observed ( never in 74.8% of respondents). Factor analysis identified two factors. Two items ( Difficulty Swallowing and Happiness ) did not load significantly (r <.3) on either of the two factors identified in the scale and their deletion increased alpha. Therefore, we retained 9 items (Table 2). The possible scores of this scale range from 0 to 45 with a higher score indicating better symptom control. The scores in our sample of completed questionnaires (n 105) ranged from 2 to 45, with a mean of 21.3 and standard deviation of 10.9. The distribution of scores (Fig. 2) was not significantly different from normal distribution (skewness.361 ±.238, kurtosis -.691 ± 472). FIG. 2. Distribution of Symptom Management at the End of Life in Dementia (SM-EOLD) scores. The factor structure of SM-EOLD is shown in Table 3. The first factor represents psychological symptoms and states ( Calm, Depression, Fear, Anxiety, Agitation, and Resistiveness to Care ) while the second factor represents physical symptoms ( Pain, Shortness of Breath, and Skin Breakdown ). Reliability coefficients alpha of the whole scale and of the first factor are excellent (.78 and.81) while the alpha of the second factor is poor (.47) caused by small number of items and low correlation among items. The scores of the first factor ranged from 0 to 30, with a mean of 13.3 and standard deviation of 8.6. The scores of the second factor ranged from 0 to 15, with mean of 7.7 and standard deviation of 4.3. However, in view of the poor alpha of this factor, these important factors of care may have to be assessed independently by single items. TABLE 2. Symptom management at the end of life in dementia (SM-EOLD) Never Once a month 2 or 3 days a month Frequencies Once a week Several days a week Every day Corrected item-total correlation Sub-scale 1 2 Pain 39.0 19.9 5.9 8.8 5.1 21.3.40.44 Shortness of breath 27.9 16.2 4.4 5.9 5.9 39.7.26.18 Skin breakdown 23.0 8.6 7.9 7.2 13.7 39.6.22.30 Calm 11.9 3.2 7.1 7.9 22.2 47.6.33.24 Depression 49.2 13.3 4.7 4.7 8.6 19.5.54.53 Fear 37.5 14.8 7.0 3.1 7.8 29.7.64.62 Anxiety 47.3 15.3 9.2 3.8 6.9 17.6.72.66 Agitation 43.0 18.5 12.6 3.7 5.2 17.0.69.64 Resistiveness to care 25.9 18.7 10.8 4.3 7.2 33.1.54.59. Total scale

198 L. VOLICER ET AL. TABLE 3. Rotated factor structure of symptom management at the end of life in dementia scale (SM-EOLD) Factor loading Factor 1 Factor 2 Calm.50.24 Depression.66.25 Fear.73.30 Anxiety.82.15 Agitation.81.10 Resistiveness to care.70.16 Pain.31.65 Shortness of breath.03.73 Skin breakdown.31.31 Eigenvalue 1.49 3.75 Percent variance 14.9 37.5 Loadings used to identify factors are underlined. Comfort Assessment in Dying With Dementia (CAD-EOLD) Answers to items of this scale indicated the condition of the care recipient during the dying process. Factor analysis showed that the scale items could be separated into four subscales. The original questionnaire contained 15 items but the item Emotional Support significantly decreased the alpha of the scale and was deleted. The possible scores of the resulting scale range from 14 to 42 with a higher score indicating better comfort level. The scores in our sample of completed questionnaires (n 95) ranged from 17 to 42, with mean of 31.4 and standard deviation of 5.9. The distribution of scores (Fig. 3) was not significantly different from normal distribution (skewness -.338 ±.247, kurtosis -.589 ±.490). FIG. 3. Distribution of Comfort Assessment in Dying with Dementia (CAD-EOLD) scores. The subscales represented Physical Distress, Emotional Distress, Well Being, and Dying Symptoms (Table 4). Physical Distress includes discomfort, pain, shortness of breath, and restlessness. Emotional Distress includes anxiety, fear, moaning, and crying. Well Being includes serenity, peace, and calm. Dying Symptoms includes choking, gurgling, difficulty swallowing, and shortness of breath. Reliability by coefficient alpha was excellent for the whole scale (.85) as well as individual subscales (Physical Distress.74, Dying Symptoms.70, Emotional Distress.82, Well Being.80). The mean scores, standard deviations and ranges of the CAD-EOLD subscales are summarized in Table 6. TABLE 4. Comfort assessment in dying with dementia (CAD-EOLD) scale A lot Frequencies Correlations Among Scales Table 7 presents a summary of the scales indicating number of items and alpha coefficients. All three total scale scores and most subscale scores were significantly correlated with each other. The only exceptions were the Emotional subscale of CAD-EOLD that did not correlate with SWC-EOLD and Dying Symptom subscale that did not correlate with Well Being subscale of CAD-EOLD. Four items were included in both SM-EOLD and CAD- EOLD subscales: pain, shortness of breath, fear and anxiety. These item pairs were moderately correlated (r.475 -.559) as anticipated because SM-EOLD reflected conditions during the last 90 days and CAD- Somewhat Corrected item-total correlation Subscales** Not Total at all scale 1 2 3 4 Discomfort 25.7 49.3 25.0.57.63 Pain 21.1 43 35.9.55.61 Restlessness 23.2 46.4 30.4.55.51 Shortness of breath 34.1 32.6 33.3.40.39.35 Choking 17.2 25.4 57.4.47.59 Gurgling 17.9 30.1 52.0.45.58 Difficulty swallowing 44.7 29.5 25.8.33.44 Fear 14.2 32.5 53.3.66.63 Anxiety 16.5 42.1 41.3.76.79 Crying 5.7 23.8 70.5.50.57 Moaning 9.8 32.0 58.2.57.62 Serenity* 20.2 55.3 24.6.26.60 Peace* 27.8 57.4 14.8.53.68 Calm* 31.4 50.0 18.6.47.68 *Reverse coded for calculation of the total score, **Subscale 1 (Factor 1) Physical distress, Subscale 2 (Factor 4) Dying symptoms, Subscale 3 (Factor 2) Emotional distress, Subscale 4 (Factor 3) Well being.

SCALES FOR EVALUATION OF END-OF-LIFE CARE IN DEMENTIA 199 TABLE 5. Rotated factor structure of comfort assessment in dying with dementia (CAD-EOLD) scale Factor loading Factor 1 Factor 2 Factor 3 Factor 4 Discomfort.85.15.13.08 Pain.83.27.08.02 Restlessness.60.25.16.21 Shortness of breath.49.19.21.50 Choking.10.31.03.73 Gurgling.11.17.07.75 Difficulty swallowing.03.11.03.76 Fear.37.58.26.24 Anxiety.57.65.20.17 Crying.07.80.11.14 Moaning.13.74.23.18 Serenity.05.08.83.04 Peace.24.27.72.06 Calm.15.21.79.09 Eigenvalue 5.14 1.94.155 1.06 Percent variance 34.3 12.9 10.4 7.1 Loadings used to identify factors are underlined. EOLD reflected conditions during the dying stage. We could not compare the means of responses with these items because the items were responded to differently in each scale. TABLE 6. Factors of the comfort assessment in dying with dementia (CAD-EOLD) scale Subscale Mean Standard deviation Range Physical distress 8.3 2.2 4 12 Dying symptoms 8.7 2.3 4 12 Emotional distress 9.9 2.2 4 12 Well being 6.2 1.7 3 9 DISCUSSION Evaluation of end-of-life care in dementia is complicated by the inability of persons with dementia to communicate their needs and levels of comfort or distress verbally. Therefore, evaluation of the care of these patients has to rely on non-verbal communication interpreted by a caregiver and on the evaluation of the patient s symptoms by a care provider. In addition, behavioral symptoms of dementia complicate the condition and management of patients. This project developed three scales that can be used to measure quality of end-of-life care for persons with dementia. Two of these scales are focused on patient s symptoms that occurred during the last 90 days of illness or during the dying process. Factor analysis of the data collected by this study indicates that four different factors can be differentiated in symptom management during the dying process. The third measure evaluated satisfaction of the family members with the process of care regarding not only the patient, but also the family caregiver. A review of existing instruments designed to measure satisfaction with care came to the conclusion that none of these instruments can be recommended for evaluation of end-of-life care (Teno, 1999). It remains to be determined if this scale can also be used for evaluation of current care and if the responses are biased by caregiver burden or availability of services. All three scales developed in this study had good internal consistency and reliability. Content validity had been built in a priori as most of the items were taken from previously developed assessments or recommended by a panel of experts. The validity of these scales was also supported by the good correlations between the three scales because it could be expected that patients receiving good care would have better scores on all three scales. However, the data were collected retrospectively and the answers may have been influenced by selective recall. Future studies should be performed to establish concurrent and predictive validity of these scales. Such studies could include correlation of SM-EOLD and CAD-EOLD scores with degree of discomfort measured by direct observation (Hurley, Volicer, et al., 1992) and measurement of relationship between SWC-EOLD and SM-EOLD scores and time to institutionalization of individuals with dementia being cared for at home. Generalizability of our results is also limited by the composition of our subject population that was mainly white and well educated because the scales may have different properties in other racially or ethnically diverse populations. The reproducibility of information collected by these scales will also have to be determined by further testing. It remains to be established if SWC-EOLD and SM-EOLD could be used to evaluate current care provided for patients with advanced dementia, and if CAD- EOLD scale can be used during the dying process of individuals suffering from diseases other than dementia who have lost the ability to communicate because of drug therapy or coma. A significant number of respondents did not completely answer SM-EOLD and CAD-EOLD scales. This lack of responses could not be explained by location of care, because the response rate was similar if the care recipients were cared for at home or at an institution. In some cases, there were only few missing responses that potentially could have been clarified by a telephone follow-up had this not been an anonymous survey. Administration of these scales by an interviewer could also

200 L. VOLICER ET AL. TABLE 7. Reliability and correlations of the scales #of Correlations # Scale Subscale items Alpha 1. 2. 3. 4. 6. 7. 8. 1. SWC-EOLD Total 10.90 2. SM-EOLD Total 10.78.28* 3. Physical 3.47.26* N.A. 4. Psychol. 6.81.20* N.A..35* 5. CAD-EOLD Total 14.86.30*.66*.51*.58* 6. Physical 4**.74.38*.61*.59*.47* 7. Dying Sx 4**.71.28*.28*.29*.18.52* 8. Emotional 4.82.13.50*.29*.46*.53*.39* 9. Well being 3.80.25*.43*.29*.45*.33*.17.49* *p < 0.05, **One item is in both subscales, N.S. not applicable. improve the number of responses, but may change the scale characteristics. Scales measuring satisfaction with care are often limited by a predominance of positive responses resulting in minimal variance and a very skewed distribution. This limits the ability of these scales to detect care improvements. In contrast to that, although the mean SWC- EOLD was higher than the mean of the scale range and for all but two items 75% or more of the respondents gave answers indicating satisfaction with the aspect of care under consideration, the distribution of scores was not significantly different from normal distribution, indicating that care modifications that would improve endof-life care for demented individuals could be readily evaluated. Similarly, distributions of scores of the other two scales were also not significantly different from normal distribution. The normal distribution of scores means that powerful parametric statistical tests requiring normally distributed outcome variables could be used in research investigating end-of-life care in dementia. We hope that availability of outcome measures for endof-life dementia care will stimulate further research in this area. REFERENCES Covinsky KE, Goldman L, Cook EF, et al. The impact of serious illness on patient s families. JAMA 1994;272:1839 44. Hurley AC, Johnson D, MacDonald S, Sgro S. Scale to assess satisfaction with care for late stage Alzheimer patients. Gerontologist 1996;36, Special issue 1:367-368.[Abstract] Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaffe MW. Studies of illness in the aged. The index of ADL: a standardized measure of biological and psychosocial function. JAMA 1963;185:914 9. Lynn J. Measuring quality of care at the end of life: A statement of principles. JAGS 1997;45:526 7. Skolnick AA. MediCaring project to demonstrate, evaluate innovative end-of-life program for chronically ill. JAMA 1998;279:1511 2. Smith SJ. Providing palliative care for the terminal Alzheimer patient. In: Volicer L, Hurley A, eds. Hospice care for patients with advanced progressive dementia. New York: Springer Publishing Company; 1998:247 56. Teno, J. M. Toolkit: Satisfaction. www.chcr.brown.edu/pcoc/satisfac.htm. 1999. Teno JM, Landrum K, Lynn J. Defining and measuring outcomes in end-stage dementia. Alz Dis Assoc Disord 1997;11:25 9. The SUPPORT Principal investigators. A controlled trial to improve care for seriously ill hospitalized patients: The study to understand prognoses and preferences for outcomes and risks of treatment (SUPPORT). JAMA 1995;274:1591 8. Volicer L, Hurley AC, Lathi DC, Kowall NW. Measurement of severity in advanced Alzheimer s disease. J Gerontol 1994;49:M223 M226.