Vol. 42 No. 3 September 2011 Journal of Pain and Symptom Management 419

Transcription

1 Vol. 42 No. 3 September 2011 Journal of Pain and Symptom Management 419 Original Article Evaluation of the Palliative Prognostic Score (PaP) and Routinely Collected Clinical Data in Prognostication of Survival for Patients Referred to a Palliative Care Consultation Service in an Acute Care Hospital Yoko Tarumi, MD, Sharon M. Watanabe, MD, Francis Lau, PhD, Ju Yang, BSc, Hue Quan, MSc, Lorelei Sawchuk, NP, BScN, MN, Donna demoissac, NP, BScN, MN, Gary Wolch, MD, and Vincent Thai, MD Division of Palliative Care Medicine (Y.T., S.M.W., G.W., V.T.), Department of Oncology, University of Alberta, Edmonton, Alberta; School of Health Information Science (F.L., J.Y.), University of Victoria, Victoria, British Columbia; Regional Palliative Care Program (H.Q.), Edmonton, Alberta; and Palliative Care Program (L.S., D.deM.), Royal Alexandra Hospital, Edmonton, Alberta, Canada Abstract Context. Patients, caregivers, and clinicians require high levels of information regarding prognosis when conditions are incurable. Objectives. 1) To validate the Palliative Prognostic Score (PaP) and 2) to evaluate prognostic capacity of used clinical tools and the diagnosis of delirium, in a population referred to a palliative care consultation service at a Canadian acute care hospital. Methods. This was a prospective observational cohort study on survival prediction based on the PaP and routinely collected clinical data, including the Palliative Performance Scale (PPS) and the Folstein Mini-Mental State Examination (MMSE). Kaplan-Meier survival curves, log-rank tests for significant differences between survival curves, and the Cox proportional hazards model were used to identify the relationship between the hazard ratio for death and the above variables. Results. Nine hundred fifty-eight cases underwent final analysis, of which 181 (19%) had a noncancer diagnosis. Median and mean survival were 35 and 131 days, respectively. The three groups, divided based on different ranges of PaP, had significantly different survival curves, with 30-day-survival rates of 78%, 55%, and 11%. Age, PPS, and PaP remained significantly associated with survival, whereas diagnosis group, MMSE, and delirium became insignificant, despite lower hazard of death for cancer vs. noncancer and higher hazard for abnormal vs. normal MMSE and presence vs. absence of delirium. Address correspondence to: Yoko Tarumi, MD, Palliative Care Program, Royal Alexandra Hospital, Room 3228, Kingsway, Edmonton, Alberta, Ó 2011 U.S. Cancer Pain Relief Committee Published by Elsevier Inc. All rights reserved. Canada T5H 3V9. Yoko.Tarumi@ albertahealthservices.ca Accepted for publication: December 21, /$ - see front matter doi: /j.jpainsymman

2 420 Tarumi et al. Vol. 42 No. 3 September 2011 Conclusion. The PaP was successfully validated in a population with characteristics that extend beyond those of the population in which it was originally developed. This is the largest sample in which the PaP has been validated to date. J Pain Symptom Manage 2011;42:419e431. Ó 2011 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved. Key Words Palliative care, end of life, prognostication, survival prediction, survival estimates Introduction Clinical experience and published data support the conclusion that patients and their caregivers from various geographical and cultural settings have high levels of information needs regarding their illness, the disease trajectory, and the dying process. 1 However, previous research has identified deficiencies in doctorpatient communication regarding prognosis and end-of-life issues for patients with advanced, progressive, life-limiting illnesses. 2,3 Multiple factors influence these deficiencies. Recurring themes in the literature include physicians inaccuracy in prediction of survival, lack of training in this topic during medical school, and lack of reliable prognostication tools that are validated in the population. 4e7 To address this shortcoming, the Research Network of the European Association for Palliative Care (EAPC) established a working group charged with the task of providing evidence-based clinical recommendations concerning prognostication in patients with advanced cancer. 8 One of the six key recommendations of the working group was the systematic use by health care providers of prognostic scores designed to divide patients into groups with significantly different survival times. They also recommended that clinical prediction of survival (CPS) be used together with prognostic factors. The two prognostic scores or tools specifically considered in the report of the working group were the Palliative Prognostic Score (PaP) and the Palliative Prognostic Index (PPI). 8 The PaP was originally developed based on an Italian multicenter study conducted in 519 patients with advanced solid tumors (patients with myeloma, renal tumors, and hematological neoplasms were excluded), who were no longer considered suitable for primary treatment and were referred for hospice care. The median survival for this patient group was 32 days. The PaP was formulated by a multiple regression model based on independent predictors that were identified after univariate analysis of 36 clinical factors, including performance status and CPS, and 19 laboratory parameters. The final model contained two clinical symptoms (anorexia and dyspnea), Karnofsky Performance Status (KPS), CPS, total white blood cell (WBC) count, and lymphocyte percentage. Three different risk groups were identified with a probability of 30-day survival of >70% (Group A), 30%e70% (Group B), and <30% (Group C), with median survival of 64 days, 32 days, and 11 days, respectively. 9,10 The same group investigated the ability of the PaP to provide prognostic accuracy in 173 patients with metastatic gastrointestinal and non-small cell lung cancer who were candidates for palliative chemotherapy. The overall median survival was 26 weeks, with only two distinct groups in median survival of 32 weeks (Group A) and eight weeks (Group B). Although the PaP could not be validated in this population, it was able to distinguish a group of patients that may benefit from chemotherapy from a group that may not. 11 The PaP was further validated in 100 patients who were referred to a palliative care consultation service in an acute care hospital in Australia and who were deemed to be terminally ill; the sample included nine (9%) noncancer patients. Median survivals for the three groups were, respectively, 60 days (95% confidence interval [CI] 41, 89 days), 34 days (25, 40), and eight days (2, 11). 12 The same group reported on the use of the PaP in 65 patients with diagnoses of nonmalignant disease. The median survival for the three groups were 266 days, 18.5 days, and five days, respectively. 13

3 Vol. 42 No. 3 September 2011 Evaluation of the Palliative Prognostic Score (PaP) 421 As different populations may have different trajectories of illness, it is important to examine the performance of prognostic scoring systems in the local population. The aim of this study was to examine the performance of the PaP in patients receiving palliative care consultation at a large acute care hospital in Edmonton, Canada. Simultaneously, we examined the applicability of routinely used clinical assessment tools in survival prognostication, including the Palliative Performance Scale (PPS) and the Folstein Mini-Mental State Examination (MMSE). 14 Finally, we aimed to provide insight into the clinical assessment of dyspnea and anorexia as part of prognostic factors within the PaP. Methods Design and Sample This prospective cohort study was conducted in a single acute care hospital from October 2006 through March The Royal Alexandra Hospital is one of the largest acute care hospitals in the city of Edmonton, Canada, with approximately 700 beds. The palliative care consultation service provides symptom management, support for establishing goals of care, and assistance with discharge coordination. Patients are referred by attending physicians, 65% of whom are internal medicine specialists. The service receives over 600 new referrals annually. The program is a part of the Regional Palliative Care Program, 15 which provides clinical care for approximately 80% of the population who die with a diagnosis of cancer annually in the Edmonton zone. 16 Although most patients who use this service have a diagnosis of cancer, the need for a palliative care approach in noncancer patients has been increasing lately. For example, the percentage of consultations provided to noncancer patients at the Royal Alexandra Hospital increased from 11% in 2006 to 24% in Approximately one-third of patients who received palliative care consultation died in the hospital, another third was discharged home, and the remaining third was transferred to inpatient hospices in the city. During the study period, all patients referred to the palliative care consultation service at the Royal Alexandra Hospital were considered to be eligible, except those who required analgesic management advice for non-life-threatening disease or who were not admitted to the hospital but seen in the emergency room or outpatient clinic. Patients also were excluded if there was no bloodwork within seven days before the initial assessment to enable completion of the PaP or if the data were not properly recorded at the time of initial consultation. Most PaPs were completed by a single physician consultant at the time of initial consultation and a smaller number by two other physician consultants and by a nurse practitioner. All physician consultants had a minimum of five years of clinical experience after completion of postgraduate training in palliative medicine. Residents and fellows who participated in patient assessment were closely supervised by the physician consultants. Patients were assessed within a median of five days of admission and were followed daily until discharged from hospital or symptoms were well controlled. The median follow-up time was 11 days. Routinely recorded clinical and administrative data included the following: age, gender, diagnosis, presence or absence of delirium (based on the criteria of the Diagnostic and Statistical Manual of Mental Disorders, 4th edition) at the time of referral, the PPS, 17 the MMSE 14 adjusted for age and education level, 18,19 and the Edmonton Symptom Assessment System (ESAS). The ESAS includes pain, tiredness, nausea, anxiety, depression, drowsiness, appetite, well-being, shortness of breath, and other symptoms, rated on a 0e10 scale; it may be completed by patients or caregivers who are sufficiently familiar with the patients if patients are unable to complete it by themselves. The following variables were recorded besides routinely documented clinical data along with partial score and total PaP: 8 1) presence (partial score 1) or absence (partial score 0) of dyspnea; 2) presence (partial score 1.5) or absence (partial score 0) of anorexia, which is part of routine clinical assessment; 3) KPS (partial score 0 for $30%, partial score 2.5 for 10%e20%); 4) clinician s estimate of survival (partial score 0 for >12 weeks, partial score 2 for >11e12 weeks, partial score 2.5 for >7e10 weeks, partial score 4.5 for 5e6 weeks, partial score 6 for 3e4 weeks, and partial score 8.5 for 1e2 weeks); 5) WBC count

4 422 Tarumi et al. Vol. 42 No. 3 September 2011 (10 to 9/1) (partial score 0 for #8.5, 0.5 for 8.6e11.0, and 1.5 for >11); and 6) lymphocyte percentage of total WBC count (partial score 0 for 20%e40%, 0.5 for 12%e19.9%, 2.5 for <12%). Laboratory values were used if they were taken within seven days before the initial consultation; no bloodwork was taken specifically for this study (Appendix). The process for assessing the presence or absence of dyspnea and anorexia, and determining KPS, is not defined in the PaP. We considered dyspnea as present if reported at rest or reported to limit the patient from carrying out normal activities in the past 24 hours. Anorexia was determined based on routine clinical assessment of anorexia-cachexia syndrome, which includes patients rating on the appetite item of the ESAS; oral intake based on history from patients and caregivers in addition to nursing records; weight history when available; and change in habit for food intake, including change in taste, volume, frequency, and texture of food. We established KPS values based on our usual clinical practice, in which the consultant develops an overall impression of functional status after review of the history, physical examination, investigations, interprofessional team charting, and informal team conference. For patients who appear to have cognitive impairment, we routinely obtain collateral history from caregivers, home care nursing staff, and family doctors who provided care while patients were at home. The study was granted approval by the Health Research Ethics Board, University of Alberta, Canada (Protocol # ). The cohort was tracked for survival status until January 16, Data Analysis Survival analysis was performed with the same variables used in the study by Lau et al., 20,21 which were the following: age, gender, diagnosis, initial PPS, and survival curve time in days. We further added unique variables: initial PaP, MMSE score, and presence/ absence of delirium on initial consultation. For age, we used the categories of <45, 45e64, 65e74, 75e84, and 85þ years. For diagnosis, we grouped the cases as cancer and noncancer, with cancer cases further grouped by tumor types based on the 2007 Alberta Cancer Registry by Alberta Health Services. 22 Survival time was defined as the difference between the death date and assessment date on which the initial PPS was obtained. For patients whose death date was unknown, survival time was censored at the last known consult date. 23,24 The censoring indicator is used in all analyses of survival times. All data were entered and stored as part of a routine administrative database except initial PaP and KPS, which were recoded by the palliative care consult team. The analysis consisted of frequency distributions of the variables in the data set; survival time in days by age, gender, diagnosis, MMSE, delirium, PPS, and PaP and Kaplan- Meier (KM) survival curves by age, gender, diagnosis, MMSE, delirium, and PaP. A multivariate Cox proportional hazards model was fitted using all the covariates, yielding hazard ratios by age group, gender, diagnosis, and PaP and survival rates in days by PaP. The proportionality assumption was tested using cox.zph in R, with a critical value of 0.05, adjusted for multiple comparisons. 25 Log-rank tests were computed for each adjacent pair of survival curves to examine if they were statistically different from each other. The analyses were conducted in the same manner as in the study by Lau et al. and with the use of survival table by days instead of mortality rates over time. All data analyses were performed using SPSS version 15 (SPSS, Inc., Chicago, IL) and R version ( Analysis of the ESAS data will be reported in a separate article. Results Patient Characteristics A total of 1459 patients were referred to the palliative care consultation service during the study. Consultation was performed within 24 hours of request and 72 hours for nonurgent consults during weekends or holidays. Of total referrals, 50 cases required chronic nonmalignant pain management without end-of-life issues, 10 cases were seen only in the emergency room or outpatient clinic and returned home on the same day, 58 cases did not have adequate laboratory data available within seven days before the consultation, and 176 cases did not

5 Vol. 42 No. 3 September 2011 Evaluation of the Palliative Prognostic Score (PaP) 423 have the PaP completed in a timely manner because of the absence of the main investigator from clinical service. Of the remaining 1165 cases, 207 cases had more than one admission occurring during the study period. After exclusion as above, a total of 958 cases underwent final analysis. Of 958 cases, 181 (19%) cases had a noncancer diagnosis. The characteristics of this cohort at the time of the first assessment by the palliative care consultation team during the study period are shown in Table 1. Overall Survival Patterns The mean, median, and range of survival of the cohort by age, gender, diagnosis group (cancer or noncancer), initial PaP score, normal or abnormal MMSE, and presence or absence of delirium are shown in Table 2. Overall survival, accounting for the 53 censored patients, was a median of 35 days (95% CI 31, 39) and a mean of 131 days (95% CI 115, 147), with a range of 1e1107 days. Diagnosis group, initial PaP, MMSE, and delirium had significant effects on overall survival (log rank P < for each variable). Age also was significant but to a lesser extent (log rank P ¼ 0.001). Gender was insignificant (log rank P ¼ 0.555). The KM survival curves stratified by initial PaP groups are shown in Fig. 1. The log-rank test for the equality of survival curves showed highly significant differences among the curves over the three PaP groups. Furthermore, pair-wise log-rank tests for adjacent pairs of PaP showed that patients in PaP Group A vs. Group B, and Group B vs. Group C, had significantly different survival curves (log-rank test P < 0.001). Survival by PaP and Covariates The Cox proportional hazards model was used to examine the relationship between the hazard ratio of death, adjusted for age, gender, diagnosis, MMSE, delirium, PPS, and PaP. Table 3 shows hazard ratios from the Cox proportional hazards model, which was used to examine the relationship between the hazard ratios of death, adjusted for age, gender, diagnosis group, PPS, PaP, MMSE, and delirium. In patients younger than 45 years compared with those older than 85 years, PPS and PaP remained significant in hazard of death, whereas diagnosis group, MMSE, and delirium became insignificant, despite lower Table 1 Patient Characteristics Variable Frequency Percent Number of patients considered 958 for final analysis Gender Male Female Number of patients per age group (from first consult date) < e e e þ Mean age in years 71 (SE 0.44) 18 w 102 Median age in years 73 Number of patients with diagnosis group Cancer Noncancer (NC) Number of patients with diagnosis groups Bones and connective tissues Breast Endocrine glands Eye, brain, and central nervous system Female genital Gastrointestinal Head and neck Hematopoietic Male genital Melanoma Nonmelanoma of the skin NC cardiopulmonary NC infectious disease NC neuromuscular NC other Respiratory Retroperitomeum Unknown primary Urinary tract Delirium No Yes MMSE Abnormal Normal Missing Status Dead Alive hazard of death for cancer vs. noncancer and higher hazard for abnormal vs. normal MMSE. Tables 4 and 5 and Figs. 2 and 3 show the impact of presence of delirium and MMSE on survival time in the three PaP groups. The presence of delirium did not significantly influence survival time in the three PaP

6 424 Tarumi et al. Vol. 42 No. 3 September 2011 Table 2 Survival Times by Age, Gender, Diagnosis, PaP, and PPS Variable Mean (95% CI) Survival Time (in Days) Median (95% CI) Range Number of Patients Percent Log Rank P-value Overall 131 (115, 147) 35 (31, 39) <1e Age group < (79, 355) 61 (22, 100) 1e e (123, 190) 45 (37, 53) <1e e (103, 171) 36 (26, 46) <1e e (89, 144) 33 (25, 41) <1e þ 75 (53, 99) 23 (12, 34) <1e Gender Male 126 (104, 149) 33 (27, 39) <1e Female 135 (111, 158) 38 (30, 46) <1e Diagnosis group <0.001 Cancer 133 (116, 150) 43 (38, 48) <1e Noncancer 120 (77, 163) 7 (5, 9) <1e Initial PaP <0.001 Group A 221 (190, 251) 89 (72, 106) <1e Group B 84 (64, 105) 35 (30, 40) 1e Group C 31 (13, 50) 4 (3, 5) <1e Initial PPS <0.001 PPS 10% 17 (0, 36) 2 (1, 3) <1e PPS 20% 21 (5, 37) 5 (3, 7) <1e PPS 30% 108 (69, 148) 20 (13, 27) <1e PPS 40% 108 (74, 142) 31 (21, 41) <1e PPS 50% 128 (97, 159) 54 (37, 71) <1e PPS 60% 193 (150, 236) 72 (54, 90) 1e PPS 70% 252 (176, 328) 116 (56, 176) 11e PPS 80% 238 (125, 352) 121 (76, 166) 8e PPS 90% 369 (192, 545) 266 (32, 500) 45e Initial MMSE <0.001 Abnormal 89 (69, 109) 14 (11, 17) <1e Normal 171 (145, 197) 61 (52, 70) 1e Delirium <0.001 No 142 (124, 162) 43 (38, 48) <1e Yes 92 (60, 123) 11 (7, 15) <1e groups; however, abnormal MMSE significantly influenced shorter survival overall, although there was a tendency to influence to a higher degree in Group A, which had the highest 30-day survival compared with Groups B and C. A similar tendency was found for delirium, for which Group A was influenced to a higher degree than Groups B and C. Survival Rate Over Time We constructed a life expectancy table from the KM survival curves to show the survival rate for each PaP category (Table 6). From this table, 30-day survival in Groups A, B, and C was 78%, 55%, and 11%, respectively. The survival times for 50% survival rate in the three groups were 90 days, 30 days, and three days, respectively. Discussion In this study, we have performed an external validation of the PaP as a prognostication tool and demonstrated satisfactory performance for patients who were referred to an inpatient palliative care consultation service in an acute care hospital in Edmonton, Canada. The PaP successfully divided this population of patients with life-threatening illness into three groups, each with a significantly different survival profile on the basis of one-month survival probability. In addition, we have shown the validity of the PaP in a population whose characteristics extend beyond those of the population in which it was originally developed. 9 Furthermore, this is the largest sample in which the PaP has been validated to date.

7 Vol. 42 No. 3 September 2011 Evaluation of the Palliative Prognostic Score (PaP) 425 Fig. 1. KM survival curves by initial PaP. PPS and Survival Prediction The PPS also has shown prognostic significance in our population through this study. The PPS is a derivative of the KPS. In the PPS, hospitalization is removed as a criterion for a lower score to be consistent with current health care because indications for hospitalization have changed since the KPS was originally developed. The PPS incorporates five observer-rated parameters: ambulation, activity, self-care, intake, and level of consciousness. The scale grades a patient s general condition from 0% (dead) to 100% (normal), in increments of 10 percentage points. 17 The PPS is a well-validated tool in the Table 3 Hazard Ratios for PaP, PPS, MMSE, Delirium, Age, Gender, and Diagnosis 95% CI for Hazard Ratio Variable Hazard Ratio Lower Upper P-value Age group (vs. 85þ) <0.001 < < e e e Gender (vs. male) Diagnosis group (vs. not cancer) Initial PPS (vs. PPS 10%) <0.001 PPS 20% PPS 30% <0.001 PPS 40% <0.001 PPS 50% <0.001 PPS 60% <0.001 PPS 70% <0.001 PPS 80% <0.001 PPS 90% <0.001 Initial PaP (vs. Group C) <0.001 Group A <0.001 Group B <0.001 MMSE (vs. normal) Delirium (vs. no)

8 426 Tarumi et al. Vol. 42 No. 3 September 2011 Table 4 Survival Times by MMSE Stratified by PaP Groups PaP MMSE Mean (95% CI) Median (95% CI) Number of Patients Percent P A Abnormal 184 (136, 233) 68 (48, 88) Normal 233 (195, 272) 101 (77, 125) Subtotal 217 (186, 247) 84 (66, 102) B Abnormal 79 (48, 110) 28 (22, 34) Normal 81 (61, 101) 39 (31, 47) Subtotal 85 (64, 106) 34 (29, 39) C Abnormal 27 (10, 45) 4 (3, 5) Normal 19 (4, 34) 6 (3, 9) Subtotal 26 (11, 42) 4 (3, 5) Overall Total 129 (112, 145) 35 (31, 39) palliative care population. It has been a part of routinely used clinical assessment tools in the Edmonton Regional Palliative Care Program since The PPS remained significant in the multivariate model although it is part of the PaP, and lower PaP also has shown significant hazard of death as shown in Table 4. This finding is consistent with previous studies that have shown its ability to prognosticate survival: a prospective study of 213 hospice patients, 17 a prospective study in 153 patients admitted to a specialist palliative care unit, 26 a retrospective study of 396 home and hospice patients, 27 a prospective study of 466 home, nursing home, hospital and hospice patients, 28 a retrospective study of 733 hospice patients, 21 a prospective cohort of 513 patients assessed by a combination of hospital-based and community-based palliative care consultation teams, 20 and a prospective study in 157 cancer patients and 104 noncancer patients. 29 Cognitive Factors and Survival Prediction In our study population, both abnormal MMSE and presence of delirium showed distinctly different survival curves, with shorter median survival time. The median survival times for normal and abnormal MMSE groups were 14 days (95% CI 11, 17) and 61 days (95% CI 52, 70), respectively (log rank P-value <0.001) (Table 2). The median survival times for the groups without and with delirium were 43 days (95% CI 38, 48) and 11 days (95% CI 7, 15), respectively (log rank P-value <0.001) (Table 2). We then applied the Cox proportional hazards model to examine the relationship between the hazard ratios of death and all the variables present. In the model, PPS and PaP remained significant in hazard of death, whereas MMSE and delirium became insignificant (Tables 3e5) (Figs. 2 and 3). In a previous validation study of the PaP, a subgroup of the original cohort was screened for the presence or absence of delirium based on the Confusion Assessment Method (CAM) algorithm. 30 This revealed that patients with delirium (109 patients, 27.1% of total screened) had a significantly different survival curve when compared with nondelirious Table 5 Survival Times by Delirium Stratified by PaP Groups PaP Group Delirium Mean (95% CI) Median (95% CI) Number of Patients Percent P A No 225 (193, 257) 97 (76, 118) Yes 189 (99, 278) 61 (36, 86) Subtotal 221 (190, 251) 89 (72, 106) B No 80 (59, 102) 35 (29, 41) Yes 97 (47, 147) 31 (20, 42) Subtotal 84 (64, 105) 35 (30, 40) C No 27 (7, 47) 5 (4, 6) Yes 35 (4, 67) 3 (2, 4) Subtotal 31 (13, 50) 4 (3, 5) Overall Total 131 (115, 147) 35 (31, 39)

9 Vol. 42 No. 3 September 2011 Evaluation of the Palliative Prognostic Score (PaP) 427 Fig. 2. KM survival curves by initial MMSE. patients (log rank, 31.6; P < ), with significantly shorter median survival. In further analysis with a multivariate model, including the PaP and the CAM algorithm outcome, the presence of delirium retained independent association with survival. 31 The PPI was devised and validated in cancer patients in a hospice inpatient unit in Japan. 32,33 It was further validated in a different population receiving palliative chemotherapy and radiotherapy in Ireland. 34 It is one of the two prognostic tools specifically considered in the report of the EAPC working group, with the PaP being the other. 8 The PPI was considered more reliable than the PaP because of the inclusion of delirium and the exclusion of CPS and laboratory testing. 34 The effect of cognitive impairment on survival of older patients with breast, colon, and prostate cancer was further studied. 35 The difference in impact of delirium on survival between our study and previous studies may be possibly explained by the different sample size and population. Delirium was detected in 321 patients (33.5%) in our study. Also, abnormal MMSE and development of Fig. 3. KM survival curves by delirium by PaP groups.

10 428 Tarumi et al. Vol. 42 No. 3 September 2011 Table 6 Survival Rate (%) by PaP in Days Days Total Cases PaP A 99% 98% 97% 96% 89% 78% 68% 60% 50% 30% 18% 418 B 98% 93% 88% 83% 73% 55% 39% 28% 19% 10% 5% 303 C 74% 54% 43% 33% 15% 11% 6% 6% 4% 3% 2% 237 delirium are often observed as the disease progresses toward the end of the illness trajectory. This may explain why neither abnormal MMSE nor delirium had an independent influence on shorter survival in the three groups, especially in Groups B and C, whereas it is possible that either or both delirium and abnormal MMSE may have shown a significant hazard ratio of death in Group A, which consists of patients who are expected to survive longer. The number of patients with abnormal MMSE was found to be greater than that with delirium in our study (Table 1). This was partly because of the inconsistent recording of the presence of delirium in our database, as delirium was identified more frequently in the descriptive initial consultation reports. At the same time, this also reflected patients with moderate-to-advanced dementia and unresponsiveness because of acute or subacute brain insults, such as hemorrhagic or embolic stroke and postresuscitative anoxic brain damage. Enigma in Application of the PaP in Clinical Practice There has been a lack of a standardized assessment system for KPS, dyspnea, and anorexia when applying the PaP. The only study to have described the process of assessing these elements is by Glare and Virik, 12 in which it is reported that a research assistant assessed dyspnea by asking In the past 24 hours, have you been short of breath?, anorexia by asking In the past 24 hours, has your appetite been decreased?, and KPS was assessed by asking If you were at home today, what sort of activities would you be able to do? To be able to adequately generalize applicability of the PaP, how the assessment should be performed needs to be clearly defined. It is always challenging to assess subjective symptoms in patients who are unable to report them because of cognitive impairment, expressive disability, or choosing not to express their symptoms because of a variety of psychosocial conditions. Dyspnea is a manifestation of a symptom secondary to multifactorial pathophysiologies, including cardiopulmonary condition, neuromuscular condition, anxiety, anorexia-cachexia syndrome, metabolic acidosis, anemia, or extrathoracic mechanical pressure, such as ascites or organomegaly. 36 Yet, the definition of dyspnea has been reliant on the patient s subjective experience of breathing discomfort that consists of qualitatively distinct sensations that vary in intensity. 36 Anorexia-cachexia syndrome is known to be one of the most frequent and devastating problems affecting patients with advanced cancer. It often manifests with loss of appetite, weight loss of more than 2% in two months or more than 5% in six months, and fatigue, often combined with chronic nausea, early satiety, and taste problems. Although some patients may not have weight loss, third space fluid accumulation, such as edema, ascites, pulmonary effusion, or increased fat tissue while losing muscle mass (sarcopenia) rapidly may occur instead. Anorexia-cachexia syndrome occurs in numerous chronic end-stage disease processes, such as cancer, acquired immunodeficiency syndrome, chronic pulmonary disease, chronic renal insufficiency, heart failure, and even in dementia. 37 However, assessment and management of anorexia has been somewhat limited partly because of the lack of clear diagnostic process/definition, clinically or physiologically. When prognosticating survival time, it is unknown how much the subjective experience and expression of dyspnea or anorexia play a role in survival, relative to the physiological parameters, such as hypoxia, decreased cardiac output, or loss of lean body mass and the reversibility of those physiological conditions. It also is possible that severe symptoms may overlap with lower performance status and shorter CPS, such that severity of symptoms

11 Vol. 42 No. 3 September 2011 Evaluation of the Palliative Prognostic Score (PaP) 429 becomes a less significant contributor to the group of patients with shorter survival time, especially the PaP Group C. Limitations The lack of consistency in the referral process for palliative care consultation and the relatively loosely defined criteria for consideration of referral may have limited the interpretation of the results. This is particularly reflected in the significantly shorter survival of the noncancer group and possibly the older age group, for which the attending teams may have waited to involve the palliative care team until the patient s condition had clearly deteriorated. In contrast, patients with cancer were often referred to the palliative care program for symptom management even before a definitive pathology-based diagnosis had been made. In our current clinical practice, over 60% of patients referred to the palliative care consultation service had no diagnosis of a palliative condition on admission to the hospital. As the prognosticating investigators were partly responsible for clinical care in coordination with the attending teams, it is possible that they influenced decision making with respect to life-sustaining treatments, such as artificial nutrition, fluid administration, transfusions, or antibiotic usage and hence influencing survival. However, the decision-making processes for end-of-life care were initiated by the attending teams in most cases, rendering it unlikely that the palliative care consultant had changed the treatment strategy. Further limitations include the timing of the prognostic assessment in relation to disease trajectory, variations in the clinical experience of the health care professionals making the prediction, and access to other clinical information when establishing prognosis. 7 Limitation in Generalizability The PaP provides only a probability of onemonth survival and the median survival range based on the population in which the PaP data were accumulated. Information, such as median survival in the Survival Times (Table 2) and probabilities of survival at different times before death in the Survival Rate by PaP (Table 6) also are helpful when assisting patients with their decisions regarding treatment or care planning by providing an actuarial estimation of survival, which uses factors established through data and research to more narrowly define an individual prognosis for patients. 38 In our study, the median survival of the population was 35 days. This is representative of selected hospitalized patients for whom referring physicians thought that palliative care consultation was useful because of the progressive nature of the disease. This particularly limits the applicability of the information to patients who retain a high performance status because they are less likely to be admitted to hospital, and this will likely influence survival information when applying the Survival Rate by PaP (Table 6). Until the PaP is tested in all hospitalized patients or all cancer patients, application of the Survival Rate by PaP requires caution. Clinicians still need to be fully aware of the limitation in using the median survival range to communicate with patients and families. Without consideration of the nature of statistics and the settings in which the statistics were formulated, the patient and family may potentially be misguided. Conclusion In this study, we were able to externally validate the PaP in a different setting from the original study with a relatively larger population. We have shown that both the abnormal MMSE and the presence of delirium showed distinctively different survival curves with shorter median survival times. However, in age younger than 45 years compared with age older than 85 years, PPS and PaP remained significant in hazard of death, whereas MMSE and delirium became insignificant, especially in the population whose expected survival times are shorter, such as the PaP Group C. When carefully applied, the results of this study, such as the Survival Time Table and the Survival Rate Table, may provide useful information regarding prognosis in the setting of end-of-life care. Disclosures and Acknowledgments The authors acknowledge their special appreciation for funding support for this work through the Canadian Institutes for Health Research New Emerging Team Grant in Hospice Palliative and End-of-life Care and the Partnership in Health System Improvement Grant on Timely Access to End-of-life Care. The authors declare no conflicts of interest.

12 430 Tarumi et al. Vol. 42 No. 3 September 2011 References 1. Parker SM, Clayton JM, Hancock K, et al. A systematic review of prognostic/end-of-life communication with adults in the advanced stage of a life-limiting illness: patient/caregiver preferences for the content, style, and timing of information. J Pain Symptom Manage 2007;34:81e Gysels M, Richardson A, Higginson IJ. Communication training for health professionals who care for patients with cancer: a systematic review of effectiveness. Support Care Cancer 2004;12:692e Sullivan AM, Lakoma MD, Matsuyama RK, et al. Diagnosing and discussing imminent death in the hospital: a secondary analysis of physician interviews. J Palliat Med 2007;10:882e Hancock K, Clayton JM, Parker SM, et al. Discrepant perceptions about end-of-life communication: a systematic review. J Pain Symptom Manage 2007;34:190e Vigano A, Dorgan M, Bukingham J, Bruera E, Suarez-Almazor ME. Survival prediction in terminal cancer patients: a systematic review of the medical literature. Palliat Med 2000;14:363e Chow E, Harth T, Hruby G, et al. How accurate are physicians clinical predictions of survival and the available prognostic tools in estimating survival times in terminally ill cancer patients? A systematic review. Clin Oncol 2001;13:209e Glare P, Virik K, Jones M, et al. A systematic review of physicians survival predictions in terminally ill cancer patients. BMJ 2003;327:195e Maltoni M, Caraceni A, Brunelli C, et al. Prognostic factors in advanced cancer patients: evidencebased clinical recommendationsda study by the Steering Committee of the European Association of Palliative Care. J Clin Oncol 2005;23:6240e Maltoni M, Nanni O, Pirovano M, et al. Successful validation of the Palliative Prognostic Score in terminally ill cancer patients. J Pain Symptom Manage 1999;17:240e Pirovano M, Maltoni M, Nanni O, et al. A new Palliative Prognostic Score: a first step for the staging of terminally ill cancer patients. J Pain Symptom Manage 1999;17:231e Tassinari D, Montanari L, Maltoni M, et al. The Palliative Prognostic Score and survival in patients with advanced solid tumors receiving chemotherapy. Support Care Cancer 2008;16:359e Glare P, Virik K. Independent prospective validation of the PaP Score in terminally ill patients referred to a hospital-based palliative medicine consultation service. J Pain Symptom Manage 2001;22:891e Glare P. The use of the Palliative Prognostic Score in patients with diagnoses other than cancer. [letter]. J Pain Symptom Manage 2003;26:883e Folstein MF, Folstein SE, McHugh PR. Minimental state. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975;12:189e Regional Palliative Care Program. Edmonton Zone, Alberta Health Services. Program description. Available from RPCP/RPCPIdx.html. Accessed March 23, Fassbender K, Fainsinger R, Jacobs P, et al. Utilization and costs of the introduction of system-wide palliative care in Alberta, 1993e2000. Palliat Med 2005;19:513e Anderson F, Downing GM, Hill J, Casorso L, Lerch N. Palliative Performances Scale (PPS): a new tool. J Palliat Care 1996;12:5e Grigoletto F, Zappala G, Anderson DW, Lebowitz BD. Norms for the Mini-Mental State Examination in a healthy population. Neurology 1999;53:315e Dufouil C, Clayton D, Brayne C, et al. Population norms for the MMSE in the very old: estimates based on longitudinal data. Mini-Mental State Examination. Neurology 2000;55:1609e Lau F, Maida V, Downing M, et al. Use of Palliative Performance Scale for end-of-life prognostication in a palliative medicine consultation service. J Pain Symptom Manage 2009;37:965e Lau F, Downing M, Lesperance M, Shaw J, Kuziemsky C. Use of Palliative Performance Scale in end of life prognostication. J Palliat Med 2006; 9:1066e Alberta Cancer Registry Annual report of cancer statistics. Available from albertahealthservices.ca/poph/hi-poph-surv-canceralta-cancer-registry-2007.pdf. Accessed March 23, Lawless J. Statistical models and methods for life time data, 2nd ed. New York: Wiley, Kalbfleisch JD, Prentice R. Statistical analysis of failure time data, 2nd ed. New York: Wiley, Grambsch P, Therneau T. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994;81:515e Virik K, Glare P. Validation of the Palliative Performance Scale for inpatients admitted to a palliative care unit in Sydney, Australia. [letter]. J Pain Symptom Manage 2002;23:455e Head B, Ritchie CS, Smoot TM. Prognostication in hospice care: can the Palliative Performance Scale help? J Palliat Med 2005;8:492e Harrold J, Rickerson E, Carroll JT, et al. Is the Palliative Performance Scale a useful predictor of mortality in a heterogeneous hospice population? J Palliat Med 2005;8:503e Olajide O, Hanson L, Usher BM, et al. Validation of the Palliative Performance Scale in the acute tertiary care hospital setting. J Palliat Med 2007;10: 111e117.

13 Vol. 42 No. 3 September 2011 Evaluation of the Palliative Prognostic Score (PaP) Inouye SK, van Dyck CH, Alessi CA, et al. Clarifying confusion: the Confusion Assessment Method. A new method for detection of delirium. Ann Intern Med 1990;113:941e Caraceni A, Nanni O, Maltoni M, et al. Impact of delirium on the short term prognosis of advanced cancer patients. Italian Multicenter Study Group on Palliative Care. Cancer 2000;89:1145e Morita T, Tsunoda J, Inoue S, Chihara S. The Palliative Prognostic Index: a scoring system for survival prediction of terminally ill cancer patients. Support Care Cancer 1999;7:128e Morita T, Tsunoda J, Inoue S, Chihara S. Improved accuracy of physicians survival prediction for terminally ill cancer patients using the Palliative Prognostic Index. Palliat Med 2001;15:419e Stone CA, Tiernan E, Dooley BA. Prospective validation of the Palliative Prognostic Index in patients with cancer. J Pain Symptom Manage 2008;35:617e Raji MA, Kuo Y, Freeman JL, Goodwin JS. Effect of a dementia diagnosis on survival of older patients after a diagnosis of breast, colon, or prostate cancer. Arch Intern Med 2008;168:2033e Anonymous. Dyspnea mechanisms, assessment, and management: a consensus statement. American Thoracic Society. Am J Respir Crit Care Med 1999; 159:321e Morley JE, Thomas DR, Wilson MG. Cachexia: pathophysiology and clinical relevance. Am J Clin Nutr 2006;83:735e Dawes RM, Faust D, Meehl PE. Clinical versus actuarial judgment. Science 1989;243:1668e1674. Appendix Palliative Prognostic Score Prognostic Factor Partial Score Dyspnea Absent 0 Present 1 Anorexia Absent 0 Present 1.5 KPS $ e e CPS >12 weeks 0 11e12 weeks 2.0 9e10 weeks 2.5 7e8 weeks 2.5 5e6 weeks 4.5 3e4 weeks 6.0 1e2 weeks 8.5 Total WBC count (cell/mm 3 ) Normal: 4800e8500 cells/ml 0 High: 8500e11,000 cells/ml 0.5 Very high: >11,000 cells/ml 1.5 Lymphocyte percentage Normal: 20.1e Low: 12.0e Very low: 0e Risk group: 30-day survival probability of A: >70% 0e5.5 B: 30%e70% 5.6e11.0 C: <30% 11.1e17.5