ORIGINAL ARTICLE. Biostatistics, Virginia Commonwealth University, Richmond, VA

Transcription

1 LIVER TRANSPLANTATION 14: , 2008 ORIGINAL ARTICLE The Utility of the Model for End-Stage Liver Disease Score: A Reliable Guide for Liver Transplant Candidacy and, for Select Patients, Simultaneous Hospice Referral Valentina Medici, 1 Lorenzo Rossaro, 1 Jacob A. Wegelin, 2 Amit Kamboj, 1 Junko Nakai, 1 Kelli Fisher, 1 and Frederick J. Meyers 1 1 Department of Internal Medicine, University of California, Davis, Sacramento, CA; and 2 Department of Biostatistics, Virginia Commonwealth University, Richmond, VA Patients with chronic liver disease are referred late to hospice or never referred. There are several barriers to timely referral. First, liver transplantation (LT) and hospice care have always been perceived as mutually exclusive. Yet the criteria for hospice referral and for LT are more similar than different (for example, advanced liver disease and imminent death). Second, physicians, patients, and families have not had a reliable metric to guide referral. However, many patients wait for transplantation but never receive an organ. We hypothesized that the Model for End-Stage Liver Disease (MELD) score already in use to prioritize LT could be used in selected patients for concurrent hospice referral. Furthermore, we hypothesized that patients awaiting LT can receive hospice care and remain eligible for transplantation. Patients with advanced or end-stage liver disease were referred to the University of California Davis Health System hospice program. We correlated the MELD score at admission to length of stay (LOS) in hospice. A total of 157 end-stage liver disease patients were admitted to the hospice service. At the time of hospice admission the mean MELD score was 21 (range, 6-45). The mean length of hospice stay was 38 days (range, days). A significant correlation was observed between hospice LOS and MELD score at hospice admission (P 0.01). Six patients were offered a liver graft while on the combined (LT and hospice) program. MELD can be used to guide clinician recommendation to families about hospice care, achieving one of the national benchmark goals of increasing hospice care duration beyond the current median of 2-3 weeks. A higher MELD score might augment physician judgment as to hospice referral. Hospice care for selected patients may be an effective strategy to improve the care of end-stage liver disease patients waiting for LT. Liver Transpl 14: , AASLD. Received 11 June 2007; accepted 19 October Worldwide, end-stage liver disease (ESLD) is one of the major causes of terminal illness, responsible for 8.8 deaths per 100,000 persons annually in the United States. 1 Patients with ESLD have significant physical and emotional suffering and should be excellent candidates for hospice care, the primary goal of which is the palliation of physical, emotional, and spiritual suffering in terminally ill patients. Hospice care can be delivered in the patient s home, allowing death to take place at home, and it may help with the relief of pain, 2 the patient s and family s satisfaction, 3 and increase cost effectiveness. 4 While these patients should be referred for hospice care at end of life, few are, and many have only a very short length of stay (LOS). Barriers to timely hospice referral in the case of ESLD are similar to those described for most terminally ill patients; however, 2 barriers are unique to ESLD patients. First, no reliable metric has been validated as a tool to guide physician referral and patient acceptance of hospice Abbreviations: ESLD, end-stage liver disease; LOS, length of stay; LT, liver transplantation; MELD, Model for End-Stage Liver Disease; MELD-A, MELD score at hospice entry; UCD, University of California, Davis; UNOS, United Network for Organ Sharing. Address reprint requests to Valentina Medici, M.D., Department of Internal Medicine, Division of Gastroenterology and Hepatology, University of California, Davis, 4150 V Street, Suite 3500, Sacramento, CA Telephone: ; FAX: ; valentina.medici@ucdmc.ucdavis.edu DOI /lt Published online in Wiley InterScience ( American Association for the Study of Liver Diseases.

2 MELD SCORE AND HOSPICE REFERRAL 1101 services. Second, many patients wait for liver transplantation (LT) and consequently delay use of hospice services. LT is currently the elective treatment option for selected patients with ESLD and its complications. Potential transplant recipients often receive disease directed therapy or so-called curative efforts in order to maintain their candidacy for transplantation, without sufficient attention to palliative or end-of-life needs. LT candidacy and enrollment in hospice care have always been perceived as mutually exclusive. 5,6 Only when the patient s medical condition worsens and the hope for LT is lost, that is, in the last weeks or days of life, are patients sometimes offered palliative or hospice care, while others are never offered this option. In this way, patients and family members have little time to benefit from the hospice care programs. 7 The result is often a poor quality of life and poor quality of care at the end of life. 8 The introduction of the Model for End-Stage Liver Disease (MELD) score as a predictor of mortality has improved the ability to predict the prognosis of ESLD and has been adopted by the United Network for Organ Sharing (UNOS) to determine priorities in allocating organs However, families recognize the cognitive dissonance between the message that a high MELD score means a shorter life span and may result in a LT, but, at the same time, hospice referral cannot be considered. Furthermore, clinical ethics should require that both options, hospice care and LT, be discussed with patients, but this rarely occurs. This ethical dilemma should be resolved using new models of care. There is increasing interest in the mutual interaction between palliative care programs or hospice services and hepatology and LT. 12 We saw an opportunity to develop a model that directly approaches this cognitive dissonance between hospice and hepatology. We hypothesized that: (1) the MELD score might be used to guide the referral of patients for hospice services; (2) patients could be maintained on a waiting list for transplantation while receiving hospice services; and (3) the hospice team could be taught the complications of ESLD such as gastrointestinal bleeding, spontaneous bacterial peritonitis, and hepatic encephalopathy that require disease specific management. 13,14 In this clinical protocol, we introduce hospice to patients with advanced liver disease, many of whom were followed by the LT program. In addition, patients with ESLD not eligible for transplantation were also admitted to hospice. In this study we describe the demographic and clinical features of patients with ESLD admitted at our hospice or at our combined hospice/liver service to determine the value of the MELD score in guiding referral for hospice care. In addition, we explore the possibility of a new integrated model between the hospice/ palliative care team and the LT team. PATIENTS AND METHODS Patients admitted with ESLD to the hospice care service of the University of California, Davis (UCD) Health System from October 2001 to December 2005 were reviewed. While this is a retrospective review, in 2000 the hospice Medical Director (F.J.M.) and the Chief of Hepatology (L.R.) discussed the feasibility of using the MELD score to facilitate hospice referral and together introduced the concept to both the LT team and the hospice team, as well as to hospital resident physicians. The UCD hospice is a Medicare-certified hospice program that serves the Sacramento region. The liver disease service and the hospice team agreed to jointly manage patients with ESLD. Patients not eligible for LT were immediately referred to hospice. Patients whose MELD score worsened or who required continuous care and support at home during their waiting period or at the time of listing were also offered hospice care. UNOS guidelines, which factor in the severity of illness and the patient s odds of benefiting from the liver transplant, were utilized for assessment and management. For each patient, the following features were recorded: age, gender, main diagnosis, comorbidities, length of hospice stay, and development of ESLD-specific complications during hospice stay (gastrointestinal bleeding, hepatic encephalopathy, or tense ascites with dyspnea). If a liver graft became available, hospice care was revoked and the patients underwent LT. The MELD score at hospice entry (MELD-A) was computed for all patients for whom sufficient information was available in the medical record. In addition, for listed patients we computed the MELD score at the time of listing and at the time of the eventual LT. Patients were followed during their postoperative course for mortality. Statistical Analysis Quantile regression was performed to assess the degree to which LOS might be related to MELD-A, gender, and age. This method was chosen rather than conventional least-squares regression because it enables the researcher to examine the distribution of LOS in considerable detail, instead of focusing merely on the mean. In the current work the 25th, 50th (median), and 75th percentiles of LOS were investigated. Thus, 3 separate regression models were fit and tests were performed to determine whether any of these percentiles of LOS was related to MELD-A, gender, or age. Since LOS is positive and skewed toward large values, this variable was logtransformed before analysis to yield multiplicative effects on the original scale. The regression formula yields an estimated LOS percentile for a given gender and given values of MELD-A and age. The formula is as follows: log Q 0 1 M 21 2 G 3 A 57, (1) where Q is the percentile being estimated; M is the MELD-A score; G 0 if the patient is male, 1 if female; A is the patient s age; and 0 through 3 are coefficients estimated by the statistical modeling software. For re-

3 1102 MEDICI ET AL. TABLE 1. Etiologies and Their Frequency of End-Stage Liver Disease Admitted to Hospice Service Etiology Number (%) Virus (HCV, HBV) 25 (15.9) ETOH 27 (17.2) Virus and ETOH 42 (26.7) End stage liver disease and HCC 44 (28%) Other (hemochromatosis, autoimmune hepatitis, NASH) 12 (7.6) Unknown 9 (5.7) Total 157 Abbreviations: ETOH: alcohol abuse; HBV, hepatitis B virus; HCC: hepatocellular carcinoma; HCV, hepatitis C virus; NASH: nonalcoholic steatohepatitis. porting, the coefficients were subsequently transformed to the original scale by the exponential function and 95% confidence intervals reported on this scale. In addition, models were fit for the quartiles without the age and gender effects, based on the following simpler formula: log Q 0 1 M 21 (2) where the coefficients 0 and 1 define the relationship between MELD-A and the percentile, averaged over all ages and both genders. This model was depicted graphically. To assess the relationship between MELD score at admission to hospice and whether or not a patient had liver cancer, a Mann-Whitney rank-sum test was performed. To assess the average level and progression of the MELD score in the patients who received liver transplants, a linear mixed-effects model was computed. In addition, the individual trajectories of MELD for these patients were depicted graphically. RESULTS Demographics of the Study Participants Patients with ESLD (n 157) were admitted at the hospice service of the UCD Health System, of which 106 were male (67.5%). The mean age was years (range, years). The most common causes of ESLD were either alcohol cirrhosis, hepatitis C virus infection, or both (Table 1). In 44 cases (28%) chronic liver disease was complicated by hepatocellular carcinoma. Common comorbidities were diabetes (18.2%), vasculopathy and hypertension (16.3%), kidney failure (11.3%), and chronic obstructive pulmonary disease (4.4%). A total of 122 (78%) patients died during the observation period. During hospice care, 3.1% of patients experienced gastrointestinal bleeding, 30% developed hepatic encephalopathy, and 22% developed tense ascites with dyspnea. Hospice was revoked in 33 patients (20.7%) either because major medical complications occurred that required hospital admission, or because the patient situation was too stable. MELD Score and Length of Hospice Stay LOS in hospice was available for all 157 patients. The mean was 38 days (range, days). MELD-A was available for 106 patients. For 51 patients the necessary data for computing MELD-A were not available. For the remaining 106 patients, MELD-A averaged 21.5 (range, 6-45). The mean LOS of patients without a MELD score available at the time of the analysis was 39 days (range, days), not significantly different from the group for which the MELD score was calculated. The regression results are displayed in Table 2. Higher MELD-A was associated with shorter LOS for all three percentiles considered (P 0.01). The 25th percentile LOS was more than twice as long for women as for men (P 0.033). The median and 75th percentile were longer for women but these differences were not statistically significant. Age was not related to LOS (all P 0.18). The quartiles of LOS and their relationship with MELD-A, averaged over all ages and both genders, are displayed in Fig. 1. We now provide an example of the way the model formula (Equation 1) works with the parameter values displayed in Table 2. Suppose a man aged 57 years is admitted with a MELD score of 31. The model provides estimated percentiles for men with MELD and age equal to his. For instance, we obtain the 25th percentile as follows. First, we remember that the logarithm must be applied to the coefficients reported in Table 2, since the values in the table have been exponentiated. Then, we plug these values into Equation 2 to obtain log(25th percentile) log(4.82) log(0.935)* (31 21) log(2.54)*0 log(0.979)*(57 57). Next we simplify the right-hand side to obtain log(25th percentile) log(4.82) log(0.935)*10 and we obtain log(25th percentile) By exponentiating both sides, we obtain a value that we can interpret:

4 MELD SCORE AND HOSPICE REFERRAL 1103 TABLE 2. Quantile Regression for Length of Stay Percentile Coefficient Interpretation Estimate CI P 0.75 exp( 0 ) LOS for average male 25.5 ( ) exp( 1 ) MELD at admission ( ) 0.001* exp( 2 ) Female gender 2.08 ( ) exp( 3 ) Age ( ) exp( 0 ) LOS for average male 11 ( ) exp( 1 ) MELD at admission ( ) 0.003* exp( 2 ) Female gender 1.94 ( ) exp( 3 ) Age ( ) exp( 0 ) LOS for average male 4.82 ( ) exp( 1 ) MELD at admission ( ) 0.009* exp( 2 ) Female gender 2.54 ( ) 0.033* exp( 3 ) Age ( ) NOTE: Coefficients were computed in comparison to a male patient of average age (57 years) and average MELD at admission (21 units). Consequently, the coefficient associated with MELD at admission is the amount by which the LOS percentile is multiplied for each unit greater than 21 or divided for each unit less than 21. The coefficient for female sex represents the amount by which the LOS percentile for men is multiplied to obtain the percentile for women. The coefficient for age is the amount by which the LOS percentile is multiplied (or divided) for each year of age greater than (or less than) 57. Abbreviations: CI, confidence interval; LOS, length of stay. *Significant P value. Thus the model estimates that the interquartile range of LOS for men aged 57 who enter hospice with MELD score of 31 is between 2.5 and 12.1 days. To obtain the 25th and 75th percentiles for women of the same age and MELD score, we note that the variable G (gender) takes on the value 1 for women, and thus we replace 0 with 1 in the third term on the right-hand side of the equation. This means that we add log(2.54) and log(2.08) to the respective equations, leading to an estimated interquartile range of 6.25 to 25.1 days for women aged 57 who enter hospice with a MELD score of 31. The same approach may be used to find the estimated 25th, 50th, or 75th percentile for men or women with any specific MELD score in the range of the current data. Figure 1. Regression-based quartiles of length of stay (LOS) as a function of MELD at admission. Higher MELD score at hospice admission was associated with shorter LOS for all 3 percentiles considered (P < 0.01). 25th percentile 2.46 days If we plug the coefficients for the 75th percentile into the formula, this takes the form: log(75th percentile) log(25.5) log(0.928) *(31 21) log(2.08)*0 log(0.998)*(57 57) By following the same process as above, we obtain 75th percentile 12.1 days Hospice Referral and Consideration for Liver Transplant The flow of patients into hospice was analyzed (Fig. 2): patients (71%) were directly admitted to the hospice service and never considered for the LT evaluation patients (19.7%) at their first evaluation were denied the LT list for medical (19) or psychosocial (12) contraindications. Their mean MELD score was 21.4 at the time of hospice admission patients were initially listed for transplantation; 7 of the 15 (4.4% of the total in the series) were referred to hospice care after being delisted from the LT waiting list because of new, severe, medical or psychosocial conditions contraindicating LT of 15 (5% of the total series) remained on the transplant list and received hospice care up to transplantation or death. In 6 patients, a liver graft became available while on the combined program. The

5 1104 MEDICI ET AL. Figure 2. Composition of our hospice service as a model of integration between palliative care and liver transplantation. hospice benefit was revoked and they received LT. Two died while waiting for a graft. In the 6 patients who underwent transplantation, underlying disease included chronic hepatitis C (3 cases), chronic hepatitis B (1 case), alcohol cirrhosis (1 case), and combination of hepatitis C virus and alcohol abuse (1 case). MELD score progression, waiting time, and LOS in hospice care are displayed in Fig. 3. For the 15 patients who were listed for LT, the average MELD at admission to hospice was 22.1 units (95% confidence interval ) and increased an average of 0.41 units per month (confidence interval , P 0.004). The mean elapsed time between the hospice admission and LT was 81 days (range, days). MELD progression for the 6 patients who received LT is displayed in Fig. 3. These patients had a longer LOS than patients not undergoing transplantation (81 versus 36 days), although the difference was not statistically significant. Five patients survived with a follow-up of 7 to 45 months. Four of the six showed an improvement of their MELD score during hospice care (not significant). One patient died after LT because of progressive heart and kidney failure. His MELD score worsened from 35 to 42 during the hospice stay. Unlike all the other listed patients, he was listed after he was admitted to hospice. In the 82 patients without liver cancer, average MELD at hospice admission was 22.3 (standard deviation 8.6); in the 24 patients with liver cancer it was 18.5 (standard deviation 7.5); P DISCUSSION We have shown a significant correlation between the MELD score and LOS in hospice service. Thus, the MELD score can be used as a quantitative metric along with clinical indicators such as the presence of hepatic Figure 3. Progression of MELD score for patients who received liver transplant. A line (solid for males, dotted for females) connects the three longitudinal MELD measurements for each individual patient. Females do not differ from males statistically in these data. Rather, different line types are assigned to the genders to visually separate the individual MELD progressions. malignancy, performance status, and weight loss to guide referral to hospice. While not discussed here in detail, the hospice interdisciplinary team and the LT team were easily able to integrate palliative and disease directed goals simultaneously. Strong interteam relationships were developed. For example, the hospice team collaboratively

6 MELD SCORE AND HOSPICE REFERRAL 1105 managed the patients with gastrointestinal bleeding (3%), portosystemic encephalopathy (30%), and tense ascites with dyspnea (22%). A previous experience with terminally ill patients who had hepatocellular carcinoma showed that the high prevalence of dyspnea with ascites and stage III-IV hepatic encephalopathy often complicated the management of these patients. 13 Hospice care may be determinant in the management of the several complications of ESLD. Hospice provides the services of physicians in order to manage pain issues, diabetes, or complications such as ascites, spontaneous bacterial peritonitis, and hepatorenal syndrome or hepatopulmonary syndrome. Hospice also provides nurses to address recurrent problems like fatigue or encephalopathy and its proper management with lactulose; dieticians to assess the frequent nutritional deficiencies and the adherence to low-salt diet; psychologists and social workers to support the social and spiritual needs of the patients and their families. Hospice offers all these services to patients whose needs are better met at home rather than with frequent visits to the emergency room or admissions to hospital. So far, hospice has been intended only for terminally-ill patients, creating some limitations in its potential uses. For this reason, we have tried a new approach to hospice and its services with the aim to improve the care of ESLD. The hospice and the LT team were able to improve hospice potentialities, by overcoming the idea that hospice is only meant to provide comfort care. Hospice has been described as the most substantial innovation to serve dying Americans in the last half century. 15 The Hastings Institute (Garrison, New York) has identified three characteristics of hospices valued by consumers: hospice responds to the consequences of profound illness, provides continuity and coordination of care across the continuum of care, and responds to the needs of a diverse community. 16 The World Health Organization s updated definition of palliative care includes relief of suffering throughout illness (not just at end of life) for adults and children. 15 Hospice can provide support to the terminal phases of life and may represent a bridge to LT. Hospice and LT may be perceived as mutually exclusive both by families and healthcare providers, while families and patients may feel that their LT team has lost hope. Communication is an essential element of this process: the hospice team has to provide the time to listen to patients and families questions, to address their anxieties and their concerns, to be available 24-hours a day and to guarantee all the efforts to keep patients and hopes for LT alive. Other barriers to timely referral to hospice are the economic issues, due to the financial incentives for patients pursuing disease-directed therapy. 17 However, these barriers can be overcome thanks to the concept that hospice can pursue both palliative and curative efforts in selected patients. One of the criteria for hospice admission is that the expected survival should be less than 6 months. However, certainty is not required and the fact that at least one-third of patients enter hospice in the last week of life is evidence that physicians need not worry about too-early referrals, rather they should be more concerned about delays in referral that erode palliation and quality. 18 Inpatient palliative care consultation service is another important resource that can facilitate the integration between the inpatient care and the hospice care and favor the transition to the hospice service in critical points of ESLD. LT was completed in a minority of patients in whom there was any intent of transplant (6/15) and even fewer patients with ESLD (6/157). Therefore, hospice referral benefited the vast majority of patients and should be considered first for these patients. The ethical dilemma around best case choices can be resolved by providing patients informed consent and the possibility of both services. This is the first report proposing the MELD score as a possible tool to guide hospice referral. A previous study on 6451 hospice patients demonstrated that most patients are admitted to the hospice service late in the course of their terminal illness, and this was particularly evident in case of patients with liver or biliary tract diseases. In fact, 22.4% of them died within 7 days of their admission. 7 It is also worth emphasizing that our mean LOS is longer than this previous experience, indicating that the approach of combined hospice/liver team managing the terminally ill patients with the help of the MELD score at the time of referral has been beneficial and effective. Hospice care can have an important role in the LT program. We previously reported a case of combined provision of preoperative palliative care through hospice for a patient awaiting LT. 19 In this case, the clinical situation rapidly worsened while the patient was on the LT waiting list, and he agreed to enter the hospice program while receiving concomitant, standard liver-directed support therapy from the transplant team. Now we have extended our experience to 6 patients, showing that a dynamic interaction between the LT list and hospice care is possible and beneficial. The model of integration between hospice/palliative care and LT is shown in Fig. 2. The transplant candidate is unique because they suffer from a terminal condition, but they require complex and complete medical care even when their condition is very advanced. 20 Hospice care, in combination with the transplant program, can improve the care of ESLD. However, as long as hospice is considered only for dying patients, individuals who are listed for LT will not experience the advantages of the hospice program; conversely, people admitted for hospice care will not be considered for LT even if they show signs of clinical improvement. Four of the 6 transplanted patients experienced an improvement of their MELD score during hospice stay, showing that this service can potentially provide effective care to terminally ill patients. Moreover, patients receiving transplantation stayed in the hospice service for a longer period than patients not receiving transplantation, indicating that either earlier referral helps patients get to transplantation, or more indolent disease is favorable for both transplantation and LOS. A study of 4301 seriously ill hospitalized patients demonstrated that a limited amount of readily available clini-

7 1106 MEDICI ET AL. cal variables determine the probability of survival as accurately as the physician clinical evaluation. 21 Although prognostic models such as the MELD score cannot completely replace the clinical evaluation, the current study adds to the evidence that such models provide important information that should be considered together with the medical opinion. Our study, as it is a retrospective analysis, presents several limitations. First, we were not able to calculate the MELD score of 51 patients due to the fact that complete blood tests were not always performed in terminally ill patients, especially when they were moved from other institutions or when they died after a short stay. Second, our study does not yield a cutoff value of the MELD score at which patients have to be referred to the hospice service. Instead, we argue that MELD is one of the tools by which the clinician can make a decision regarding referral to hospice. We note, however, that in the current study almost all patients with a MELD score above 18 died within several months. We anticipate that future studies and clinical care can use the MELD score as a tool for the implementation of palliative services. More information needs to be collected on differences in pain management and patient quality of life between patients in hospice and nonhospice care. This analysis presents the first evidence of a significant correlation between hospice LOS and MELD score, advancing the possibility that the MELD score may serve as a guide for timely hospice referral. A prospective study is required to apply the MELD score in terminally ill patients in order to define a cutoff value of the score as a guide for admission to hospice. Such a cutoff could change the enrollment criteria of ESLD patients. ACKNOWLEDGMENT We thank Erica Chedin for editing the manuscript. REFERENCES 1. Centers for Disease Control and Prevention. Monthly vital statistics report. Available at: hus.htm. Accessed December 5, Greer DS, Mor V, Morris JN, Sherwood S, Kidder D, Birnbaum H. An alternative in terminal care: results of the National Hospice Study. J Chronic Dis 1986;39: Wallston KA, Burger C, Smith RA, Baugher RJ. Comparing the quality of death for hospice and non-hospice cancer patients. Med Care 1988;26: Brooks CH, Smyth-Staruck K. Hospice home care cost savings to third-party insurers. Med Care 1984;22: Centers for Medicare and Medicaid Services. Medicare hospice benefits: a special way of caring for people who have terminal illnesses. Available at: gov/publications/pubs/pdf/02154.pdf. Accessed December 5, Sanchez W, Talwalkar JA. Palliative care for patients with end-stage liver disease ineligible for liver transplantation. Gastroenterol Clin North Am 2006;35: Christakis NA, Escarce JJ. Survival of Medicare patients after enrollment in hospice programs. N Engl J Med 1996; 335: Roth K, Lynn J, Zhong Z, Borum M, Dawson NV. Dying with end stage liver disease with cirrhosis: insights from SUPPORT. Study to understand prognoses and preferences for outcomes and risks of treatment. J Am Geriatr Soc 2000;48:S122 S Wiesner R, Edwards E, Freeman R, Harper A, Kim R, Kamath P, et al. Model for End-Stage Liver Disease (MELD) and allocation of donor livers. Gastroenterology 2003;124: Wiesner RH, McDiarmid SV, Kamath PS, Edwards EB, Malinchoc M, Kremers WK, et al. MELD and PELD: applications of survival models to liver allocation. Liver Transpl 2001;7: McGaughan GW, Stresser SL. To MELD or not to MELD? Hepatology 2001;34: Larson AM, Randall Curtis J. Integrating palliative care for liver transplant candidates. JAMA 2006;10:295: Lin MH, Wu PY, Tsai ST, Lin CL, Chen TW, Hwang SJ. Hospice palliative care for patients with hepatocellular carcinoma in Taiwan. Palliat Med 2004;18: Iwashyna TJ, Zhang JX, Christakis NA. Disease-specific patterns of hospice and related healthcare use in an incidence cohort of seriously ill elderly patients. J Palliat Med 2002:5: Sepulveda C, Marlin A, Yoshida T, Ulrich A. Palliative care: the World Health Organization s global perspective. J Pain Symptom Managet 2002;24: Jennings B, Ryndes T, D Onofrio C, Baily MA. Access to hospice care: expanding boundaries, overcoming barriers. Hastings Cent Rep 2003;35. Available at: thehastingscenter.org/publications/reports.asp. Accessed on September Wright AA, Katz IT. Letting go of the rope aggressive treatment, hospice care, and open access. N Engl J Med 2007;357: Gazelle G. Understanding hospice- an underutilized option for life final chapter. N Engl J Med 2007;26:357: Rossaro L, Troppman C, McVicar JP, Sturges M, Fisher K, Meyers FJ. A strategy for the simultaneous provision of pre-operative palliative care for patients awaiting liver transplantation. Transpl Int 2004;17: Cardenas A, Gines P. Management of complications of cirrhosis in patients awaiting liver transplantation. J Hepatol 2005;42:S124 S Knaus WA, Harrell FE, Lynn J, Goldman L, Phillips RS, Connors AF, et al. The SUPPORT Prognostic Model: objective estimates of survival for seriously ill hospitalized adults. Ann Int Med 1995;122: