LIVER TRANSPLANTATION HAS ADvanced

Transcription

1 ORIGINAL CONTRIBUTION Resource Utilization in Liver Transplantation Effects of Patient Characteristics and Clinical Practice Jonathan Showstack, PhD, MPH Patricia P. Katz, PhD John R. Lake, MD Robert S. Brown, Jr, MD, MPH R. Adams Dudley, MD, MBA Steven Belle, PhD Russell H. Wiesner, MD Rowen K. Zetterman, MD James Everhart, MD, MPH for the NIDDK Liver Transplantation Database Group LIVER TRANSPLANTATION HAS ADvanced markedly over the past decade. One-year survival rates, which were 30% in 1980, are now approaching 90%, with 75% of patients surviving 5 years or more. 1,2 Improvements in immunosuppression, antiviral prophylaxis, and the management of other common posttransplant complications continue to improve patient and graft survival as well as posttransplant quality of life. 3 Although there is considerable pressure to make liver transplantation more cost-effective, few data are available on patient characteristics and clinical practices that are associated with the costs of transplantation. A 1983 analysis of liver transplants paid for by Blue Cross/ Blue Shield of Massachusetts estimated the average cost to be $ in the first year, although this estimate was disputed by several local transplant programs. 4,5 Kankaanpaa 6 studied 32 patients who received 43 transplants between 1981 and 1986 at the University of Pittsburgh, Pittsburgh, Pa, and For editorial comment see p Context Liver transplantation is among the most costly of medical services, yet few studies have addressed the relationship between the resources utilized for this procedure and specific patient characteristics and clinical practices. Objective To assess the association of pretransplant patient characteristics and clinical practices with hospital resource utilization. Design Prospective cohort of patients who received liver transplants between January 1991 and July Setting University of California, San Francisco; Mayo Clinic, Rochester, Minn; and the University of Nebraska, Omaha. Patients Seven hundred eleven patients who received single-organ liver transplants, were at least 16 years old, and had nonfulminant liver disease. Main Outcome Measure Standardized resource utilization derived from a database created by matching all services to a single price list. Results Higher adjusted resource utilization was associated with donor age of 60 years or older (28% [$53 813] greater mean resource utilization; P =.005); recipient age of 60 years or older (17% [$32 795]; P =.01); alcoholic liver disease (26% [$49 596]; P =.002); Child-Pugh class C (41% [$67 658]; P.001); care from the intensive care unit at time of transplant (42% [$77 833]; P.001); death in the hospital (35% [$67 076]; P.001); and having multiple liver transplants during the index hospitalization (154% increase [$ vs $ for 1 transplant]; P.001). Adjusted length of stay and resource utilization also differed significantly among transplant centers. Conclusions Clinical, economic, and ethical dilemmas in liver transplantation are highlighted by these findings. Recipients who were older, had alcoholic liver disease, or were severely ill were the most expensive to treat; this suggests that organ allocation criteria may affect transplant costs. Clinical practices and resource utilization varied considerably among transplant centers; methods to reduce variation in practice patterns, such as clinical guidelines, might lower costs while maintaining quality of care. JAMA. 1999;281: found that charges and length of stay averaged $ and 64 days. A comprehensive study was performed by Evans and colleagues 7 in 1988 involving more than 70% of the programs active in the United States. They reported median hospital charges of $ with an average hospital stay of 33 days. The database created by Evans and colleagues consisted of information aggregated at the transplant center level and did not allow linkage of specific patient and clinical characteristics with costs. Studies of transplants other than Author Affiliations: Department of Medicine (Drs Showstack, Katz, and Dudley) and Liver Transplant Program (Drs Lake and Brown), University of California, San Francisco; Department of Epidemiology, University of Pittsburgh, Pittsburgh, Pa (Dr Belle); Department of Medicine, Mayo Clinic, Rochester, Minn (Dr Wiesner); Department of Medicine, University of Nebraska, Omaha (Dr Zetterman); and Branch of Epidemiology and Clinical Trials, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, Md (Dr Everhart). Dr Lake is now with the Liver Transplant Program, University of Minnesota, Minneapolis, and Dr Brown is with the Center for Liver Disease and Transplantation, Columbia University College of Physicians and Surgeons, New York, NY. A complete list of the members of the NIDDK Liver Transplantation Database Group appears at the end of this article. Corresponding Author and Reprints: Jonathan Showstack, PhD, MPH, Institute for Health Policy Studies, University of California, San Francisco, CA ( jas1@itsa.ucsf.edu) American Medical Association. All rights reserved. JAMA, April 21, 1999 Vol 281, No

2 livers have shown that many factors can influence costs, including disease origin, comorbidities, duration of disease, pretransplant treatments, and type of immunosuppressive therapy. 8,9 Discussions about the important, but contentious, issue of organ allocation policies have focused on clinical criteria and outcomes, with little attention paid to cost-effectiveness. 10,11 Recent changes in these criteria require that more severely ill patients receive transplants before those less severely ill. 12 The study presented here provides estimates of the costs associated with different levels of severity and patient characteristics, thus providing economic information that may be used to improve the cost-effectiveness of liver transplantation and to provide information to create organ allocation policies. All previous analyses of the costs of liver transplantation used nonstandardized hospital charges as their measure of resource utilization. Furthermore, because of limitations in their databases, none of these studies was able to assess the association of the demographic and clinical characteristics of specific donors and patients with the services provided to those patients Through the construction and analysis of a database that contained detailed donor, recipient, and clinical information, the independent correlates of standardized resource utilization were assessed for 711 patients who received liver transplants between 1990 and METHODS This study is a component of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Liver Transplantation Database, which was established to study the demographic and clinical characteristics and outcomes of patients evaluated for and undergoing liver transplantation. 17 The NIDDK Liver Transplantation Database contains extensive pretransplant and posttransplant data for 916 transplant recipients from 3 clinical centers: the Mayo Clinic, Rochester, Minn, the University of Nebraska Medical Center, Omaha, and the University of California, San Francisco (the identities of the clinical centers in the results described below have been masked). The NIDDK Liver Transplantation Database was maintained by the Epidemiology Data Coordinating Center at the University of Pittsburgh Graduate School of Public Health. Patient recruitment began in April 1990 and ended in June Information on the specific data collected from each patient has been published previously. 18 The results reported below are for the initial transplant hospitalization (from day of transplantation through discharge) for 711 patients who were aged 16 years or older, had nonfulminant liver disease, and for whom a liver was the only organ transplanted. The study was approved by the institutional review boards at the 3 study centers and at the data coordinating center. All patients gave informed consent to participate in the study. A resource utilization database was constructed to add economic data to the demographic and clinical data previously collected in the NIDDK Liver Transplantation Database. A detailed description of the methods used in the construction of the resource utilization database can be found elsewhere. 19 Standardization of resource utilization across centers and over time was accomplished by applying a 1995 price list from 1 of the 3 centers to the hospital services provided to patients at each center. (At the index center, charges and costs are highly correlated and the ratio of charges to costs is relatively consistent among cost centers.) Professional fees were not included. The dollar results reported below should be interpreted as the amounts that would have been charged for hospital care if all 711 patients had received their transplants at the index center in July By standardizing resource utilization, these methods allow an assessment of the relative contribution of patient demographic and clinical characteristics to resources utilized for the transplant hospitalization. Statistical models were constructed for the associations of patient demographic and clinical characteristics at the time of transplantation with standardized hospital resource utilization (measured from the day of transplantation to the day of discharge). Frequency distributions, scatter plots, and first-order correlations were computed to assess the distributions of, and correlations among, the independent variables. Multivariate models were computed to assess the independent association of resource utilization with patient demographic and clinical characteristics. Independent variables were included in the multivariate analyses based on their a priori theoretical association with and influence on resource utilization. To avoid capitalizing on chance associations among the dependent and independent variables, iterative procedures that select variables based on their statistical properties were avoided. Resource utilization and length of stay often have high outlier values that create skewed distributions. To avoid the potential bias that this non-normal distribution might contribute to multivariate analyses, several techniques were used to normalize the data. Among the methods used in separate analyses were truncation of the distributions of the dependent variables to 3 SDs from the mean, a logarithmic conversion of the dependent variables, and elimination of all cases with values of the dependent variables greater than 3 SDs from the mean. Since each of these analyses produced similar results, the results for the first technique (truncation) are described below. (The mean nontruncated length of stay was 22.0 days compared with 21.3 for the truncated distribution; similarly, the respective means for total resource utilization were $ and $ ) Many of the clinical variables were both theoretically and empirically correlated with each other. For example, while both Child-Pugh class and Karnofsky score were associated with resource utilization, they were also highly corre JAMA, April 21, 1999 Vol 281, No American Medical Association. All rights reserved.

3 lated with each other. To avoid collinearity in the multivariate analyses, only 1 of a pair of highly correlated variables was included: those that had the highest firstorder correlation with the dependent variable being studied (in the above example, the Child-Pugh class). The Child- Pugh class is a categorization of clinical severity (with class A being least severe and class C most severe) that is based on the degree of ascites and encephalopathy, and abnormalities in albumin, bilirubin, and coagulation. Separate linear regression models were computed for length of stay and resource utilization. The multivariate models included as independent variables were: transplant center, donor age, recipient age, sex, diagnosis of alcoholic liver disease (chronic disease, with at least 6 months abstinence from alcohol prior to surgery), diagnosis of cholestatic liver disease, obesity (body mass index 30 kg/m 2 ), renal function (creatinine levels of 88.4, , , and 265 µmol/l [ 1, 1-1.9, 2-2.9, and 3 mg/dl], and/or patient receiving dialysis), Child-Pugh class, gastrointestinal tract bleeding during immediate pretransplant hospitalization, dialysis at the time of transplantation, identical donor-recipient blood types, immunosuppressive regimen (tacrolimus vs cyclosporine, based on intention to treat), nutritional status, United Network for Organ Sharing (UNOS) status (at home, in hospital but not in the intensive care unit [ICU], or in the ICU), prior hepatobiliary surgery (excluding portosystemic shunting), and prior portosystemic shunt. In additional models, length of stay was added as an independent variable both to assess the effect of length of stay on resource utilization and to compare the association of resource utilization with the independent variables when length of stay was included in, then excluded from, the model. Other variables added to later models included death in the hospital and multiple transplants during the index hospitalization. All analyses were performed using the SAS statistical package (SAS Inc, Cary, NC). A result with P.05 was judged to be statistically significant. RESULTS The demographic and clinical characteristics of the study population are displayed in TABLE 1. The study population had a slightly greater proportion of men than women, was predominantly white, and had a mean age of 49.5 years. The most common primary diagnoses were chronic hepatitis (38%), cholestatic liver disease (31%), and alcoholic liver disease (13% without and 7% with hepatitis C). Almost half (49%) of the patients were classified in Child-Pugh class B, and 35% were class C. Nearly one quarter (26%) had prior hepatobiliary surgery, 10% had a prior porto-systemic shunt procedure, and 6% had gastrointestinal tract bleeding just prior to transplantation. Four percent had creatinine levels greater than 265 µmol/l (3 mg/dl), and 3% were receiving dialysis. Thirty-three percent of the study patients were in the hospital at the time of the transplant (UNOS status in hospital or in the ICU), with a quarter of those in the ICU. Nutritional status was fair for the majority (62%) of the study population. Cytomegalovirus serolgic status for the donor and/or recipient were unavailable for 119 patients (16%). Of the 592 patients for whom cytomegalovirus data were available, 44% were donorrecipient positive, 10% donor-recipient negative, 29% donor negative-recipient positive, and 16% donor positiverecipient negative. The donor-recipient ABO match was identical for 94% of the patients. Seven percent of donors and 19% of recipients were aged 60 years or older. Twenty-two patients (3% of the study population) required retransplantation and 39 patients (6%) died during the index hospitalization. Mean length of stay (from day of transplantation to day of discharge) was 21.2 days and mean total resource utilization (in 1995 dollars) was $ (TABLE 2). Also shown in Table 2 are the mean resource utilization and percentages of total resource utilization contributed by each category of service utilization. Room and care averaged $ or 26% of total resource utilization, of which almost half was due to special care (nurse to patient ratios of 1:2). Diagnostics (imaging, laboratory, and pathology) averaged $ Table 1. Demographic and Clinical Characteristics of Study Patients* Characteristic No. (%) Women 320 (45) Race White 575 (81) Hispanic 63 (9) Black 27 (4) Asian 25 (4) Other 21 (3) Age, mean (SD), y 49.5 (11) Primary diagnosis Chronic hepatitis 270 (38) Cholestatic liver disease 222 (31) Alcoholic liver disease 89 (13) Alcoholic liver disease and 51 (7) hepatitis C Other 79 (11) Child-Pugh Class A 119 (17) B 345 (49) C 247 (35) Prior hepatobiliary surgery 183 (26) Prior porto-systemic shunt 68 (10) Current gastrointestinal tract bleeding 40 (6) Dialysis 18 (3) Obese (body mass index 128 (18) 30 kg/m 2 ) Creatinine, µmol/l (mg/dl) 88.4 ( 1) 381 (54) 88.4 and 177 ( 1 and 2) 266 (38) 177 and 265 ( 2 and 3) 33 (5) 265 ( 3) 26 (4) United Network for Organ Sharing status At home 482 (68) In hospital 162 (24) In intensive care unit 62 (9) Nutritional status Excellent 158 (22) Fair 434 (62) Poor 114 (16) Donor-recipient cytomegalovirus match Donor +, recipient (44) Donor, recipient 61 (10) Donor, recipient (29) Donor +, recipient 95 (16) Donor-recipient ABO match Identical 667 (94) Donor age 60 y 51 (7) Recipient age 60 y 133 (19) Primary medication Tacrolimus 56 (8) Retransplantation 22 (3) Died 39 (6) *Total number of single-organ, nonfulminant, adult transplant recipients is 711. A plus sign indicated positive; minus sign, negative. Excluding patients with prior porto-systemic shunt. Data missing for 119 matches. During study hospitalization only American Medical Association. All rights reserved. JAMA, April 21, 1999 Vol 281, No

4 or 17% of total resource utilization, with laboratory charges accounting for the majority of diagnostic charges. Treatments, including operating room charges, medications, blood products, and respiratory services, averaged $67 826, or 33% of total resource utilization, with medications accounting for approximately one third of treatment resources (12% of total resources). The multivariate models provided an estimate of the association of the independent variables with adjusted mean length of stay and adjusted mean resource utilization. The comparisons between different categories of a particular independent variable (eg, women compared with men) are adjusted for all of the other independent variables in the multivariate models. Compared with center 1, center 2 used $ (20%) more resources (P =.005), with no difference in adjusted mean length of stay, and center 3 used $ (44%) more resources (P.001) and had a significantly longer length of stay (7.6 days, P.001). Patients with transplants from older donors had longer lengths of stay and used significantly more resources compared with younger donors (5.9 Table 2. Mean Resource Use by Service Category* Resource Mean (%), Organ acquisition (15) Room and care (26) Special care (12) Diagnostics Imaging 6502 (3) Chest radiography 1936 (1) Ultrasound 1074 (1) Laboratory (13) Immunosuppressive 1650 (1) monitoring Pathology 1279 (1) Total (17) Treatment Operating room (8) Medications (12) Immunosuppressives (7) Anti-infectives 8714 (4) Blood products (10) Respiratory services 7174 (4) Total (33) Other Material services (8) Miscellaneous 3137 (2) Total (9) Total Resource Use (100) *Resource use is calculated from day of transplantation to discharge and is not truncated. The denominator for each percentage is total resource use. Subtotals may not add to totals due to rounding. days, P =.01; $ [ 28%], P =.005). Compared with recipients 59 years old or younger, older recipients had increased length of stay and resource utilization (3.9 days, P =.01; and $ [ 17%], P =.01). There were no significant differences between women and men as recipients in either length of stay or resource utilization. Of the clinical variables in the model, alcoholic liver disease was associated with a 5.5-day increase in length of stay (P=.004) and $ (26%) greater resource utilization (P=.002). Neither length of stay nor resource utilization was associated with renal function (creatinine level) at the time of transplantation. Patients in Child-Pugh class B did not differ significantly from patients in Child-Pugh class A, while those in Child- Pugh class C had significantly longer lengths of stay (7.5 days, P.001) and greater resource utilization ($ [ 41%], P.001). Compared with patients at home, patients in the hospital (but not in ICU) at the time of the surgery stayed longer and used more resources (3.7 days, P =.02; and $ [ 15%], P =.04, respectively), while patients in the ICU at the time of the surgery had significantly longer stays (9.3 days, P.001) and greater resource utilization ($ [ 42%], P.001). There was a trend for patients with prior hepatic surgery to use more resources ($ [11%], P =.07), with no difference in length of stay. Patients who died utilized $ (35%) more resources (P=.001) than those alive at discharge, yet had shorter lengths of stay (5.2 days shorter than those alive at discharge, P.049). Patients with multiple transplants during the index hospitalization stayed an additional 23.7 days on average (P.001) and used more than times the amount of resources ($ compared with $ ; P.001). Other independent variables included in the primary models (prior porto-systemic shunt, cholestatic liver disease, gastrointestinal tract bleeding, dialysis, obesity, nutritional status, identical blood type, donor-recipient cytomegalovirus serologic match, and medications) were not significantly associated with either length of stay or resource utilization, nor was patient ethnicity. Among the individual components of the Child-Pugh score, the presence of encephalopathy and a lower albumin concentration were associated with greater resource utilization (P.001 and P =.03, respectively). Preoperative ascites, prolonged prothrombin time, and total bilirubin were not significantly associated with resource utilization. In multivariate models for resource utilization that included length of stay as an independent variable, center 2 utilized 16% more resources per day (P.001) than center 1, and center 3 used 9% more per day (P=.004). For most of the other independent variables the increases in length of stay and resource utilization were similar, implying that the primary mechanism for the increased resource utilization was through added length of stay (as opposed to more services per day). The proportional effects of selected independent variables on adjusted mean length of stay and adjusted mean resource utilization are shown in the FIGURE. Multivariate analyses were also performed to assess the association of clinical characteristics with the categories of resource utilization shown in Table 2. These analyses produced results consistent with those for total resource utilization, with generally the same independent variables having significant associations with resource utilization. COMMENT Liver transplantation is both lifesaving and one of the most resourceintensive procedures in medicine. The results reported here help to quantify the economic consequences of difficult clinical and social issues, such as the greatly increased use of services associated with a pretransplant diagnosis of alcoholic liver disease. In addition, differences in clinical practices were shown to result in large variations in resource utilization. Multivariate analyses suggested that the clinical characteristics most associated with resource utilization were do JAMA, April 21, 1999 Vol 281, No American Medical Association. All rights reserved.

5 nor age older than 59 years, being severely ill at the time of transplantation (Child-Pugh class C and/or in the ICU), and alcoholic liver disease. The principal mechanism through which the clinical characteristics caused higher resource utilization was to increase length of stay and, to a lesser extent, require higher nurse to patient ratios (eg, time spent in special care compared with ward care). After adjusting for severity of illness, the 3 study centers had differences in lengths of stay and in the levels of nursing care provided to similar patients, which are both, to a degree, discretionary and based on patterns of practice. For example, a program may use a nonhospital setting for antibiotic administration, wound care, or rehabilitation, which would reduce hospital length of stay. In another example, the ICU may be used for longer periods of time, which would increase the cost of nursing care. Our results imply that some savings might be obtained (with no decrement in clinical outcomes) through adjustments in length of stay and level of nursing care services. Donor age was highly associated with resource utilization. There is some evidence that outcomes in liver transplantation are associated with the characteristics of the liver at the time of harvest. 20 A poor-quality liver may produce increased morbidity in the transplanted patient, such as early allograft dysfunction, thereby causing increased resource utilization. 21 To the extent that older donors may have poorer-quality livers, this may be the mechanism through which donor age is associated with higher resource utilization. After adjustment for other characteristics, patients with alcoholic liver disease used approximately 26% more hospital resources than did other patients. Alcoholic liver disease is a relatively common, if controversial, indication for liver transplantation, with many programs having specific criteria for such patients awaiting transplants. 22 Issues centering around a patient s personal responsibility, mental health, and social circumstances create difficult dilemmas when allocating a scarce resource such as livers among alcoholic and nonalcoholic patients These results have substantial relevance to the current controversy concerning organ allocation and distribution. 26 UNOS recently changed the criteria for status 2 designation on the liver transplantation waiting list. To be listed as status 2B, patients must have a Child-Pugh score of 10 or greater (Child-Pugh class C). To be listed as status 2A (the highest urgency for a patient with chronic liver disease), the patient must be in class C and experience a complication that requires ICU care. This change makes it likely that most liver transplants will be for recipients with these clinical characteristics. Our data suggest that this policy could lead to a substantial increase in resources utilized for liver transplantation in the United States. The increasing concern about costs on the part of both public and private payers for medical care suggests that costs will continue to have a role in resource allocation decisions, although these decisions become considerably more complex when trying to balance costs against clinical effectiveness. 27 The interests of equity and clinical necessity should, of course, provide a necessary balance against economic considerations for patients with end-stage liver disease. In addition, these results provide strong support for the need to adjust payments for medical care according to a patient s clinical characteristics and severity of illness. 28 This study examined a limited group of patients (adults with nonfulminant liver disease who received single-organ transplants) who may be the least expensive to treat with liver Figure. Percentage Increase in Adjusted Mean Length of Stay and Mean Resource Utilization for Selected Clinical, Donor, and Patient Characteristics Center 2 Center 3 Donor Age 60 y Recipient Age 60 y Women Alcoholic Liver Disease Prior Hepatic Surgery Child-Pugh Class B Child-Pugh Class C UNOS Status: in Hospital UNOS Status: in Intensive Care Unit P =.62 P =.64 P =.78 P =.16 P =.07 P =.15 P =.11 P =.005 P =.01 P =.01 P =.02 P =.04 P =.01 P =.005 P =.004 P =.001 Length of Stay Resource Use P =.001 P = Increase Compared With Baseline, % Length of stay and resource utilization were adjusted for transplant center, donor age, recipient age and sex, alcoholic liver disease, body mass index of more than 30 kg/m 2, blood urea nitrogen, Child-Pugh class, cholestatic liver disease, renal function, current gastrointestinal tract bleeding, currently receiving dialysis, identical donor-recipient blood types, immunosuppressive regimen, nutritional status, United Network for Organ Sharing (UNOS) status, and prior hepatobiliary surgery (excluding porto-systemic shunting), and prior portosystemic shunt. The comparison categories for the variables were center 1, donor and recipient ages of 59 years or younger, men as recipients, no alcoholic liver disease, no prior hepatic surgery, Child-Pugh class A, and at home for United Network for Organ Sharing status American Medical Association. All rights reserved. JAMA, April 21, 1999 Vol 281, No

6 transplantation. This was not a costeffectiveness study, rather a study of factors that affect hospital resource utilization in light of the generally good outcomes of the patients studied. The methods permitted the identification of association, not causation, and we did not assess the appropriateness of the services provided to the study patients. The patients treated and the care provided at these 3 large transplant centers may not represent patients or care patterns at other centers. The NIDDK Liver Transplantation Database, however, contains comprehensive clinical data for a large study population, and the multivariate models included a wide range of independent variables. Thus, the results support relatively robust inferences about the association of the resources utilized in liver transplantation with patient characteristics and clinical care patterns. In summary, these results highlight the clinical, economic, and ethical dilemmas in liver transplantation. Recipients who were older, had alcoholic liver disease, and/or were severely ill were the most expensive to treat, which suggests that recent changes in organ allocation criteria may produce higher average transplant costs. Clinical practices and resource utilization varied considerably among transplant centers. Similar to other types of medical care with high variations in use for similar patients, 29 it appears that significant savings might result from guidelines or other methods to standardize resource utilization for liver transplantation. Funding/Support: This study was supported by National Institutes of Health contracts N01-DK , N01- DK , N01-DK , and N01-DK Members of the NIDDK Liver Transplantation Database Group: Katherine M. Detre, MD, DrPH (principal investigator), A. Jake Demetris, MD (coinvestigator), Yuling L. Wei, MS (coinvestigator), Manuel Lombardero, MS (statistician), Eric Seaberg, MPH (statistician), Sharon Lawlor, MBA (data manager/ analyst), Heather Eng (data analyst), JoAnn Naujelis (data manager), Jacqueline Haber (data manager assistant), Shannon Fitz Gerald, MS (graduate research assistant), and Gerald L. Swanson (systems analyst) from the Department of Epidemiology, University of Pittsburgh Graduate School of Public Health, Pittsburgh, Pa; Ruud A. F. Krom, MD (coinvestigator), Michael K. Porayko, MD (coinvestigator), and Lori Schwerman, RN (research nurse coordinator) from the Departments of Medicine and Surgery, Mayo Clinic Foundation, Rochester, Minn; Byers W. Shaw, Jr, MD (coinvestigator) and Karen Taylor, RN, BSN (research nurse coordinator) from the Departments of Medicine and Surgery, the University of Nebraska Medical Center, Omaha; Nancy L. Ascher, MD, PhD (principal investigator), Cherie Bremer-Kamp, RN (research nurse coordinator), and M. Esther Colwell (programmer) from the Departments of Medicine and Surgery, University of California, San Francisco; and Tommie Sue Tralka (program director) and Jay H. Hoofnagle, MD (division director) from the Division of Digestive Diseases and Nutrition, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, Md. REFERENCES 1. Starzl TE, Demetris AJ, Van Thiel D. Liver transplantation. N Engl J Med. 1989;329: , Lake JR. Changing indications for liver transplantation. Gastroenterol Clin North Am. 1993;22: United Network for Organ Sharing. Waiting list charts by organ by ABO blood group, gender and age. UNOS Update. 1994;10: Blue Cross/Blue Shield. The Final Report of the Task Force on Liver Transplantation in Massachusetts. Boston, Mass: Blue Cross/Blue Shield; Luebs HW. Cost considerations. Semin Liver Dis. 1985;5: Kankaanpaa J. Cost-effectiveness of liver transplantations: how to apply the results in resource allocation. Prev Med. 1990;19: Evans RW, Manninen DL, Dong FB. An economic analysis of liver transplantation: costs, insurance coverage, and reimbursement. Gastroenterol Clin North Am. 1993;22: Showstack J, Katz P, Amend W, et al. The effect of cyclosporine on the use of hospital resources for kidneytransplantation.nengljmed.1989;321: Showstack J, Katz P, Amend W, Salvatierra O. The association of cyclosporine with the 1-year costs of cadaver-donor kidney transplants. JAMA. 1990;264: Steinbrook R. Allocating livers: devising a fair system. N Engl J Med. 1997;336: Shaw BW. A collective wisdom. Liver Transpl Surg. 1997;3: UnitedNetworkforOrganSharing.AmendedUNOS policy 3.6: allocation of livers. World Wide Web address: Accessed March 5, Staschak S, Wagner S, Block G, et al. A cost comparison of liver transplantation with FK 506 or CyA as the primary immunosuppressive agent. Transplant Proc. 1990;22: Pageaux GP, Souche B, Perney P, et al. Results and cost of orthotopic liver transplantation for alcoholic cirrhosis. Transplant Proc. 1993;25: Wiesner RH, Porayko MK, Dickson ER, et al. Selection and timing of liver transplantation in primary biliary cirrhosis and primary sclerosing cholangitis. Hepatology. 1992;16: Lake JR, Gorman KJ, Esquivel CO, et al. The impact of immunosuppressive regimens on the cost of liver transplantation: results from the US FK-506 multicenter trial. Transplantation. 1995;60: Detre KM. The NIDDK liver transplantation database. In: Terasaki P, ed. Clinical Transplants, Los Angeles, Calif: UCLA Tissue Typing Laboratory; 1986: Wei YL, Detre KM, Everhart JE, and the NIDDK Liver Transplantation Database. The NIDDK Liver Transplantation Database. Liver Transpl Surg. 1997; 3: Katz PP, Showstack JA, Lake J, et al. Methods to estimate and analyze medical care resource use: an example from liver transplantation. Int J Technol Assess Health Care. In press. 20. Hoofnagle JH, Lombardero M, Zetterman RK, and the NIDDK Liver Transplantation Database. Donor age and outcome of liver transplantation. Hepatology. 1997;24: Deschenes M, Belle SH, Krom RAF, Lake JR, for the National Institute of Diabetes and Digestive and Kidney Diseases Liver Transplantation Database. Early allograft dysfunction following liver transplantation: a definition and predictors of outcome. Transplantation. 1998;66: Everhart JE, Beresford TP. Liver transplantation for alcoholic liver disease: a survey of transplantation programs in the United States. Liver Transpl Surg. 1997; 3: Moss AH, Siegler M. Should alcoholics compete equally for transplantation? JAMA. 1992;265: Benjamin M. Transplantation for alcoholic liver disease: the ethical issues. Liver Transpl Surg. 1997;3: Ubel PA. Transplantation in alcoholics: separating prognosis and responsibility from social biases. Liver Transpl Surg. 1997;3: Stapleton S. Physicians, states, hospitals challenge transplant rules. American Medical News. October 19, 1998;41: Hadorn DC. Setting health care priorities in Oregon: cost-effectiveness meets the rule of rescue. JAMA. 1991;265: Dudley RA, Harrell FE, Smith LR, et al. Comparison of analytic models for estimating the effects of clinical factors on the cost of coronary artery bypass graft surgery. J Clin Epidemiol. 1993;46: Wennberg JE, Cooper MM. The Dartmouth Atlas of Health Care in the United States. Chicago, Ill: American Hospital Publishing; JAMA, April 21, 1999 Vol 281, No American Medical Association. All rights reserved.