RISKS AND COSTS OF END-STAGE RENAL DISEASE AFTER HEART TRANSPLANTATION^1,2

Abstract

Objectives. 

To estimate the risks and costs of end-stage renal disease (ESRD) after heart transplantation.

Background. 

Previous studies have shown high rates of ESRD among solid-organ transplant patients, but the relevance of these studies for current transplant practices and policies is unclear. Limitations of prior studies include relatively small, single-center samples and estimates made before implementing suggested practice changes to reduce ESRD risk.

Methods. 

Medicare beneficiaries who underwent heart transplantation between 1989 and 1994 were eligible for study inclusion (n=2088). Thirty-four patients undergoing dialysis or who had the diagnosis of ESRD before or at transplantation were excluded from the study. ESRD was defined as any patient undergoing renal transplantation or requiring dialysis for more than 3 months. Mortality and ESRD events were recorded up to 1995. ESRD risk was estimated using the Kaplan-Meier product-limit estimator and logistic regression analyses. Linear regression was performed to determine expenditures for treating ESRD, and we developed long-term models of the risk and direct medical costs of ESRD care.

Results. 

The annual risk of ESRD was 0.37% in the first year after transplant and increased to 4.49% by the sixth posttransplant year. There was no significant trend in the risk of ESRD based on the year of transplantation, even after adjusting for patient characteristics. The average cumulative 10-year direct cost of ESRD per patient undergoing heart transplantation exceeded $13,000.

Conclusions. 

In a large, national sample of patients undergoing heart transplantation, ESRD is not rare, even for patients undergoing transplant after the development of new practices intended to reduce its occurrence. ESRD remains an important component of the costs of heart transplantation.

*Abbreviations: CAPD, continuous ambulatory peritoneal dialysis; CCPD, continuous cycling peritoneal dialysis; CPI, Consumer Price Index; DRG, diagnosis-related group; ESRD, end-stage renal disease; HCFA, Health Care Financing Administration; ICD, international classification of diseases.

Nephrotoxicity, and the associated risk of end-stage renal disease (ESRD*), has been a primary concern for patients undergoing transplantation. Since 1984, 18 studies were published that reported follow-up of heart transplant patients for at least 1 year and documented the number of patients who subsequently became dialysis-dependent ^(1-18) (Table 1). In all, the 18 studies included 2017 patients, with lengths of follow-up between 1 and 8 years. The mean number of patients per study was 141, with a minimum of 41 and a maximum of 430. An average of 5.0% of patients became dialysis-dependent (range, 0-14.3%). In the study with the longest follow-up, Myers et al. ⁽¹²⁾ estimated that the cumulative risk of ESRD at 8 years after transplantation was 9.8%.

Table 1: Studies reporting end-stage renal disease after cardiac transplantation

These remains, however, some doubt as to the significance of these findings to current practice ^(8,9,16). For example, Greenberg et al. ⁽⁵⁾ concluded in 1990 that "the risk of end-stage renal disease is significant, but not prohibitive." Previous study designs varied in a number of ways that could leave considerable doubt about the risk of ESRD. First, many of these studies were conducted in an era before the adoption of new practices to reduce the risk of nephrotoxicity. Since concerns about the risk of nephrotoxicity were first reported in 1984 ^(1,10), the suggested practice changes include cyclosporine dose reductions, close monitoring of cyclosporine levels, and the use of newer, potentially less nephrotoxic drug formulations ^(1,4,6). Many of the patients included in the 18 studies underwent transplantation in the early-to-mid 1980s, before such practices may have been fully implemented. Second, estimating the risk of ESRD was not the main endpoint for most of these studies. The definition of ESRD used in the study was not always stated clearly, leaving some doubt as to whether patients were accurately classified as having ESRD. Third, most studies presented single-center experiences, often containing too few patients with too brief a follow-up interval to accurately estimate the long-term risk of ESRD. Finally, no study documented the economic consequences of ESRD in heart transplantation. In summary, the considerable heterogeneity among these studies limits the ability to make reliable conclusions about the current risk of ESRD after heart transplantation and its consequences for health care costs.

The primary goal of this study, therefore, was to obtain a more generalizable estimate of the risk of ESRD in patients undergoing heart transplantation during a period when suggestions for practices intended to reduce the risk of ESRD had already been widely disseminated. The second goal of this study was to estimate the expected costs of posttransplant ESRD in heart transplant recipients. To achieve these goals, we analyzed data from a national sample of patients who had undergone heart transplantation from 1989 to 1994.

METHODS

The study was conducted in two parts. In the first part, we assessed the incidence of ESRD and medical expenditures of care for patients who had undergone heart transplantation between 1989 and 1994 and who were Medicare beneficiaries at the time of their transplant. Medicare began reimbursing for heart transplantation in 1984. Because one of the primary goals of this article is to provide reliable estimates of the economic consequences of ESRD after heart transplantation, which may then influence policy decisions, a relevant issue is deciding the time horizon (i.e., the longest number of years of follow-up posttransplant) for estimating these costs. The U.S. Public Health Service's Panel on Cost-Effectiveness in Health and Medicine recommended that the time horizon for a cost analysis "should be long enough to capture all relevant future effects of a health care intervention" ⁽¹⁹⁾. However, predicting ESRD risk and associated costs much beyond the period where data is available is problematic. The median patient survival worldwide now exceeds 8 years from time of transplant and 11 years for patients who survive the first posttransplant year ⁽²⁰⁾. We, therefore, decided in our reference-case model to estimate costs using a Markov model ⁽¹⁹⁾ for each year up to 10 years posttransplantation. Because more than 40% of patients are expected to survive beyond 10 years, we also report the consequences of extending the model's time horizon to 20 years.

ESRD Risk and Expenditures-Medicare Data

Data sources. We identified 2088 persons from the Health Care Financing Administration's (HCFA) Medicare Provider Analysis and Review file who were discharged from a hospital between 1989 and 1994 with a principal procedure of heart transplantation (diagnosis-related group [DRG] code 103 or international classification of disease [ICD]-9-CM code of 37.5). Patients were not eligible if they had had a heart transplant within the previous year (n=1). We also eliminated patients who were unlikely to have complete and reliable records of transplant care and complications. Patients were excluded if they (1) were under the age of 18 years (n=6); (2) resided outside of the U.S. (n=6); (3) were enrolled in an HMO (n=10); (4) had noncontinuous enrollment in Medicare Part A (n=16) or Part B (n=49) over the period after their initial heart transplant admission (through 1995 data); or (5) were admitted to a non-acute care hospital for their transplant (n=1). We excluded the three patients discharged alive with new heart transplants after a length of stay of only 1 day, which, given the short duration, were likely to have been miscoded as being admitted for a heart transplant procedure. The 34 patients undergoing dialysis or who were designated as eligible for Medicare because of ESRD at the time of heart transplantation also were excluded. The final heart transplant population for these analyses equaled 1963.

We used Medicare administrative claims files to estimate the incidence of ESRD and expenditures associated with the following five categories: inpatient admissions, outpatient services, home health visits, hospice services, and physician services. The Appendix lists the Medicare claims files used in this study. Because of data limitations involving the large number of claims involved, HCFA produced comprehensive physician and nonhospital supplier billing data for a 5% random sample. We restricted analysis of physician and supplier services to this 5% sample (n=99) of transplant recipients.

Study endpoints. The health outcomes of the study were patient survival and the occurrence of ESRD. We defined the occurrence of ESRD as a patient who was recorded as eligible for Medicare because of ESRD or received maintenance dialysis treatments (e.g., hemodialysis 2-3 times per week) over at least a 90-day period after cardiac transplantation. Patients receiving dialysis over a shorter interval, even if the interval ended in death, were not considered to have experienced ESRD. Relevant DRG, ICD-9, and dialysis procedure codes are shown in the Appendix. Others have documented the reliability of applying these criteria to Medicare data files for diagnosing the incidence of ESRD ⁽²¹⁾.

Data analysis. We report patient demographic (mean age, sex, and race) and geographic (region) variables, in addition to patient survival rates and ESRD incidence in the posttransplant follow-up period (up to 6 years). We estimated the annual risk of ESRD and mortality, weighted for the proportion of patients in each cohort, using Kaplan-Meier product-limit estimation methods. We used logistic regression analyses to estimate ESRD risk adjusted for age, sex, race, and year of transplant.

For each posttransplant year, we calculated the mean total expenditure for covered services, including expenditures for inpatient acute and non-acute care, outpatient care, home health care, hospice care, and physician visits. These expenditures are reported in 1997 dollars, inflated by using the overall Consumer Price Index (CPI). We used the overall CPI, instead of the Medical CPI, because experts recently have argued that the latter overestimates price inflation ^(19,22).

Linear regression models of total expenditures were estimated for each posttransplant year. Covariates in the model included age, sex, race, and the year that the patient underwent transplantation. The model included main effects and second-order interactions between patient demographic variables. We estimated the average added expense per patient of treating ESRD (dialysis or transplantation) by including a variable indicating whether the patient had been treated for ESRD at any time during that year. This variable included two types of patients: those who were being treated for ESRD at the start of the year, regardless of whether they survived to the end of the year, and those who were not treated for ESRD at the start of the year, but subsequently survived at least 90 days and were treated for ESRD.

Other comorbidities, such as peripheral (reno) vascular disease or hypertension, are not reported reliably in Medicare databases. As a result, patients who actually develop ESRD might have been more costly to treat because these unobservable characteristics may correlate with ESRD risk and with medical expenditures. To control for this potential source of bias, we included a variable for whether a patient subsequently developed ESRD posttransplant. Thus, our estimate of the cost of ESRD is based on the increase in costs after developing ESRD. We control for any cost differential that existed between ESRD and non-ESRD patients before the development of ESRD. We report those effects that are significantly different from 0 at the 0.01 significance level.

The total expense of ESRD treatment per transplant recipient was estimated by combining the estimates from the logistic regressions of adjusted ESRD risk and the cost per year of treating ESRD. Specifically, for each posttransplant year, we multiplied the risk of developing ESRD in that year by the extra per-patient expense of ESRD treatment obtained from the linear regressions.

Costs are estimated for acute inpatient admissions by multiplying reported charges for each admission with hospital-specific calculated cost-to-charge ratios by department. The ratio is computed by dividing the Medicare operating cost (from the HCFA cost reports) by the Medicare covered charge (from the claim file).

Reference-Case Markov Model

To estimate the long-term cost impact of ESRD, we developed a 5-state Markov model of progression into ESRD. The five states included in the model were (1) functioning heart transplant without ESRD, (2) functioning heart transplant with ESRD undergoing dialysis, (3) functioning heart transplant with ESRD in the first posttransplant year after having undergone renal transplantation, and (4) functioning heart transplant with ESRD in succeeding postrenal transplant years, and (5) death.

The risks of developing ESRD and probability of transplantation for years 1 to 6 after transplant were based on estimates from our analysis of risks and cost impact of ESRD using the Medicare claims files described above. The question arises about how to assign ESRD risk in subsequent years. As will be shown below, the risk of ESRD increases sharply in the fifth and sixth posttransplant years; thus, using the common technique of "carrying the last value forward" may provide an upwardly biased estimate of the annual risk. The average of all prior annual rates equals 1.6% per year. In the reference case, to be conservative, we assigned a rate of ESRD equal to 1% per year. Using this estimate gives a cumulative risk of ESRD of 7.8%, which is less than the 9.8% risk reported by Myers et al. ⁽¹²⁾. In a sensitivity analysis, we increased the risk of ESRD to 2.8% per year, so that the 8-year risk equaled the estimate in Myers et al.'s published report.

The annual costs of ESRD care (dialysis or transplant) for patients who have not undergone heart transplantation has been published previously ⁽²³⁾. For example, the first-year costs of renal transplantation exceed $100,000, primarily due to procurement and procedure costs will not vary significantly because the patient had a previous heart transplant. In subsequent postrenal transplant years, the annual costs for maintenance care, excluding the costs of immunosuppressive drugs (which heart transplant patients will already be taking), is approximately $10,000 ^(23,24). The average treatment cost of chronic dialysis per year has been estimated as $53,300 ^(23,25). However, some costs are related to managing underlying illnesses, such as heart disease or diabetes, and should not be included in this model as an added cost of developing ESRD after heart transplantation. To estimate costs of chronic dialysis after heart transplantation, we computed the average added ESRD costs for dialysis care among Medicare beneficiaries as equal to $45,267.

Indirect and other nonmedical costs, such as those associated with lost work productivity, caregiver burden, and transportation, are often considered important elements of a societal cost analysis. However, the U.S. Panel recommends that the "costs and outcomes that are insignificant in the context of the analysis can reasonably be excluded" ⁽¹⁹⁾. Because indirect costs are believed to represent a relatively small fraction of the total cost of ESRD care and reliable estimates of indirect costs are not readily available, we estimated only direct medical costs for the Markov model.

The annual mortality rate in heart transplant patients who develop ESRD is higher than the average reported for all ESRD patients ^(2,21). We, therefore, assume an annual mortality rate of 12% for heart transplant patients who subsequently develop ESRD and undergo a renal transplant. The annual mortality rate for heart-transplant patients undergoing chronic dialysis was set as 24%.

The cost of care for patients after heart transplantation has steadily increased over the past decade, presumably because of a change in medical resource use and prices (e.g., increasing wage rates or use of new medical technologies). To account for these trends, we inflated costs based on the average of the overall CPI, as described before. To adjust dollars back to net present value in 1997 dollars, according to the Panel's recommendation, we discounted costs at a fixed annual rate of 3% ⁽²⁶⁾.

RESULTS

We first present the results of the analysis of ESRD risk and expenditures in Medicare beneficiaries. We then present the results of the Markov model.

ESRD Risk and Expenditures-Medicare Data. The mean patient age was 54.5 (±9.8) years (Table 2). Approximately 15% of transplant recipients were women and 7.4% were black. Disability was the primary reason for Medicare entitlement (85%).

Table 2: Patient characteristics^a

The mean 1-year survival rate among all cohorts was 82.6%, with higher survival rates among patients who underwent transplantation in recent years (Table 3). These rates are comparable to the average survival rates reported by the United Network for Organ Sharing ⁽²⁷⁾. For example, the average annual 1-year survival rate was 83.0% for the patients who underwent transplantation between 1987 and 1995.

Table 3: Cumulative patient survival in percent

Forty-one patients developed ESRD during the study period. Ten percent of these patients underwent kidney transplantation. Figure 1 shows the annual risk of ESRD for each cohort, unadjusted for baseline characteristics. The figure also shows the combined weighted annual risk of ESRD for all six cohorts.

Figure 1: The unadjusted annual risk of ESRD per survivor from time of heart transplant is shown for each patient cohort, grouped according to year of transplant. The widest line shows the pooled estimate of the unadjusted risk of ESRD across all patient cohorts. In the first year after transplantation, the ESRD risk is 0.37%. ESRD risk increases, to reach a high of 4.5% risk per year by the sixth posttransplant year.

The average cost of transplantation in the first 90 days equaled $106,250. Much of this cost was associated with the combined cost of organ procurement and the transplant procedure. The cost of transplantation then dropped during the period from day 91 to the end of the first year to $28,390. In the following 5 years, the average annual cost of care for transplant patients was relatively steady, at about $11,500 per year. These estimates do not include the cost of maintenance immunosuppressive drugs given to heart (and kidney) transplant patients.

Table 4 shows the results of the expenditure and cost regression analyses. The main difference between the cost and expenditure regressions is that between day 91 and the end of the first posttransplant year, the cost of ESRD care is approximately five times higher than expenditures. In contrast, the cost of ESRD is just 20% higher, on average, than expenditures beyond year 1. One hypothesis explaining the larger difference between cost and expenditures in year 1, compared with later years, may be the relatively smaller number of patients who have experienced ESRD, thereby affecting the precision of these cost and expenditure estimates. The average additional Medicare cost and expenditures associated with ESRD care were estimated to be $6,106 and $5,075, respectively, per heart transplant recipient for the 6 years for which data were available for this study (Table 5).

Table 4: Expenditures and total costs of Medicare beneficiaries after heart transplantation, with and without ESRD, by time posttransplant^a,b

Table 5: ESRD-related costs after heart transplantation

Reference-Case Markov Model.Table 5 shows the risk and cost of ESRD per transplant recipient by posttransplant year. Notably, the adjusted risk of ESRD increased substantially in the sixth posttransplant year to more than 4%. The cumulative 10-year cost of ESRD per transplant patient is $13,213. When the estimated risk of ESRD is increased to 2.8% (in our sensitivity analysis, the cumulative risk of ESRD equals the 9.8% reported by Myers et al. ^[12]), then the additional costs of ESRD increases to $17,352 per patient over a 10-year horizon.

The expected postprocedure cost of heart transplantation (i.e., from day 91-year 10), is estimated to be $142,668. Thus, the costs of ESRD care for transplanted patients account for 9.3% ($13,213 ÷ $142,668) of the total postprocedure costs of heart transplantation. Over a 20-year time horizon, the predicted costs of heart transplantation is $284,888, of which $22,514 is associated with the cost of ESRD care.

A 1991 National Institutes of Health-sponsored study of the potential cost-effectiveness of the artificial heart projected that heart transplantation was cost-effective compared with conventional medical therapy, equaling approximately $32,000 ($40,500 when adjusted by 4% per year to 1997) per year of life gained when considering adjustments for quality of life ⁽²⁸⁾. The study, however, did not explicitly include estimates of the added costs of ESRD care that were comparable to our study. We provide an estimate of the overall cost-effectiveness of heart transplantation in light of the added costs of ESRD found in our study. Patient life expectancy, unadjusted and adjusted for quality of life, is estimated to be 9.4 and 5.8 years, respectively. Our revised estimate of the cost-effectiveness of heart transplantation is $48,054 per quality-adjusted year of life gained. Even when using an extreme extrapolation of ESRD risk after year 6 (i.e., by carrying the last value of 4.9% per year forward for each of the subsequent 14 years up to 20 years), the cost-effectiveness of heart transplantation increases to $52,793.

DISCUSSION

This study is the first to document the incidence of ESRD after heart transplantation in a national sample. Moreover, we describe the costs associated with ESRD treatment in this population. In summary, the cumulative 6-year risk of developing ESRD after heart transplantation exceeds 6%. Although the risk of ESRD is just under 0.5% in the first posttransplant year, ESRD risk increases steadily with time, as was shown in other long-term studies ⁽¹²⁾. These findings, from a large, nationally representative sample of heart-transplant patients, provide evidence that ESRD risk is not lower among patients who underwent transplantation during or after 1989 compared with those treated before 1989.

Although most experts agree that there is an increased risk of ESRD in patients undergoing heart transplantation, some still suggest that they have taken measures to reduce ESRD risk, so that it should not be considered as significant an issue as in the past. This study of patients who underwent transplantation in more recent years affirms that ESRD risk has not been curtailed to the extent hoped for by suggested changes in practice. The consequences of ESRD for the health care costs of transplant recipients remain substantial. Our best estimate of the cumulative cost of treating ESRD per transplant recipient exceeds $13,000 over a 10-year horizon, accounting for more than 9% of the overall postprocedure costs of transplantation. These estimates of health care costs do not account for the serious consequences of developing ESRD for a patient's quality and quantity of life. Although mitigating the risk of ESRD after heart transplantation should save costs of dialysis and renal transplantation to health care systems, our analyses do not suggest that the cost of ESRD care precludes heart transplantation from being a cost-effective alternative to conventional medical therapy alone.

This study has several limitations. First, the sample was restricted to patients who were Medicare beneficiaries. However, the baseline characteristics of patients (see Table 2) are consistent with national averages, and patients' survival rates were similar to those reported for a national cohort. Second, our definition of ESRD was conservative in that patients had to have received dialysis for at least 90 days before being included in the analysis. If we had included all patients receiving any renal-replacement therapy after transplantation, the number of patients in the final analysis would have almost doubled, from 41 to 75. Finally, because more recent data were not available, we were restricted to follow-up of patients only through 1995. Perhaps trends of patients treated in 1996 and 1997 may reveal a changing pattern of ESRD risk or costs not found in the currently available cohorts. However, recommendations to alter immunosuppressive practices to reduce ESRD risk were published well before the beginning of our study period ^(1,4,6).

Finally, because our study does not include detailed information on drug use and other risk factors for ESRD, we do not settle the question of how to reduce the risk of ESRD after transplantation. Evaluating the mechanisms of ESRD will require analyses of databases with more detailed information on the use of cyclosporine and other drugs associated with nephrotoxicity, as well as on patients' clinical complications. Further studies using long-term registry data and patient follow-up as part of routine medical practice will provide more insights into these issues. Linking these clinical case records to the Medicare data used here on the occurrence and costs of ESRD would represent an even more powerful approach to assessing the clinical determinants of long-term health and economic outcomes in transplant patients. Such studies clearly seem to be worthwhile to help address the continuing problem of ESRD after transplantation.

Acknowledgments. The authors thank the Health Care Financing Administration for providing access to data. They also thank Lou Garrison, John Scandling, Clive Ward-Able, and an anonymous reviewer for invaluable comments on the paper.